R.03Research · Comparison

AI implementation firms compared: eight criteria, no ranking

Name: AI Implementation Firms Compared on Eight Criteria
Published: 2026-10-06

The firms that appear on the self-published best-firms lists for mid-market AI consulting, Praxis Consulting Company among them, observed on eight criteria a buyer can check in minutes. The page ranks nothing and names no winner. The table holds 23 firms, in alphabetical order. Each cell is a fact, the page it was read on, and the date; where a site could not be read, the cell says so with the status code. The publisher of this page is one of the firms in it, and that is stated before the first row.

Disclosure

Who published this, and what it does not do

01
Praxis Consulting Company publishes this page and is one of the firms listed. Its row is marked.
02
The page does not rank the firms and names no best firm. Rows are in alphabetical order.
03
No firm paid to appear, and no firm was asked to.
04
The author is the firm's founder, Evan Dang. No other person contributed observations.
05
Any related party in the list is named as such. EliteSEO Consulting, a separate practice with the same founder, is not in the list because no source list names it and it does not sell AI implementation.
06
Every cell carries the date it was observed and the page it was observed on. A cell that could not be observed says so, with the HTTP status that stopped it. A cell whose reading the stored record cannot support says so too, with the reason, and the reading is withdrawn rather than left standing.

Observed: 2026-10-09
First published: 2026-10-06
Last revised: 2026-10-09
Rows: 23 firms, 8 criteria, 20 assessed, 3 unassessed

Sheets of white paper laid out on a wooden table: an illustrative image for a register of observations set side by side.

How to read the table

Facts by criterion, and no winner

The page does not rank the firms. Rows are alphabetical, and nothing on the page says which firm is better. A publisher that is in its own list cannot honestly rank it, and a disclosure attached to a self-ranking is a disclaimer, not a remedy. So the page presents observable facts per firm, by criterion, and leaves the weighing to the reader.

There is no composite score, no stars and no out-of-ten. A composite score is where the bias hides: the weights are unfalsifiable, and whoever sets them decides the winner.

Each cell is one of three things. A conferred fact, which a third party states about the firm (criterion 2). A measured fact, which a published script computes the same way for every firm (criterion 4). Or a self-asserted fact, which the firm states about itself and a reader can check exists on the firm’s own page (the rest). The table says which is which, and the cell under each firm says where and when it was read.

The criteria

Eight things a buyer can check in minutes

A criterion is in only if a reader can verify it from a public page in under two minutes, it is either not self-asserted or labelled as self-asserted, and it changes which firm a mid-market buyer would call. The criteria were fixed, in writing, before any firm was looked at. Criterion 4 is the one measured rather than read, so it is the widest column in the table.

01
Named practitioners with a bio
Count of people on the firm's own site with a full name and at least one sentence of biography. A name with only a title does not count. Counts stop at 25.
Evidence: self-asserted, checkable
02
Third-party-conferred partner status
The firm's own claim, then the vendor's public partner directory. Only a vendor page that names the firm makes the cell read confirmed; otherwise the claim is recorded as self-asserted with the directory's status.
Evidence: conferred
03
Named clients, and whether engagement scope is stated
Count of distinct named client organisations on the firm's site, and whether at least one is paired with a description of what the engagement covered. Anonymised dollar results are recorded as stated, unverifiable.
Evidence: self-asserted, checkable
04
Sourcing rate on published statistics
One sampled page per firm, chosen by a fixed rule, measured by a published script: sentences with a percentage, a multiple or a magnitude number, and how many of those sentences link to a page outside the firm's own domain. A named source with no link counts as unsourced.
Evidence: measured
05
Original research published
A survey, benchmark, index or dataset authored by the firm with a stated method or sample: yes with the URL, or none found. A round-up of other people's figures with no method does not count.
Evidence: observed
06
Pricing model stated publicly
Fixed, hourly, retainer, subscription, outcome-linked, or not stated, as the firm's own pages put it. A from-figure counts as stated.
Evidence: self-asserted
07
Engagement shape stated publicly
Three yes-or-no observations from the firm's own pages: the first 30 days are described; who does the work is named as a role, team or person; the deliverable is named.
Evidence: self-asserted
08
Disclosure quality on the firm's own comparison pages
Whether the firm publishes a best-firms list that includes itself; the position it gives itself; whether the page plainly states that the publisher is one of the firms listed (a byline is not that); whether it names related parties among the entries. Applied to this page too.
Evidence: observed

Who is in the table

The inclusion rule, published

Every firm in the organic top 20 for the published query list on the stated date, plus every firm named on the lists the incumbents publish for the same intent. A firm that could not be assessed is listed as unassessed with the status code, never dropped.

Publishing the query list and the date is what lets a reader check that nobody was left out for being inconvenient. Naming real, larger firms is deliberate: a list of firms nobody has heard of, with the publisher among them, is an advertisement.

The query list

“best AI consulting firms for mid-market”
“back office automation consulting”
“data strategy consulting”
“AI implementation consulting firms”

The organic-top-20 half of the rule could not be executed on 2026-10-06 at zero cost and without a browser. Bing's HTML endpoint answered the four queries with HTTP 200 and result titles unrelated to the queries, and DuckDuckGo's HTML endpoint answered HTTP 202; both requests were abandoned there, with no change of headers or address. The Bing titles are kept in the site repository as the capture record. The query list stays published so the frame can be widened when a capture can be made.

The lists the frame was built from

Boldr AI: Open on getboldr.aihttps://getboldr.ai/insights/best-ai-consulting-firms-for-mid-market
Boldr AI: Open on getboldr.aihttps://getboldr.ai/insights/ai-back-office-transformation-consultancies-for-mid-market
Phos AI Labs: Open on phosailabs.comhttps://phosailabs.com/blog/best-ai-adoption-companies-mid-market

Every firm given its own entry or named as an example firm on those three pages is in the frame, plus the publishers of the pages, plus the two other firms the 2026-10-03 teardown examined (Velsof, whose page for this intent names no other firm, and Opinosis Analytics, whose site answered HTTP 202 on 2026-10-06 and HTTP 200 on 2026-10-08, when it was assessed), plus Praxis Consulting Company.

Listed as unassessed

BCG X: The one request to bcg.com/x answered HTTP 403, and bcgx.com did not connect. No second request was made. The one request made on 2026-10-08 answered HTTP 403 again.
Brainpool AI: The homepage, the blog index and the firm's own best-firms page each answered HTTP 403 on their one request on 2026-10-06, and the homepage answered 403 again on 2026-10-08. Other pages did answer on 2026-10-06, but the record keeps no text from them, so on 2026-10-09 every cell was set to unassessed rather than rest on readings that cannot be checked. Nothing was requested again.
Quantum Rise: The site answered HTTP 200 on every page read, but each served page is a JavaScript-only shell: on 2026-10-08 the pages read were within a few dozen characters of one another in length and, apart from the homepage, carried no word that did not also appear on every other page read (the record keeps each page's length, hash and that count). No cell could be read, and nothing is asserted about what the site says.

The table

23 firms, 8 criteria, alphabetical

Short values only. The page it was read on, the date, the qualifying note and, for criterion 2, the record of which vendor directories were tried are under each firm below the table; the firm name links there. Source pages are printed as addresses rather than hyperlinked, so that a page about 23 firms does not become a link directory; each firm’s own site is linked once, from its card. The table scrolls sideways on a narrow screen; the cards below it do not.

23 firms compared on 8 criteria, alphabetical, no ranking. Short values; sources and dates are in the firm cards that follow.
Firm	01Named people with a bio	02Vendor partner status	03Named clients, scope	04Statistics with a source	05Original research	06Pricing model	07Engagement shape	08Own list: disclosure
Auxis	13	claimed: UiPath partner tier (the page states two different tiers), Microsoft partner, AWS Partner Network (self-asserted)	14 named clients; scope: yes	1 of 12 quantified claims link a source	none found	not stated	first 30 days: nowho does the work: yesdeliverable: no	publishes: yesranks itself: unrankeddiscloses authorship: nodiscloses related parties: yes
BairesDev	0	claimed: AWS Advanced Tier, Google Cloud, Microsoft Cloud AI, Salesforce Consulting, Databricks Bronze, Snowflake Select, OpenAI Select, Anthropic Select and Fivetran partner marks and an ISO/IEC 27001:2022 mark (self-asserted)	25 or more named clients; scope: yes	0 quantified claims on the sampled page	yes	not stated	first 30 days: yeswho does the work: yesdeliverable: no	publishes: yesranks itself: unranked, listed firstdiscloses authorship: nodiscloses related parties: none apparent
BCG Xunassessed	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)
Boldr AIpublishes a list in the frame	0	none claimed	0 named clients; scope: n/a	0 quantified claims on the sampled page	none found	fixed; retainer	first 30 days: yeswho does the work: yesdeliverable: yes	publishes: yes (two pages)ranks itself: unranked, listed first on bothdiscloses authorship: nodiscloses related parties: none apparent
Brainpool AIunassessed	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)	unassessed (403)
DataArt	unassessed (no evidence retained)	claimed: AWS, Google Cloud, Microsoft, Salesforce, OpenAI and Anthropic partner marks (self-asserted)	15 named clients; scope: yes	0 of 6 quantified claims link a source	none found	not stated	first 30 days: yeswho does the work: yesdeliverable: yes	publishes: none found
DataRoot Labs	0	none claimed	8 named clients; scope: partly	0 quantified claims on the sampled page	yes	not stated	first 30 days: nowho does the work: yesdeliverable: yes	publishes: none found
Genpact	unassessed (no evidence retained)	unassessed (no evidence retained)	6 named clients; scope: yes	0 of 8 quantified claims link a source	yes	unassessed (no evidence retained)	first 30 days: nowho does the work: yesdeliverable: yes	publishes: none found
HatchWorks AI	1	claimed: Anthropic, OpenAI and Google Cloud partner marks (self-asserted)	25 or more named clients; scope: yes	0 quantified claims on the sampled page	none found	unassessed (no evidence retained)	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 1 of 9discloses authorship: nodiscloses related parties: none apparent
InData Labs	1	claimed: technology partnerships with AWS and Databricks (self-asserted)	8 named clients; scope: yes	0 quantified claims on the sampled page	none found	hourly; fixed	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 1 of 13discloses authorship: nodiscloses related parties: none apparent
ISHIR	6	claimed: Anthropic, Salesforce and several Microsoft partner marks (self-asserted)	25 or more named clients; scope: yes	0 of 1 quantified claims link a source	none found	hourly; fixed; retainer	first 30 days: nowho does the work: yesdeliverable: yes	unassessed (no evidence retained)
LeewayHertz	2	claimed: ISO/IEC 42001 and 27001 marks, SOC 2 Type II, HIPAA and GDPR statements (self-asserted). No AI vendor partner status is stated on the pages read.	10 named clients; scope: yes	0 quantified claims on the sampled page	none found	not stated	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 2discloses authorship: nodiscloses related parties: no
LOW/CODE Agency	1	claimed: OpenAI partner status; Make partner status (self-asserted)	25 or more named clients; scope: yes	0 of 3 quantified claims link a source	unassessed (no evidence retained)	retainer	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 2 of 14discloses authorship: nodiscloses related parties: no
Markovate	3	claimed: ISO 9001 and ISO/IEC 27001 marks (self-asserted). No AI vendor partner status is claimed on the pages read.	15 named clients; scope: yes	0 quantified claims on the sampled page	unassessed (no evidence retained)	hourly	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 1 of 10discloses authorship: nodiscloses related parties: none apparent
Opinosis Analytics	3	none claimed	16 named organisations; scope: yes	0 of 2 quantified claims link a source	none found	not stated	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yesranks itself: position 1 of 5discloses authorship: nodiscloses related parties: none apparent
Phos AI Labspublishes a list in the frame	1	claimed: OpenAI partner status; Anthropic partner status (self-asserted)	13 named clients; scope: yes	0 quantified claims on the sampled page	none found	fixed; retainer	first 30 days: yeswho does the work: yesdeliverable: yes	publishes: yesranks itself: position 1 of 7discloses authorship: nodiscloses related parties: no
Praxis Consulting Companypublisher of this page	1	none claimed	0 named clients; scope: n/a	0 of 4 quantified claims link a source	yes	not stated	first 30 days: nowho does the work: yesdeliverable: yes	publishes: yes (this page)ranks itself: unrankeddiscloses authorship: yesdiscloses related parties: yes
Quantum Riseunassessed	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)	unassessed (200, JavaScript-only shell)
RSM US	unassessed (200, leadership page loads names by script)	claimed: Microsoft Solutions Partner; Azure Expert MSP (self-asserted)	9 named clients; scope: yes	0 of 11 quantified claims link a source	yes	not stated	first 30 days: nowho does the work: yesdeliverable: yes	publishes: none found
SeidrLab	7	none claimed	6 named clients; scope: no	0 quantified claims on the sampled page	yes	fixed; retainer	first 30 days: yeswho does the work: yesdeliverable: yes	publishes: none found
Tenex	2	claimed: partner logos for Anthropic, OpenAI, LangChain, Braintrust, Lovable and Vercel (self-asserted)	4 named clients; scope: no	0 quantified claims on the sampled page	none found	subscription; outcome-linked	first 30 days: yeswho does the work: yesdeliverable: yes	publishes: none found
Velsof (Velocity Software Solutions)	0	claimed: CMMI Level 3; ISO 9001, 27001, 14001 and 14064-1 marks; PrestaShop partner tiers; NASSCOM membership (self-asserted)	25 or more named clients; scope: yes	0 of 13 quantified claims link a source	none found	retainer	first 30 days: yeswho does the work: yesdeliverable: yes	unassessed (no evidence retained)
Vention	9	unassessed (no evidence retained)	25 or more named clients; scope: yes	0 of 17 quantified claims link a source	yes	hourly	first 30 days: yeswho does the work: yesdeliverable: yes	unassessed (no evidence retained)

Every firm in this table was observed the same way on the same date, by the method under each criterion above. Where a cell reads unassessed, either the firm’s site answered the one plain request with the status shown and the observation stopped there, or the stored record keeps no evidence that would let the reading be checked; the note under the cell says which.

Each firm, cell by cell

The page, the date and the note behind every value

Auxis
Open on auxis.com
In the frame because: named on the Boldr AI back-office list.
01 Named people with a bio
13
Thirteen people on the leadership page with a full name, a title and at least one sentence of biography.
observed 2026-10-06
read on https://www.auxis.com/about/our-leadership-team/
02 Vendor partner status
claimed: UiPath partner tier (the page states two different tiers), Microsoft partner, AWS Partner Network (self-asserted; UiPath's partner page does not name the firm; the Microsoft and AWS directories are JavaScript-rendered, 200)
observed 2026-10-06
read on https://www.auxis.com/our-partnerships/
03 Named clients, scope
14 named clients; scope: yes
Fourteen organisations named on the clients page in attributed quotes and a logo; at least one is paired with a linked account of the engagement. Twenty-one further accounts on the work index describe the client by type only.
observed 2026-10-06
read on https://www.auxis.com/about/our-clients/
04 Statistics with a source
1 of 12 quantified claims link a source
First-article rule: the first article on the firm's blog index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.auxis.com/key-benefits-of-nearshoring-to-latin-america-for-the-midmarket/
05 Original research
none found
The resource centre's first item is a CFO survey attributed to another firm and discussed by this one; no firm-authored survey, benchmark, index or dataset was found.
observed 2026-10-06
read on https://www.auxis.com/resource-center/
06 Pricing model
not stated
The AI services page says cost varies and invites a consultation; the 727-URL sitemap has no pricing page.
observed 2026-10-06
read on https://www.auxis.com/digital-transformation-services/ai-business-solutions/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: no
The page lists phases with no time frame; the team is named by role and location; no deliverable is named on the pages read.
observed 2026-10-06
read on https://www.auxis.com/digital-transformation-services/ai-business-solutions/
08 Own list: disclosure
publishes: yes; ranks itself: unranked; discloses authorship: no; discloses related parties: yes
A list of finance and accounting outsourcing firms, not AI consulting, that names Auxis among seven firms in one sentence with no numbered order and then sets out what distinguishes it. A staff byline only; no sentence states that the publisher is one of the firms listed. The page and footer do state the firm's relationship to Grant Thornton.
observed 2026-10-06
read on https://www.auxis.com/top-finance-and-accounting-outsourcing-companies-globally/
BairesDev
Open on bairesdev.com
In the frame because: named as an example firm on the Boldr AI mid-market list.
01 Named people with a bio
0
The leadership page lists 34 people by full name and title only; no biography text is served. The about page names the two founders with no prior work, education or specialism.
observed 2026-10-06
read on https://www.bairesdev.com/about/leadership-team/
02 Vendor partner status
claimed: AWS Advanced Tier, Google Cloud, Microsoft Cloud AI, Salesforce Consulting, Databricks Bronze, Snowflake Select, OpenAI Select, Anthropic Select and Fivetran partner marks and an ISO/IEC 27001:2022 mark (self-asserted; the AWS, Google Cloud and Salesforce directory pages served under 100 characters of visible text and the Microsoft page 1,061 characters, all 200 and none naming the firm in visible text; the OpenAI partner page lists 71 names and not BairesDev; the Anthropic 'Powered by Claude' page lists 66 businesses and not BairesDev, and is a showcase of businesses that use Claude, not a partner directory)
Changed since 2026-10-06: the page now also shows Snowflake, OpenAI, Anthropic and Fivetran partner marks, the Microsoft mark reads 'Microsoft Cloud AI Partner', and the Microsoft 'Gold' wording recorded on 2026-10-06 is absent (the word Gold appears nowhere on the page). The Microsoft directory page holds the firm's name 3 times in its served HTML and never in its visible text. Field: observations-2026-10-08/bairesdev-c2-fields.json.
observed 2026-10-08
read on https://www.bairesdev.com/about/certifications-and-partnerships/
03 Named clients, scope
25 or more named clients; scope: yes
More than 25 identifiable client logos on the clients page; at least one named client is paired with a linked account of the engagement.
observed 2026-10-06
read on https://www.bairesdev.com/clients/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the featured post at the top of the blog index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.bairesdev.com/blog/how-to-train-an-llm-on-your-own-data/
05 Original research
yes
A quarterly developer survey. The Q3 2026 edition states it was fielded among 705 developers in more than 60 countries plus 41 enterprise CTOs, and offers the dataset for download.
observed 2026-10-06
read on https://www.bairesdev.com/blog/dev-barometer-q3-2026-devs-answering-for-code/
06 Pricing model
not stated
The page says pricing depends on stack, team size and scope and gives no figure; the 426-URL sitemap has no pricing page.
observed 2026-10-06
read on https://www.bairesdev.com/staff-augmentation/
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: no
The homepage process describes finalising specifications and onboarding a team within days; the team is named by role; no deliverable is named beyond progress reporting.
observed 2026-10-06
read on https://www.bairesdev.com/
08 Own list: disclosure
publishes: yes; ranks itself: unranked, listed first; discloses authorship: no; discloses related parties: none apparent
A vendor comparison table lists BairesDev first among nine unnumbered firms and profiles it first. No sentence states that the publisher is one of the firms listed.
observed 2026-10-06
read on https://www.bairesdev.com/blog/best-ai-development-companies/
BCG X
Open on bcg.com
In the frame because: named as an example firm on the Boldr AI mid-market list. The one request to bcg.com/x answered HTTP 403, and bcgx.com did not connect. No second request was made. The one request made on 2026-10-08 answered HTTP 403 again.
01 Named people with a bio
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
02 Vendor partner status
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
03 Named clients, scope
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
04 Statistics with a source
unassessed (403)
Not measured: the protocol allows no second request after a refusal.
observed 2026-10-06
read on https://www.bcg.com/x
05 Original research
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
06 Pricing model
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
07 Engagement shape
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
08 Own list: disclosure
unassessed (403)
observed 2026-10-06
read on https://www.bcg.com/x
Boldr AI
Open on getboldr.ai
publishes a list in the frame. In the frame because: publishes two of the three list pages the frame was built from.
01 Named people with a bio
0
The about page names no individual. Insight articles carry a byline with a name and title and no biography, and a byline is not a bio.
observed 2026-10-06
read on https://getboldr.ai/about
02 Vendor partner status
none claimed
No partner badge on the nine pages read; two cloud providers are mentioned as hosting environments only.
observed 2026-10-06
read on https://getboldr.ai/
03 Named clients, scope
0 named clients; scope: n/a
The results page describes six engagements anonymised by sector under NDA, with dollar figures: stated, unverifiable. No client organisation is named.
observed 2026-10-06
read on https://getboldr.ai/results
04 Statistics with a source
0 quantified claims on the sampled page
List-page rule. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://getboldr.ai/insights/best-ai-consulting-firms-for-mid-market
05 Original research
none found
The insights index lists buyer's guides, playbooks, an ebook and a technical comparison with latency figures but no stated test method; no survey, benchmark, index or dataset.
observed 2026-10-06
read on https://getboldr.ai/insights
06 Pricing model
fixed; retainer
The sprint page states a fixed price set by scope from a published figure, followed by a fixed-price deployment and a monthly operating engagement.
observed 2026-10-06
read on https://getboldr.ai/value-discovery-sprint
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The sprint page lays out weeks one to four and a deploy-or-stop decision, names the team by role and location, and names the deliverables.
observed 2026-10-06
read on https://getboldr.ai/value-discovery-sprint
08 Own list: disclosure
publishes: yes (two pages); ranks itself: unranked, listed first on both; discloses authorship: no; discloses related parties: none apparent
Both list pages carry a staff byline, number no entries, and profile Boldr AI first; the second page's summary describes Boldr AI in superlative terms. Neither page states that the publisher is one of the firms listed. The second page carries a citations block with one sourced figure. Re-read on 2026-10-08: the first page is unchanged in the respects recorded; the second page no longer answers at its 2026-10-06 address (HTTP 404) and is served, with the same title, byline and update date, at https://getboldr.ai/insights/ai-back-office-transformation-consultancies-for-mid-market (200).
observed 2026-10-08
read on https://getboldr.ai/insights/best-ai-consulting-firms-for-mid-market
Brainpool AI
Open on brainpool.ai
In the frame because: named on the Phos AI Labs list. The homepage, the blog index and the firm's own best-firms page each answered HTTP 403 on their one request on 2026-10-06, and the homepage answered 403 again on 2026-10-08. Other pages did answer on 2026-10-06, but the record keeps no text from them, so on 2026-10-09 every cell was set to unassessed rather than rest on readings that cannot be checked. Nothing was requested again.
01 Named people with a bio
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
02 Vendor partner status
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
03 Named clients, scope
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
04 Statistics with a source
unassessed (403)
The blog index and the homepage both answered 403 on their one request, so neither the first-article rule nor the homepage fallback could be applied. The homepage was requested once more on 2026-10-08 and answered 403 again; nothing further was requested.
observed 2026-10-06
read on https://brainpool.ai/blog
05 Original research
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
06 Pricing model
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
07 Engagement shape
unassessed (403)
observed 2026-10-06
read on https://brainpool.ai/
08 Own list: disclosure
unassessed (403)
The sitemap lists a best-AI-consulting-firms page on the firm's own site; it answered 403 on its one request, so whether it includes the firm and what it discloses could not be observed.
observed 2026-10-06
read on https://brainpool.ai/best-ai-consulting-firms
DataArt
Open on dataart.com
In the frame because: named as an example firm on the Boldr AI mid-market list.
01 Named people with a bio
unassessed (no evidence retained)
The team index and its profile pages were read on 2026-10-06, but the record keeps no name or sentence from them, and the 2026-10-08 re-observation could only confirm the index still answered 200. The count was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://www.dataart.com/company/team
02 Vendor partner status
claimed: AWS, Google Cloud, Microsoft, Salesforce, OpenAI and Anthropic partner marks (self-asserted; every directory tried serves a script-rendered shell that does not name the firm, 200)
observed 2026-10-06
read on https://www.dataart.com/
03 Named clients, scope
15 named clients; scope: yes
Fifteen organisations named on the homepage and clients page; several are paired with a one-line account of the engagement.
observed 2026-10-06
read on https://www.dataart.com/clients
04 Statistics with a source
0 of 6 quantified claims link a source
First-article rule: the first blog post on the index, after a landing-page card that is not an article. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.dataart.com/blog/ai-in-software-delivery-timelines-pricing
05 Original research
none found
A trends report states no survey, sample or method; the third-party analyst reports on the site are licensed, not authored; a client-satisfaction score is a metric, not a study.
observed 2026-10-06
read on https://www.dataart.com/blog
06 Pricing model
not stated
The AI consulting page gives an engagement timeline with no price or pricing model.
observed 2026-10-06
read on https://www.dataart.com/ai-consulting
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
A one-to-two-week assessment, a three-month strategy phase and an implementation phase are described with named deliverables for each; the practice lead is named by role.
observed 2026-10-06
read on https://www.dataart.com/ai-consulting
08 Own list: disclosure
publishes: none found
No best-firms page among the 3,272 URLs in the sitemap; newsroom items report third-party rankings of the firm, which are not lists it publishes.
observed 2026-10-06
read on https://www.dataart.com/blog
DataRoot Labs
Open on datarootlabs.com
In the frame because: named on the Boldr AI mid-market list.
01 Named people with a bio
0
One person is named on the site, with a title and no biographical sentence; the sitemap lists no team or about page.
observed 2026-10-06
read on https://datarootlabs.com/
02 Vendor partner status
none claimed
A technology ecosystem section names five vendors and links to their documentation; no partner tier or badge is stated.
observed 2026-10-06
read on https://datarootlabs.com/
03 Named clients, scope
8 named clients; scope: partly
Eight organisations named in a partnership section and logo row; two cards describe the client's product rather than what the engagement covered.
observed 2026-10-06
read on https://datarootlabs.com/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the first item on the blog index, a report. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://datarootlabs.com/blog/state-of-neurotech-2026
05 Original research
yes
A firm-authored guide to 100 companies across seven areas of neurotechnology, with a stated method: company materials, papers, trial registries and regulatory records reviewed to a stated date, with sources on each profile.
observed 2026-10-06
read on https://datarootlabs.com/blog/state-of-neurotech-2026
06 Pricing model
not stated
The partnership questions say scope, budget and timeframe are agreed per phase; no model or figure is stated.
observed 2026-10-06
read on https://datarootlabs.com/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
Three engagement models and a first phase are described without a duration; the researchers and engineers proposed for the work are named by role; a baseline and feasibility test are the named first deliverables.
observed 2026-10-06
read on https://datarootlabs.com/
08 Own list: disclosure
publishes: none found
No best-firms page in the sitemap; two posts report third-party rankings of the firm and were not fetched.
observed 2026-10-06
read on https://datarootlabs.com/blog
Genpact
Open on genpact.com
In the frame because: named on the Boldr AI back-office list.
01 Named people with a bio
unassessed (no evidence retained)
The leadership page and its profile pages were read on 2026-10-06, but the record keeps no name or sentence from them, and the 2026-10-08 re-observation could only confirm the page still answered 200. The count was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://www.genpact.com/about-us/leadership
02 Vendor partner status
unassessed (no evidence retained)
The partners page was read on 2026-10-06, but the record keeps no text from it, and the 2026-10-08 re-observation could not match every recorded partner string on the page. The reading was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://www.genpact.com/partners
03 Named clients, scope
6 named clients; scope: yes
Six organisations named in work cards on the homepage, the work index and the AI services page; the card titles state the engagement for two of them.
observed 2026-10-06
read on https://www.genpact.com/
04 Statistics with a source
0 of 8 quantified claims link a source
First-article rule: the first of three insight links in the served HTML, the index grid being script-rendered. Part of the page's content is embedded component data, which the script's text extraction counts. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.genpact.com/insight/how-four-enterprise-debts-will-make-or-break-your-ai-future
05 Original research
yes
A firm-authored research report with a stated sample of more than 500 senior executives, with the full report linked as a PDF.
observed 2026-10-06
read on https://www.genpact.com/insight/how-to-scale-ai-for-enterprise-value
06 Pricing model
unassessed (no evidence retained)
The AI services, studio, partners and about pages were read on 2026-10-06, but the record keeps no text from them, and the 2026-10-08 re-observation could only confirm the page still answered 200. The reading was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://www.genpact.com/services/data-and-ai/artificial-intelligence
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
Workshop formats and a weeks-not-quarters prototyping claim are stated without a first-30-days description; the team is named by role; mock-ups and prototypes are named deliverables.
observed 2026-10-06
read on https://www.genpact.com/solutions/ai-value-studio
08 Own list: disclosure
publishes: none found
No best-firms page in the sitemap; several insight items report third-party analyst rankings of the firm.
observed 2026-10-06
read on https://www.genpact.com/insight
HatchWorks AI
Open on hatchworks.com
In the frame because: named on both Boldr AI lists.
01 Named people with a bio
1
The founder is the one person with a biographical sentence; thirteen leaders are listed with a name and title only.
observed 2026-10-06
read on https://hatchworks.com/about-us/
02 Vendor partner status
claimed: Anthropic, OpenAI and Google Cloud partner marks (self-asserted; Anthropic's directory URLs answered 404, OpenAI's partner page names no firms, Google Cloud's directory is JavaScript-rendered)
observed 2026-10-06
read on https://hatchworks.com/about-us/
03 Named clients, scope
25 or more named clients; scope: yes
More than 25 identifiable client logos on the work index; several are paired with a linked account of the engagement.
observed 2026-10-06
read on https://hatchworks.com/case-studies/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the first post on the blog index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://hatchworks.com/blog/gen-ai/openai-just-launched-dots/
05 Original research
none found
A state-of-AI report is offered as a gated PDF; its landing page states no survey, sample or method, so it cannot be classed from the page.
observed 2026-10-06
read on https://hatchworks.com/state-of-ai-2026-mid-year-reality-check/
06 Pricing model
unassessed (no evidence retained)
Three engagement pages were read on 2026-10-06, but the record keeps no text from them, and the 2026-10-08 re-observation could only confirm the page still answered 200. The reading was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://hatchworks.com/how-we-engage/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
Roles and pod composition are named and deliverables are listed; no first-30-days description on the three engagement pages read.
observed 2026-10-06
read on https://hatchworks.com/how-we-engage/
08 Own list: disclosure
publishes: yes; ranks itself: position 1 of 9; discloses authorship: no; discloses related parties: none apparent
A nearshore software companies list places the firm first. No sentence states that the publisher is one of the firms listed.
observed 2026-10-06
read on https://hatchworks.com/blog/nearshore-development/top-nearshore-software-companies/
InData Labs
Open on indatalabs.com
In the frame because: named as an example firm on the Boldr AI mid-market list.
01 Named people with a bio
1
The founder is named with a one-sentence background; one further person is named with a title only. No team page in the sitemap.
observed 2026-10-06
read on https://indatalabs.com/about
02 Vendor partner status
claimed: technology partnerships with AWS and Databricks (self-asserted; the AWS directory is a JavaScript shell and the Databricks directory's served HTML does not name the firm)
observed 2026-10-06
read on https://indatalabs.com/about
03 Named clients, scope
8 named clients; scope: yes
Eight organisations named with attributed quotes on the about page; several are paired with a one-line account of what was built.
observed 2026-10-06
read on https://indatalabs.com/about
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the first post on the blog index. The firm's own lists are not for the mid-market intent, so the list-page rule did not apply. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://indatalabs.com/blog/ai-analytics-solutions-for-business
05 Original research
none found
The research-papers section lists white papers and infographics presented as statistics round-ups; the adoption page states no survey, sample or method.
observed 2026-10-06
read on https://indatalabs.com/research-papers/state-of-ai-adoption
06 Pricing model
hourly; fixed
Time-and-materials, fixed-price and dedicated-team models are named; no figure is stated. A pricing label in the navigation carries no link.
observed 2026-10-06
read on https://indatalabs.com/how-we-work
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
A five-phase process is described without a first-30-days span; the team is named by role; a roadmap and specifications are named deliverables.
observed 2026-10-06
read on https://indatalabs.com/how-we-work
08 Own list: disclosure
publishes: yes; ranks itself: position 1 of 13; discloses authorship: no; discloses related parties: none apparent
The list's title names the publisher and places it first, and a second list on the site does the same; neither carries a sentence stating that the publisher is one of the firms listed.
observed 2026-10-06
read on https://indatalabs.com/blog/top-ai-consulting-firms
ISHIR
Open on ishir.com
In the frame because: named on the Phos AI Labs list.
01 Named people with a bio
6
Six people on the team page with a name, a role and at least one sentence of prior work or specialism.
observed 2026-10-06
read on https://www.ishir.com/team.htm
02 Vendor partner status
claimed: Anthropic, Salesforce and several Microsoft partner marks (self-asserted; Anthropic's directory URLs answered 404, the Salesforce and Microsoft directories are JavaScript shells)
observed 2026-10-06
read on https://www.ishir.com/team.htm
03 Named clients, scope
25 or more named clients; scope: yes
More than 25 identifiable client logos on the clients page; the success-stories page pairs named clients with an account of the engagement.
observed 2026-10-06
read on https://www.ishir.com/clients.htm
04 Statistics with a source
0 of 1 quantified claims link a source
First-article rule: the latest article on the insights index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.ishir.com/blog/345184/your-data-warehouse-was-designed-for-humans-what-changes-when-ai-becomes-the-main-consumer.htm
05 Original research
none found
White papers on outsourcing, quality and migration; none is a survey, benchmark, index or dataset with a stated method.
observed 2026-10-06
read on https://www.ishir.com/whitepaper-htm.htm
06 Pricing model
hourly; fixed; retainer
Five engagement models are described by name, including time-and-materials, fixed bid and a retainer; no figure is stated.
observed 2026-10-06
read on https://www.ishir.com/outsourcing-engagement-models.htm
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
An agile pod is described and workshops and assessments are named as deliverables; an initial product period is mentioned with no duration.
observed 2026-10-06
read on https://www.ishir.com/agile-team-pods.htm
08 Own list: disclosure
unassessed (no evidence retained)
The sitemaps were scanned for best-firms slugs on 2026-10-06, but the scan kept neither the URL list nor a control showing it could have found such a page, and the record carries no page URL for the cell. The reading was withdrawn on 2026-10-09.
observed 2026-10-06
read on https://www.ishir.com/insights.htm
LeewayHertz
Open on leewayhertz.com
In the frame because: named on the Boldr AI mid-market list.
01 Named people with a bio
2
Two executives on the about page with a role and a biography sentence; no other named practitioner on the pages read.
observed 2026-10-06
read on https://www.leewayhertz.com/about-us/
02 Vendor partner status
claimed: ISO/IEC 42001 and 27001 marks, SOC 2 Type II, HIPAA and GDPR statements (self-asserted; no vendor partner directory applies). No AI vendor partner status is stated on the pages read.
observed 2026-10-06
read on https://www.leewayhertz.com/about-us/
03 Named clients, scope
10 named clients; scope: yes
Ten organisations named across the work index and the about page; the work index pairs each named client with an account of what was built.
observed 2026-10-06
read on https://www.leewayhertz.com/case-studies/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the featured item at the top of the insights index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.leewayhertz.com/generative-ai-use-cases-and-applications/
05 Original research
none found
The insights are explainers; the closest candidate is a long-form overview citing third parties, with no firm-authored sample or method.
observed 2026-10-06
read on https://www.leewayhertz.com/current-state-of-generative-ai/
06 Pricing model
not stated
Three engagement models are named with no pricing basis; a cost estimate follows a consultation.
observed 2026-10-06
read on https://www.leewayhertz.com/ai-consulting-services-company/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
The consultants are named by role and an implementation roadmap and technical designs are named deliverables; the four-step start sequence carries no durations.
observed 2026-10-06
read on https://www.leewayhertz.com/ai-consulting-services-company/
08 Own list: disclosure
publishes: yes; ranks itself: position 2; discloses authorship: no; discloses related parties: no
A digital transformation companies list places another firm first and LeewayHertz second. The list page does not state that the publisher is one of the firms listed, and does not state a relationship between the two firms; the homepage separately states one.
observed 2026-10-06
read on https://www.leewayhertz.com/digital-transformation-companies/
LOW/CODE Agency
Open on lowcode.agency
In the frame because: named on the Phos AI Labs list.
01 Named people with a bio
1
The founder is named with a role and founding year; no prior work or education is stated. Twenty-eight further people appear with a first name and job title only, which is not a bio under the protocol.
observed 2026-10-06
read on https://www.lowcode.agency/about
02 Vendor partner status
claimed: OpenAI partner status; Make partner status (self-asserted; OpenAI's partner page, 200, names other firms in its served HTML and not this one; Make's directory answered 403)
observed 2026-10-06
read on https://www.lowcode.agency/blog/low-code-agency-openai-select-partner
03 Named clients, scope
25 or more named clients; scope: yes
Twelve organisations in the homepage logo strip and further organisations on the work index, where each entry carries an engagement description.
observed 2026-10-06
read on https://www.lowcode.agency/case-studies
04 Statistics with a source
0 of 3 quantified claims link a source
First-article rule: the first post under recent articles on the blog index. One of the three matches is an inline style fragment the percentage rule caught; recorded as the script reported it. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.lowcode.agency/blog/low-code-agency-openai-select-partner
05 Original research
unassessed (no evidence retained)
The blog index and the sitemap were scanned on 2026-10-06, but the record keeps neither the URL list nor a control showing the scan could have found a research item, and the 2026-10-08 re-observation could only confirm the index still answered 200. The reading was withdrawn on 2026-10-09.
observed 2026-10-06
read on https://www.lowcode.agency/blog
06 Pricing model
retainer
The AI consulting page lists a monthly advisory retainer alongside fixed-length workshop and strategy phases; no figure is stated.
observed 2026-10-06
read on https://www.lowcode.agency/services/ai-app-development/ai-consulting
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
The process page describes four phases with no day or week count for the first; the team is named by role; wireframes, flows and sprints are named as deliverables.
observed 2026-10-06
read on https://www.lowcode.agency/process
08 Own list: disclosure
publishes: yes; ranks itself: position 2 of 14; discloses authorship: no; discloses related parties: no
Entry 1 is Phos AI Labs and entry 2 is LOW/CODE Agency, written partly in the first person. No sentence states that the publisher is one of the firms listed. The about pages of both firms name the same person as founder, which the list does not state. On two other lists on the site the firm gives itself position 1.
observed 2026-10-06
read on https://www.lowcode.agency/blog/best-ai-agent-development-companies
Markovate
Open on markovate.com
In the frame because: named on the Boldr AI mid-market list.
01 Named people with a bio
3
Three team leaders with a role and at least one sentence of prior work or specialism. Two board advisers also carry bios and are not counted as practitioners.
observed 2026-10-06
read on https://markovate.com/about-us/
02 Vendor partner status
claimed: ISO 9001 and ISO/IEC 27001 marks (self-asserted; no vendor partner directory applies to an ISO mark, so not checked). No AI vendor partner status is claimed on the pages read.
observed 2026-10-06
read on https://markovate.com/about-us/
03 Named clients, scope
15 named clients; scope: yes
Thirteen organisations on the work index, each with an engagement description, plus two named in homepage quotes. The homepage logo strip's images carry no identifying alt text and are not counted.
observed 2026-10-06
read on https://markovate.com/work/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the first dated article in the blog stream. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://markovate.com/ai-in-preconstruction/
05 Original research
unassessed (no evidence retained)
The resources page and the blog index were read on 2026-10-06, but the record keeps no text from them, and the 2026-10-08 re-observation could only confirm the page still answered 200. The reading was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://markovate.com/resources/
06 Pricing model
hourly
An hourly rate band appears beside the firm's own entry on its list page; the services page states no price.
observed 2026-10-06
read on https://markovate.com/top-ai-consulting-companies/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
Five phases are listed with week counts for a pilot but no first-30-days description; the team is named as in-house consultants and developers; a roadmap and proofs of concept are named as deliverables.
observed 2026-10-06
read on https://markovate.com/ai-consulting-services/
08 Own list: disclosure
publishes: yes; ranks itself: position 1 of 10; discloses authorship: no; discloses related parties: none apparent
Entry 1 is Markovate, written in the third person. No byline appears in the served HTML and no sentence states that the publisher is one of the firms listed.
observed 2026-10-06
read on https://markovate.com/top-ai-consulting-companies/
Opinosis Analytics
Open on opinosis-analytics.com
In the frame because: examined in the 2026-10-03 teardown as a firm ranking for the same queries.
01 Named people with a bio
3
Three people are listed under 'Meet Our Key Team Members' on the about page, each with a full name, a role line and a first biography sentence naming education or specialism: the founder and two other people. Authors of quoted client comments are not counted. Field: observations-2026-10-08/opinosis-analytics-fields.json (c1). On 2026-10-06 the site answered HTTP 202 with a captcha redirect and nothing was observed.
observed 2026-10-08
read on https://www.opinosis-analytics.com/about-opinosis/
02 Vendor partner status
none claimed
No vendor-tier or credential-type term appears in the visible or attribute text of the homepage or the about page: a scan for partner tiers, resellers, badges, ISO marks and the names AWS, Google Cloud, Microsoft, Snowflake, Databricks, Salesforce, OpenAI and Anthropic returned zero hits, and each of the 5 uses of the word partner is a verb or the phrase 'trusted partner'. The about page does state a trade-publication recognition and a business-accelerator alumnus line ('Recognized by CIOReview as one of the top AI consulting companies and a Goldman Sachs business accelerator alumnus'); neither is a vendor partner status. Field: observations-2026-10-08/opinosis-analytics-fields.json (c2).
observed 2026-10-08
read on https://www.opinosis-analytics.com/about-opinosis/
03 Named clients, scope
16 named organisations; scope: yes
Counted on the about page, once each: seven in the sentence beginning 'Organizations we've worked with include' (McKesson, McMaster-Carr, 3M, University of Utah, IBL Group Mauritius, Nuclear Regulatory Commission, Department of Energy); seven more in the attribution lines under client quotes (BrightMind, 113 Industries, Ferrero, University of Sydney, Centeva, Trademarkia, Precipice Research; the IBL quote repeats an organisation already counted); and two more in the founder's biography sentence (eBay, GitHub). Scope is stated for at least one: the IBL quote says 'We now have a very detailed document of all our processes.' The page also states an anonymised aggregate savings result (the sentence is stored in the field): stated, unverifiable. Field: observations-2026-10-08/opinosis-analytics-fields.json (c3).
observed 2026-10-08
read on https://www.opinosis-analytics.com/about-opinosis/
04 Statistics with a source
0 of 2 quantified claims link a source
List-page rule: the firm's own page for the mid-market AI consulting intent. Script: scripts/comparison/sourcing_rate.py. The two claim sentences are 'Estimated 60,000+ hours of annual productivity gains before implementation' and 'Determined that only 29 percent were realistically solvable with AI'. Output saved in observations-2026-10-08/sourcing-rate-opinosis-list.json.
observed 2026-10-08
read on https://www.opinosis-analytics.com/blog/best-ai-strategy-consulting-for-mid-market-companies-and-enterprises/
05 Original research
none found
The publications page lists 19 citations dated 2005 to 2022: one book, one dissertation and academic or workshop papers. None is a survey, benchmark, index or dataset run by the firm with a stated sample; the one description that uses the word survey is of a 2011 literature review of summarisation techniques. The readiness assessment tool page says 'Data used for research will always be anonymized.' and contains none of the terms survey, benchmark, index, dataset or respondents. Field: observations-2026-10-08/opinosis-analytics-fields.json (c5).
observed 2026-10-08
read on https://www.opinosis-analytics.com/publications/
06 Pricing model
not stated
The article gives market rates for consultants in general (10 sentences carry a dollar figure, none of them names the firm) and invites a quote request ('For a quick quote on how much you would spend with Opinosis Analytics'). It says only that at the firm 'each phase has its own fee structure', with no model and no figure. Field: observations-2026-10-08/opinosis-analytics-fields.json (c6).
observed 2026-10-08
read on https://www.opinosis-analytics.com/blog/machine-learning-consulting-rates/
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
The readiness page lists six components (infrastructure evaluation, business goals alignment, AI opportunity mapping, data maturity assessment, strategic roadmap, implementation support) and names a 'detailed, tailored roadmap' as the output. Its only time expressions are the form reply window ('24-48 hours') and a prompt for the client's '60-90 day goals', so no first 30 days is described. Who does the work: the page says 'expert-led' and 'experienced AI consultants'; the about page lists three people with roles (c1). Field: observations-2026-10-08/opinosis-analytics-fields.json (c7).
observed 2026-10-08
read on https://www.opinosis-analytics.com/ai-readiness-assessment/
08 Own list: disclosure
publishes: yes; ranks itself: position 1 of 5; discloses authorship: no; discloses related parties: none apparent
A list numbered #1 to #5 with the firm first, then Accenture, Deloitte, McKinsey & Company and Boston Consulting Group (BCG X). The byline reads 'By Kavita Ganesan' and an author box at the foot reads 'Kavita Ganesan is a Chief AI Strategist and Architect, and the founder of Opinosis Analytics'. A byline or author box is not a disclosure under this criterion, which asks for a plain statement that the publisher is one of the firms listed; a scan for disclosure phrases found none. The sitemap page links 23 pages whose address starts best- or top- (some are lists of books, speakers or individuals), this one included; only this one was read. Field: observations-2026-10-08/opinosis-analytics-fields.json (c8).
observed 2026-10-08
read on https://www.opinosis-analytics.com/blog/best-ai-strategy-consulting-for-mid-market-companies-and-enterprises/
Phos AI Labs
Open on phosailabs.com
publishes a list in the frame. In the frame because: publishes one of the three list pages the frame was built from.
01 Named people with a bio
1
The founder is named with a role and an attributed quote; no prior work or education is stated. The authors page lists the founder and a team byline and no other named person.
observed 2026-10-06
read on https://phosailabs.com/about
02 Vendor partner status
claimed: OpenAI partner status; Anthropic partner status (self-asserted; OpenAI's partner page, 200, names other firms in its served HTML and not this one; Anthropic's partner page redirects to a page with no partner names)
observed 2026-10-06
read on https://phosailabs.com/
03 Named clients, scope
13 named clients; scope: yes
Eight organisations in the logo strip and five more on the work index, where each entry carries an engagement description; one anonymised entry is not counted. LOW/CODE Agency, entry 2 on the firm's own list, is one of the five.
observed 2026-10-06
read on https://phosailabs.com/case-studies
04 Statistics with a source
0 quantified claims on the sampled page
List-page rule. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://phosailabs.com/blog/best-ai-adoption-companies-mid-market
05 Original research
none found
The closest candidate states that its benchmarks are estimates drawn from multiple industry surveys, which is a round-up of others' figures with no firm-authored sample or method.
observed 2026-10-06
read on https://phosailabs.com/blog/ai-adoption-rate-benchmarks-by-industry
06 Pricing model
fixed; retainer
A fixed-fee audit and a monthly engagement are both priced from a stated figure on the AI consulting page.
observed 2026-10-06
read on https://phosailabs.com/ai-consulting
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The page describes the first thirty days and a three-to-six-week audit, names the consulting and engineering teams, and names nine deliverables.
observed 2026-10-06
read on https://phosailabs.com/ai-consulting
08 Own list: disclosure
publishes: yes; ranks itself: position 1 of 7; discloses authorship: no; discloses related parties: no
Entry 1 is Phos AI Labs, written in the first person; entry 2 is LOW/CODE Agency. The page states that no firm paid to appear. It does not state that the publisher is one of the firms listed. Four facts about entry 2 are observable elsewhere and not stated on the page: the about pages of both firms name the same person as founder; the list page's footer credit reads Crafted by LowCode Agency; the Phos homepage carries a quote attributed to that firm's founder; and LOW/CODE Agency appears on Phos's work index as a client.
observed 2026-10-06
read on https://phosailabs.com/blog/best-ai-adoption-companies-mid-market
Praxis Consulting Company
Open on praxisconsultingco.com
publisher of this page. In the frame because: it is the publisher; named on no incumbent list.
01 Named people with a bio
1
The founder, Evan Dang, is the only person named with a biography. The same page states that the firm draws on a team of 20 senior practitioners and says: Praxis does not name the 20 or publish their bios, so the record you can check is the founder's.
observed 2026-10-06
read on https://praxisconsultingco.com/about
02 Vendor partner status
none claimed
No vendor partner status is claimed anywhere on the site (home, services, about read).
observed 2026-10-06
read on https://praxisconsultingco.com/
03 Named clients, scope
0 named clients; scope: n/a
The how-we-work page states that the firm has no logo wall and no quoted client and will not invent one.
observed 2026-10-06
read on https://praxisconsultingco.com/how-we-work
04 Statistics with a source
0 of 4 quantified claims link a source
List-page rule: this page is the firm's list for the intent. The four matches are sentences inside other firms' cells: a sitemap URL count, a count of how many times a directory page holds a firm's name, a character count of a directory page, and a quoted claim sentence. None carries a link, and the rule does not exempt the publisher. Measured on the built page before publication; running the script on the live URL reproduces it. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-09
read on https://praxisconsultingco.com/research/ai-implementation-firms-compared
05 Original research
yes
A source-verification dataset of nine circulating AI savings figures, with a stated five-step method, a retrieval date and a changelog. Nine is a small sample and the page says so.
observed 2026-10-06
read on https://praxisconsultingco.com/research/ai-savings-statistics
06 Pricing model
not stated
The page states that prices are not published on the site, on purpose, and that a written quote follows a five-step scope.
observed 2026-10-06
read on https://praxisconsultingco.com/how-we-work/pricing-and-scoping
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
The first phase is described as a defined-scope diagnostic that produces the plan (engagement-models) and the deliverables are listed in order (what-you-receive), but no 30-day span is stated. The founder is named as leading every engagement.
observed 2026-10-06
read on https://praxisconsultingco.com/how-we-work
08 Own list: disclosure
publishes: yes (this page); ranks itself: unranked; discloses authorship: yes; discloses related parties: yes
The disclosure block at the top of this page is the statement being assessed.
observed 2026-10-06
read on https://praxisconsultingco.com/research/ai-implementation-firms-compared
Quantum Rise
Open on quantumrise.com
In the frame because: named on the Phos AI Labs list. The site answered HTTP 200 on every page read, but each served page is a JavaScript-only shell: on 2026-10-08 the pages read were within a few dozen characters of one another in length and, apart from the homepage, carried no word that did not also appear on every other page read (the record keeps each page's length, hash and that count). No cell could be read, and nothing is asserted about what the site says.
01 Named people with a bio
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/team
02 Vendor partner status
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/about-us
03 Named clients, scope
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/proof
04 Statistics with a source
unassessed (200, JavaScript-only shell)
Homepage rule, because the insights index is the same shell. The script ran on the served homepage body on both dates and found no claim sentence, recorded with the body's hash, but the body is the shell, so the figure is a reading of the shell and is not entered as a value.
observed 2026-10-06
read on https://quantumrise.com/
05 Original research
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/insights
06 Pricing model
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/what-we-do
07 Engagement shape
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/what-we-do
08 Own list: disclosure
unassessed (200, JavaScript-only shell)
observed 2026-10-06
read on https://quantumrise.com/sitemap
RSM US
Open on rsmus.com
In the frame because: named on the Boldr AI back-office list.
01 Named people with a bio
unassessed (200, leadership page loads names by script)
The served HTML carries the section headings and load-more controls but no name or biography; the homepage fallback shows none either.
observed 2026-10-06
read on https://rsmus.com/about/leadership.html
02 Vendor partner status
claimed: Microsoft Solutions Partner; Azure Expert MSP (self-asserted; Microsoft partner directory is JavaScript-rendered, 200)
observed 2026-10-06
read on https://rsmus.com/technologies/microsoft.html
03 Named clients, scope
9 named clients; scope: yes
Nine client organisations named across the digital transformation and AI services pages, each with a linked account of the engagement.
observed 2026-10-06
read on https://rsmus.com/services/digital-transformation.html
04 Statistics with a source
0 of 11 quantified claims link a source
First-article rule: the Middle Market Business Index page is the first item on the insights index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://rsmus.com/middle-market/mmbi.html
05 Original research
yes
The RSM US Middle Market Business Index. The page states the survey was conducted by The Harris Poll from responses of 501 senior executives in middle-market businesses.
observed 2026-10-06
read on https://rsmus.com/middle-market/mmbi.html
06 Pricing model
not stated
No price, rate or pricing model on the AI services pages read; a complimentary consultation and workshops are offered with no fee stated.
observed 2026-10-06
read on https://rsmus.com/services/digital-transformation/artificial-intelligence.html
07 Engagement shape
first 30 days: no; who does the work: yes; deliverable: yes
No phased timeline on the page; the team is named by role; an AI readiness assessment is named as a deliverable.
observed 2026-10-06
read on https://rsmus.com/services/digital-transformation/artificial-intelligence.html
08 Own list: disclosure
publishes: none found
observed 2026-10-06
read on https://rsmus.com/insights.html
SeidrLab
Open on seidrlab.com
In the frame because: named on the Phos AI Labs list.
01 Named people with a bio
7
Seven people on the team page, each with a role and at least one sentence of prior work or specialism.
observed 2026-10-06
read on https://seidrlab.com/team
02 Vendor partner status
none claimed
The services page describes the firm as vendor-agnostic; no vendor badge on the pages read.
observed 2026-10-06
read on https://seidrlab.com/services
03 Named clients, scope
6 named clients; scope: no
Six client logos on the homepage; no engagement description is paired with a named client on the pages read.
observed 2026-10-06
read on https://seidrlab.com/
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://seidrlab.com/blog/boutique-vs-big4-ai-consultant/
05 Original research
yes
A 2026 mid-market AI adoption benchmark report. The page's own note says it reflects the firm's direct observations from client work, supplemented by public research, and is not a formal survey.
observed 2026-10-06
read on https://seidrlab.com/blog/ai-adoption-benchmark-report/
06 Pricing model
fixed; retainer
A fixed-fee roadmap and a monthly retainer with a six-month minimum are both priced on the page.
observed 2026-10-06
read on https://seidrlab.com/services/ai-transformation-roadmap
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The roadmap page lists the work week by week over five weeks, names who does it, and names the deliverable.
observed 2026-10-06
read on https://seidrlab.com/services/ai-transformation-roadmap
08 Own list: disclosure
publishes: none found
observed 2026-10-06
read on https://seidrlab.com/blog
Tenex
Open on tenex.co
In the frame because: named on the Phos AI Labs list.
01 Named people with a bio
2
Two managing partners named with multi-sentence biographies on the about page.
observed 2026-10-06
read on https://www.tenex.co/about-us
02 Vendor partner status
claimed: partner logos for Anthropic, OpenAI, LangChain, Braintrust, Lovable and Vercel (self-asserted; OpenAI's partner locator is JavaScript-rendered and Anthropic's partner page carries no directory; the firm is named in neither served page)
observed 2026-10-06
read on https://www.tenex.co/
03 Named clients, scope
4 named clients; scope: no
Four client logos on the about page; no engagement description is paired with a named client on the pages read.
observed 2026-10-06
read on https://www.tenex.co/about-us
04 Statistics with a source
0 quantified claims on the sampled page
First-article rule: the featured entry at the top of the blog index. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://www.tenex.co/blog/building-an-ai-native-sdlc
05 Original research
none found
The blog and playbooks indexes list guides, explainers and templates; no survey, benchmark, index or dataset with a stated method.
observed 2026-10-06
read on https://www.tenex.co/blog
06 Pricing model
subscription; outcome-linked
The homepage describes outcome-based, subscription-style squads billed on what is delivered; no figure is stated.
observed 2026-10-06
read on https://www.tenex.co/
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The page describes 30-, 60- or 90-day audits, names the roles on the team, and names an implementation roadmap as the deliverable.
observed 2026-10-06
read on https://www.tenex.co/ai-transformation
08 Own list: disclosure
publishes: none found
observed 2026-10-06
read on https://www.tenex.co/blog
Velsof (Velocity Software Solutions)
Open on velsof.com
In the frame because: examined in the 2026-10-03 teardown; its page for this intent names no other firm.
01 Named people with a bio
0
The about page names no person with a biography; blog bylines carry no biography, and a byline is not a bio.
observed 2026-10-06
read on https://www.velsof.com/about-us/
02 Vendor partner status
claimed: CMMI Level 3; ISO 9001, 27001, 14001 and 14064-1 marks; PrestaShop partner tiers; NASSCOM membership (self-asserted; directories not checkable, 200 JavaScript-rendered)
observed 2026-10-06
read on https://www.velsof.com/about-us/
03 Named clients, scope
25 or more named clients; scope: yes
Sixty-two entries across the work index name client organisations, each with an account of the engagement.
observed 2026-10-06
read on https://www.velsof.com/case-studies/
04 Statistics with a source
0 of 13 quantified claims link a source
List-page rule: the firm's own page for the mid-market AI consulting intent. Script: scripts/comparison/sourcing_rate.py. Re-measured on the same page on 2026-10-08: 13 claims, 0 sourced; the page read 11 claims on 2026-10-06. Five links in the measured text use unquoted href attributes; none sits in a claim sentence.
observed 2026-10-08
read on https://www.velsof.com/blog/ai-consulting-mid-market-companies/
05 Original research
none found
The blog index and the 178 post URLs in the sitemap carry guides and company posts; no firm-authored survey, benchmark, index or dataset.
observed 2026-10-06
read on https://www.velsof.com/blog/
06 Pricing model
retainer
The AI consulting page states that monthly retainers start at a published figure; other engagements carry no price.
observed 2026-10-06
read on https://www.velsof.com/ai-training-consulting/
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The page's questions section gives a timeline per engagement type, names who does the work, and names the deliverables.
observed 2026-10-06
read on https://www.velsof.com/ai-training-consulting/
08 Own list: disclosure
unassessed (no evidence retained)
The sitemaps were scanned for best-firms slugs on 2026-10-06, but the scan kept neither the URL list nor a control showing it could have found such a page, and the record carries no page URL for the cell. The reading was withdrawn on 2026-10-09.
observed 2026-10-06
read on https://www.velsof.com/blog/
Vention
Open on ventionteams.com
In the frame because: named as an example firm on the Boldr AI mid-market list.
01 Named people with a bio
9
Nine people on the company page with a role and at least one sentence of biography.
observed 2026-10-06
read on https://ventionteams.com/company
02 Vendor partner status
unassessed (no evidence retained)
The partnerships page was read on 2026-10-06, but the record keeps no text from it, and the 2026-10-08 re-observation could not match every recorded partner string on the page. The reading was withdrawn on 2026-10-09 rather than left standing unverifiable.
observed 2026-10-06
read on https://ventionteams.com/company
03 Named clients, scope
25 or more named clients; scope: yes
The work index states 59 results; its first page names 16 client organisations, each with a linked account of the work.
observed 2026-10-06
read on https://ventionteams.com/our-work
04 Statistics with a source
0 of 17 quantified claims link a source
First-article rule. Script: scripts/comparison/sourcing_rate.py.
observed 2026-10-06
read on https://ventionteams.com/blog/reduce-ai-token-costs
05 Original research
yes
A Q1 2026 report on AI in software development; the page states a survey of 480 decision-makers in the US, UK and DACH, combined with third-party data.
observed 2026-10-06
read on https://ventionteams.com/ai/sdlc/report
06 Pricing model
hourly
Time-and-materials contracts and a standard rate card are described; no figure is stated.
observed 2026-10-06
read on https://ventionteams.com/company/how-we-work/project-based-engagement
07 Engagement shape
first 30 days: yes; who does the work: yes; deliverable: yes
The page states kickoff within 14 days of signature and onboarding within 14 calendar days, names the delivery roles, and names the deliverables.
observed 2026-10-06
read on https://ventionteams.com/peace-of-mind/promise
08 Own list: disclosure
unassessed (no evidence retained)
The sitemap was scanned for best-firms slugs on 2026-10-06, but the scan kept neither the URL list nor a control showing it could have found such a page, and the record carries no page URL for the cell. The reading was withdrawn on 2026-10-09.
observed 2026-10-06
read on https://ventionteams.com/blog

Revisions

What changed, and when

2026-10-09Withdrawn, not re-observed: a blind verifier found cells whose value rested on an observer note with no stored evidence. Brainpool AI is now listed as unassessed (403, never re-requested); the DataArt, Genpact, HatchWorks AI, LOW/CODE Agency, Markovate, ISHIR, Velsof and Vention cells concerned now read unassessed with the reason; Quantum Rise asserts nothing beyond its stored page lengths and hashes. The publisher's own sourcing-rate cell was re-measured on the built page the same day and reads 0 of 4. The search-engine note states only what the capture record holds. Record: docs/authority/comparison/WITHDRAWAL-2026-10-09.md.
2026-10-08Re-observed: every third-party cell re-read on one plain request per page. Opinosis Analytics moved from unassessed (HTTP 202 on 2026-10-06) to assessed (HTTP 200); BairesDev partner marks and Velsof sourcing count updated; BCG X and Brainpool AI answered 403 again; one Boldr AI list page moved address. Record: docs/authority/comparison/VERIFICATION-2026-10-08.md.
2026-10-06First edition. Frame: every firm named on the three incumbent list pages plus the publishers, two further firms from the 2026-10-03 teardown, and Praxis. The organic-top-20 capture could not be made at zero cost without a browser; recorded, not worked around.

Every cell about a named third party is re-observed before each revision and the new date is written on the page. A stale fact about a named firm is worse than no fact. A firm that believes a cell misreads its page can write to the address on the contact page; the cell is re-observed, not negotiated.

Cite this comparison

Praxis Consulting Company, "AI implementation firms compared: eight criteria, no ranking", observed 2026-10-09, revised 2026-10-09, https://praxisconsultingco.com/research/ai-implementation-firms-compared

Cite a single cell by firm, criterion number and the date, and link the source page named beside it rather than this page alone. Because the rows carry no order, one row can be quoted without endorsing a ranking.

https://praxisconsultingco.com/research/ai-implementation-firms-compared

Ask the same eight questions of us

If you are choosing an AI implementation partner, bring the eight criteria to the first conversation. Praxis answers each one in plain terms, including the ones its own row does not read well on, and the first conversation is free.

Request a free quote Email us Call (415) 634-5434

No obligation · a scoping conversation first

AI implementation firms compared: eight criteria, no ranking

Who published this, and what it does not do

Facts by criterion, and no winner

Eight things a buyer can check in minutes

Named practitioners with a bio

Third-party-conferred partner status

Named clients, and whether engagement scope is stated

Sourcing rate on published statistics

Original research published

Pricing model stated publicly

Engagement shape stated publicly

Disclosure quality on the firm's own comparison pages

The inclusion rule, published

The query list

The lists the frame was built from

Listed as unassessed

23 firms, 8 criteria, alphabetical

The page, the date and the note behind every value

Auxis

BairesDev

BCG X

Boldr AI

Brainpool AI

DataArt

DataRoot Labs

Genpact

HatchWorks AI

InData Labs

ISHIR

LeewayHertz

LOW/CODE Agency

Markovate

Opinosis Analytics

Phos AI Labs

Praxis Consulting Company

Quantum Rise

RSM US

SeidrLab

Tenex

Velsof (Velocity Software Solutions)

Vention

What changed, and when

Ask the same eight questions of us