The idea
Most companies buy an AI tool. Very few change how the work gets done. The second one is the part that pays.
Buying a tool is not the same as changing the work
A company buys a chatbot. Six months later the same people are doing the same jobs, plus one more tab to check. Nothing got faster. Nothing got cheaper. The tool was fine. The work never moved.
We start from the other end, with how a job travels through your business from the first touch to the money landing in the bank. A lead arrives. Someone answers it, or does not. Someone writes it down, or forgets. Someone follows up on Thursday. Someone raises the invoice. Someone chases the payment. That whole line is the thing we automate, not one box on it.
That is why we ask about sales and finance in the same conversation. A faster reply is worth little if the invoice still sits in a drawer for nine days. The gain shows up when the chain runs on its own, and a person only steps in where a person is genuinely needed.
Praxis is founder-led. Evan Dang leads every engagement and draws on a team of 20 senior practitioners across the disciplines the firm advises on.
The deal
How the engagement works
Nothing about this is unusual until the last step. The first three cost you nothing.
- 01
Free consultation
One call, usually 30 to 45 minutes. You tell us how the business runs and where the day gets eaten. We tell you which of that work AI can take over, which of it cannot be taken over yet, and what we would build first. You leave with a plain list. If the answer is that there is nothing here worth building, we say that on the call.
What happens on the call - 02
Free two-week diagnostic
If the call goes somewhere, the team spends two weeks inside the real work. We sit with the people doing the jobs, count how long things take, and read what your systems already hold. At the end you get a written map of every job AI could do, ranked by what it is worth and how hard it is. That document is yours whether or not you go further. It still costs nothing.
What the two weeks produce - 03
Build and handover
We build the automation, connect it to the systems you already own, and run it beside the old way until it is trusted. We train the people who will live with it. Then we hand it over with notes a normal person can read.
How a build runs - 04
A share of the measured gain
Before anything is built, both sides write down what the business does today. That is the baseline. Afterwards, the same measure is taken again. The fee is a share of the measured gain. If there is no measured gain, there is no share.
How you pay
What counts as incremental profit
Incremental means new. It is the difference between what the business earns after the automation runs and what it earned doing the same work the old way.
The baseline is the old way, written down and agreed before anything is built. It is usually three things: what the work costs now, how long it takes now, and how much of it gets finished now. Your finance people pick the measure. Praxis does not get to choose it afterwards, and neither do you.
Some of the gain is money saved. Some is money earned that was being dropped, like a lead that went unanswered on a Saturday. Both are counted the same way: measured against the baseline, using your own records, not ours.
The industry calls this gainshare, or outcome-based pricing, or a share of savings. The mechanics matter more than the name, and they are set out in full on how the profit share is measured.
AI implementation
AI implementation across five business functions
These are the five parts of a business where the work is repetitive enough for AI to take real weight off people, and countable enough for a baseline to hold. Most companies have the same problems in the same places.
- Today
- A lead that arrives in the evening sits in a shared inbox until a rep sees it the next morning, guesses which territory it belongs to, and forwards it by hand.
- After
- Every inbound lead gets a real reply within seconds, at any hour, scored and routed to whoever is actually working, with the CRM record filled in from the call and email automatically.
- Measured by
- Minutes from a lead arriving to its first real reply, and the share of leads that get one, read from the CRM's own timestamps.
- Today
- Routine questions and complicated ones land in the same queue and wait the same amount of time, with the account history scattered across systems a person has to piece together before answering.
- After
- Routine questions are answered immediately, day and night, and the rest reach a person fast, in the right queue, with the account and billing history already attached.
- Measured by
- What share of tickets close without ever reaching a person, and how long the rest wait before landing with someone, read from the help desk's own ticket log.
- Today
- Invoices are read and coded by hand in whatever layout the sender used, expenses get a spot check rather than a full one, and the close runs through email and spreadsheets until a variance nobody can explain turns up days later.
- After
- Invoices are read, matched to the purchase order and coded automatically with only the exceptions sent to a person, every expense claim is checked against policy, and reconciliations and variance explanations are drafted before anyone opens the file.
- Measured by
- Cost to process one invoice, days to complete the close, and late-payment or duplicate-payment charges, all read from the ledger the finance team already keeps.
- Today
- Interview scheduling runs across several calendars by email, onboarding paperwork gets chased one document at a time, and nobody has told the candidate an automated tool is even in use.
- After
- Scheduling, onboarding paperwork and routine candidate questions are handled automatically, built to sit inside the audit, notice and human-oversight duties that now apply to hiring tools, with no automated rejection of any candidate.
- Measured by
- Days to fill a role, hours spent on scheduling and onboarding admin, and whether every audit and notice duty was met on time, read from the applicant tracking system's own records.
- Today
- Password resets, approvals and access requests move through email one at a time, a report gets rebuilt by hand from several exports, and meeting notes live only in whoever happened to be taking them.
- After
- Tickets are read and provisioned against the rules automatically, approvals route by policy and chase whoever is holding one up, and dashboards, meeting notes and routine email arrive already drafted for a person to check.
- Measured by
- How many tickets and approvals close without a person, and how long the rest take from request to resolution, read from the service desk and approval system's own logs.
- 06AI Enablement and Workflow Automation at EliteSEO Consulting (opens in a new tab)Search and marketing operations (not ours)This one Praxis does not sell. Search and marketing automation lives at EliteSEO Consulting, an affiliated practice with the same founder. The two practices share no clients, engagements or staff, and we would rather say that plainly than take work twice.
If you want advice on which AI tools to buy rather than someone to build it, that is a different piece of work and it lives here: advice on which AI tools to buy.
The through-line
The work travels as one line, from the first touch to the money landing. That whole line is the thing we automate.
Numbers, and whose they are
The case for AI, in published figures
These figures come from two different places: government surveys measuring how many businesses use AI at all, and what executives at one company say they expect AI to save. None of it is a Praxis result, and none of it is a prediction about your business. Each one is listed with its publisher and its date on where every number comes from.
15%
more issues resolved per hour by customer support agents given a generative AI assistant
The Quarterly Journal of Economics, Generative AI at Work (Brynjolfsson, Li and Raymond), vol. 140, issue 2, Oxford University Press, published . Read back .
Concentrated among the least experienced agents. Sample of 5,172 agents at one company. A measured result, not a forecast, and a productivity result rather than a cost saving.
11%
lower costs, which surveyed executives expect within 18 months of deploying and scaling generative AI
Accenture, The front-runners' guide to scaling AI: Lessons from industry leaders (Accenture Research), published . Read back .
What executives expect, not a measured result.
37%
of firms with 250 or more employees report using AI
US Census Bureau, Large Firms With at Least 20 Employees Biggest AI Users (Business Trends and Outlook Survey), by Adam Grundy, Cory Breaux and Dhanapati Khatiwoda, published . Read back .
Adoption among the smallest firms, four employees or fewer, is far lower. Measures adoption only, not cost or productivity.
18%
of US firms had adopted AI by the end of 2025
Board of Governors of the Federal Reserve System, FEDS Notes: Monitoring AI Adoption in the US Economy, by Jeffrey S. Allen, published . Read back .
Most of the labor force already works at a firm that has adopted AI, even though most individual firms have not. Measures adoption only; no measured productivity or cost effect.
The founder
Before Praxis, the founder worked on an AI transformation program at a prior engagement at a global consulting firm, targeting $150M in annual synergies. That is the founder's prior engagement, not a Praxis client result. The full background is on the full background.
No client names appear on this site because there are none to name. When there are, they will appear with permission and with numbers you can check.
The terms
The fee is a share of the measured gain. If there is no measured gain, there is no share.
Estimate
Estimate the savings in your business
There is a small tool that takes your industry, your headcount and the parts of the work that still feel manual, and turns them into a range using published figures. It shows the arithmetic so you can check it. It is an estimate built from other companies' published studies. It is not a quote, and it is not a claim about your business.
Estimate what AI could save in your businessSelf-assessment
AI readiness assessment
Eight questions, answered by you, scored in your browser. Nothing is sent anywhere. It gives you a reading of where you stand before you ever speak to us, and it tends to be the fastest way to find out whether the free call is worth booking.
The same scorecard also runs at its own address, with no email step.
AI Readiness Scorecard
Eight questions, scored in your browser. Your answers are sent only when you ask for your scorecard at the end.
~2 minutes
Fit
Who this is for
Businesses of any size
A one-person business and a company with a few hundred staff both have the same problem, in different clothes: work that a person does by hand because nobody ever changed it. For a small business, the version of this offer is here: AI implementation for a small business. For a company that already owns its systems, the work is joining them up rather than picking new ones: AI implementation for a mid-sized company.
Very large companies are a different shape. Before anyone starts there, you need a master agreement, insurance, a security review and a vendor risk assessment. Praxis does not have those in place, so we will say so on the call rather than three months later.
What stays out of scope
- Work where nobody will agree a baseline. If the gain cannot be measured, this arrangement has no honest basis and we would rather not start.
- Hiring decisions made by a machine with no person accountable for them.
- Anything that needs a person's judgement about a person: firing, medical calls, safety calls.
- Automating a broken process. If the process itself is wrong, automating it just makes the wrong thing happen faster, and we will tell you that in the free two weeks.
- Work we cannot see. If we cannot get access to how the job is really done, we cannot count it, and neither can you.
Questions
What should you know before the first call?
- How does Praxis charge for AI implementation work?
- Praxis charges a share of the measured gain the automation creates, compared against a baseline agreed in writing before any building starts. There is no fee up front, no hourly rate and no retainer. If there is no measured gain, there is no share.
- What does the free two-week diagnostic actually produce?
- It produces a written map of every job in the business that AI could take over, ranked by what that job is worth and how hard it would be to change. It also names the jobs that should stay with people. The document belongs to the business whether or not it goes on to build anything.
- Which parts of a business can AI realistically take over today?
- The work that repeats and follows rules: answering routine questions, sorting and routing requests, reading and coding invoices, checking expenses, scheduling, writing first drafts of reports, and moving information between systems. Judgement calls about people, safety and strategy stay with people.
- How is the baseline for an AI profit share agreed?
- The baseline is a written record of what the work costs, how long it takes and how much of it gets completed before anything changes. It is taken from the company's own records, chosen by the company's own finance people, and fixed before the build begins so it cannot be adjusted afterwards.
- Does this work for a small business with no IT department?
- Yes. A small business is usually the fastest case, because there is nobody to negotiate with and the tools are off the shelf. The work starts with whoever answers the phone and whoever does the admin, since that is where a small owner's day disappears.
- Who runs the automation after it is built?
- The business does. Everything is built on systems the company already owns or can own outright, handed over with written notes, and the people who use it are trained before handover. It is not built to keep anyone dependent on Praxis.
Start with a free consultation
Tell us how the business runs today and where the day gets eaten. We will tell you what AI can take over, what it cannot, and what we would build first. The call costs nothing and ends with a plain list either way.
No obligation · a scoping conversation first