R.00Research
Research
The data the practice publishes and stands behind: figures checked against their primary sources, benchmarks measured on stated methods, and the record of what could not be verified. Each artefact lives at a stable URL, states its retrieval and revision dates, and is revised in place rather than re-published.
Why a research section
Numbers with a document behind them
Most of the statistics that circulate about consulting, AI and operations arrive with a firm’s name attached and no document behind it. The rule this practice works to is plain: a figure may be used only once someone has opened the source it is credited to and found the number in it. Applied to the nine AI savings figures most often quoted in decks and articles, that rule failed 7 of 9. The record of that check is the first artefact here.
Everything on these pages is somebody else’s published number, reported in the publisher’s own terms with the document and date attached, or a measurement made on a method the page states. None of it is a claim about what an engagement with us will produce. Where a source could not be read, that is recorded as a state of its own rather than smoothed into a finding. The Journal carries the reasoning and the argument; this section carries the data, and the two link to each other.
The artefacts
Published datasets
- R.01revised 2026-10-06AI savings statistics ledgerNine widely repeated AI cost-savings figures, each checked against the document it is credited to. Seven could not be published as claimed. Every entry, URL and verdict, and the one correction to this ledger's own verdicts.Open the ledger
- R.02revised 2026-10-06Mid-market AI claim auditA recurring, named audit of the quantified AI claims mid-market consulting pages publish. Forty-three claims, each traced to the document it is credited to. Thirty-four could not be. Praxis carries the only proven miss.Open the ledger
- R.03revised 2026-10-09AI implementation firms comparedThe firms on the incumbents' own best-firms lists, Praxis among them, compared on eight checkable criteria. Every cell is an observed fact with its URL and date. No ranking, no winner.Open the comparison
Further artefacts are added here as they are finished, under the same rule. Each keeps its URL, states its method and its revision date, and carries a changelog so that a reader who cited an earlier revision can see what changed. To cite a page, use the citation line printed at the foot of it.
How the data is used
From ledger to page
- 01
A figure resolves to an entry, or it is not used.
Every number rendered on the AI pages carries a reference that resolves to a ledger entry with a verdict of verified. An entry marked not found or unverifiable cannot be rendered, quoted, paraphrased or rounded into copy anywhere on the site.
- 02
The caveat travels with the figure.
An expectation stays labelled as an expectation, a forecast as a forecast, a productivity result as productivity rather than cost. The ledger records the caveat next to the figure so that a page cannot quietly drop it.
- 03
Revisions are dated, not re-published.
When a source changes, as a legal deadline did between the plan and the page in the first ledger, the entry is corrected in place, the revision date moves, and the changelog says what changed. The URL does not.
The figures that survived the first ledger are the only ones used on the AI savings pages, and the savings calculator works from a reader’s own numbers rather than from any of them. The practice the ledger was built for is AI implementation consulting.
Have a business case checked this way
If you are evaluating an AI business case and want the numbers in it traced to their documents, tell us what you are looking at. The first conversation is free and ends with a plain list of what held and what did not.
No obligation · a scoping conversation first