What should you be holding in your hands at the end of an AI assessment? Ask that question of most business owners and the answer is some version of a score: a percentile, a maturity tier, a red and yellow and green chart. Those are survey artifacts, and a survey is exactly what the first stage of serious AI work is not. When Assess is done correctly, you finish holding two things a survey cannot produce. The first is a baseline of your operation that every later result gets measured against. The second is one problem, named in a sentence and sized in a number, that the first build will remove. Everything inside the stage exists to manufacture that pair, and this piece opens the stage up to show the machinery: three distinct motions, the artifact each one produces, the failure each one prevents, and the test that tells you whether an assessment, from Omnine or from anyone else, has done its job.
The methodology overview in Assess, Architect, Activate, Adapt: how an Omnine engagement runs walks all four stages in sequence. This piece stays inside the first, because the first is where most AI initiatives are already lost before anyone notices they are losing.
Why the stage exists
The current adoption numbers describe an enormous population of organizations that started and stalled. McKinsey finds 88 percent of organizations now using AI in at least one business function, while only 7 percent have scaled it across the enterprise (McKinsey, 2025). The strugglers did not fail at the technology. Most of them failed earlier, at the selection of what the technology should do, because something got built or bought before anyone measured where the leverage was. My doctoral research found the same failure a full decade before AI: the leaders who adopted successfully started from a recognized, concrete problem, usually a communication breakdown they could point at, while the leaders who stalled had no named problem to anchor a decision to (Goodwin, 2014). Assess exists to guarantee the anchor.
There is a quieter reason too. The problem a business feels and the problem a business has are frequently different things, and the gap between them is expensive. Advocate Roofing, the Alabama roofing company whose full story is in how a roofing company stopped rebuilding every quote by hand, arrived certain its quoting software was broken after a version migration. Assessment work found two perfectly healthy systems that had stopped speaking the same language: product names in the estimate templates no longer matched the supplier library they pulled from, and a redundant supplier connection was interfering underneath. The felt problem pointed at a purchase. The actual problem pointed at an afternoon of repair on a naming layer. An owner who skips the stage builds on the felt problem, and the felt problem lies often enough to make that a real gamble.
Motion one: the readiness read
The first motion measures the person and the posture before it measures the operation, because the research is blunt about where adoption is decided. In my study, budget did not separate the leaders who moved from the leaders who stalled. Neither did company size, and neither did age. What separated them were levers that live with the leader: whether they could explain the technology in plain terms, whether they had engaged with it personally, whether they could see the unified system underneath the pile of tools, whether a real problem was driving the decision, and how they carried their organization while it changed (Goodwin, 2014). The full map of those levers is in the 5 drivers of AI adoption.
The readiness read is a structured instrument, deliberately short, that scores where you sit on those levers today. Its brevity is a design decision rather than a limitation. The levers are few, they are personal, and each one is measurable with a small number of direct questions, so an instrument that runs to sixty items has usually wandered into inventorying your software, which is a different job that a later motion handles better. What the read produces is a position: a plain statement of which lever is currently holding you back, specific enough to act on. What it prevents is the most personal failure in the whole sequence, which is prescribing a system to an operation whose leader is not yet positioned to carry it. A build delivered to a leader who cannot explain what it does, has never touched the tools it runs on, and feels no problem it solves will become shelfware with an invoice attached, and no amount of technical quality downstream repairs that. The read is also the one motion you can run on yourself tonight without a phone call, and I will come back to that at the end.
Motion two: the problem inventory
The second motion turns from the leader to the operation, and its job is surfacing: finding the expensive, recurring failures in how information moves through the business. The raw material is the daily texture of the work. The customer record rekeyed into a second system is a translation failure. The status that has to be chased through messages is a pulling failure that costs two people at once. The lead that arrived and was never called is a silent drop, priced in revenue rather than minutes. The context reconstructed in a meeting because it was never captured where the work happened is rebuilding. The full taxonomy of those four failures, and the one week audit that logs them, is the subject of the communication tax your business pays every day, and that audit is the inventory motion in its self-serve form: a week of light logging under four headings, tallied for thirty minutes on Friday, with one strict rule, which is to observe for the week rather than fix, because teams that start fixing mid week stop logging.
What the motion produces is a written inventory with minutes attached, and the word written carries weight. An inventory that lives in the owner’s head is a mood, and moods rank problems by irritation. A written inventory with counts and minutes is evidence, and evidence ranks problems by cost. What the motion prevents is diagnosis by anecdote. The Advocate engagement is again the instructive case. The inventory question is never which tool is misbehaving. The inventory question is where the information stops flowing, and asked that way, the roofing problem stopped being a software complaint and became a traceable path with a break at a specific, repairable point. Symptoms nominate themselves loudly. Failures have to be found, and the inventory is the finding instrument.
Motion three: the leverage ranking
The third motion is selection, and it is where the stage earns its keep, because an inventory of fifteen problems is not a plan. The ranking scores every inventory item against a small set of questions. How often does it occur, because repetition is what automation is for, and a maddening failure that happens twice a year is a poor first target. How many minutes does it consume per occurrence, counted across every person it touches, since chasing consumes the asker and the answerer both. Does it cross at least two tools, because failures that live between systems are the ones a flow redesign removes outright rather than relocates. And does it ever touch revenue, because a dropped inquiry or a late quote is priced in customers, and one of those can outweigh a month of everyone’s rekeying. The selection discipline behind those questions, and the trap of automating the most irritating item instead of the most expensive one, is laid out in how to pick the one task to automate first.
What the motion produces is the second artifact the stage exists for: one named, sized first move. Not a transformation roadmap with twelve initiatives, one move, chosen because the evidence says it carries the most leverage, scoped small enough to build in days, and measured against the baseline the earlier motions established. What it prevents is dispersion, the many shallow pilots pattern the adoption research keeps finding, and it points the first build at where the documented value lives. McKinsey’s finding on this is specific: of all the changes organizations make around AI, redesigning a workflow end to end is the one most associated with bottom line impact, and only about one in five organizations using the technology has done that work (McKinsey, 2025). A ranked inventory is what makes depth a decision instead of an accident.
What the stage measures, in order
Notice the sequence the three motions trace, because the order is doing quiet work. The stage measures the leader first, the operation second, and the leverage third, and each measurement makes the next one trustworthy. An operational inventory taken before the readiness read lands on a leader who may not be positioned to act on it. A leverage ranking run before the inventory ranks anecdotes. Run in order, the motions convert three questions every owner half asks anyway, where do I stand, what is broken, what should I do first, into three artifacts that can survive contact with a skeptical accountant. That conversion, from questions into evidence, is what the word measures in this piece’s title is claiming, and it is the difference between an assessment and an opinion with formatting.
The artifact test
Put the motions back together and the stage’s real product comes into focus. The readiness read produces a position. The inventory produces evidence. The ranking produces a decision. Stacked, they are the two artifacts named at the top: a baseline, which is the scored position plus the measured inventory, and a named, sized first problem, which is the ranking’s output. That pair is also the exit gate. In the methodology piece I put the gate this way: you leave Assess only when the problem fits in one sentence with a number attached. We lose this many hours a week to that. This many inquiries a month go unanswered after hours. If the sentence needs three qualifiers, or the number is a shrug, the stage is not finished, whoever performed it and whatever the deliverable says on its cover.
The test matters because it is how you audit any assessment, including one you paid a stranger for. A deliverable that ends in a maturity score has measured you against an abstraction. A deliverable that ends in a sentence and a number has measured your operation against itself, and it hands you something with three unusual properties. It is falsifiable, because the number can be measured again after the build and the build either moved it or did not. It is portable, because you can act on it with any provider or with none. And it is small, because a first move that cannot fail informatively was scoped wrong. Scores flatter or alarm, and either way they mostly get filed. Baselines and named problems get built against.
Fair questions
Can a three minute online instrument replace the full stage? No, and it does not claim to. The public readiness read is motion one in a self-serve form. It scores the adoption levers and points at the one currently setting your ceiling, which is real information, and it stops there. It does not walk your operation, log your week, or rank your inventory, which is exactly the difference between a read and an assessment. Treat it as the stage’s honest first motion rather than a compressed whole, and treat any short instrument that claims to be a complete assessment with suspicion, since the claim itself reveals a misunderstanding of what the stage produces.
We already know our problem, so can we skip to building? Sometimes you do know it, and the stage confirms it quickly at almost no cost. But the felt problem and the actual problem diverge often enough that the confirmation is the cheapest insurance in the method, and the roofing case is what the divergence looks like with money attached: certainty said replace the software, and measurement said repair a naming layer in an afternoon. Skipping Assess never skips the assessment. It just performs the assessment implicitly, badly, and after the spend instead of before it.
Is this discovery consulting under a new label? The label is not the difference. The artifacts are. Discovery that ends in a proposal has produced a sales document whose incentives you can guess. Assess ends in a baseline and a named problem that stand on their own, serve as the measuring stick for everything built afterward, and leave you free to stop with full value in hand. That stopping point is designed in rather than tolerated, and the gates between all four stages, along with why a method with real exits is the opposite of a funnel, are covered in the methodology piece.
Run the first motion tonight
The stage is not an abstraction you have to hire someone to understand, because two of its three motions have self-serve forms. Motion two is the one week audit from the communication tax piece: four headings, light logging, thirty minutes on Friday, and a defensible number at the end. Motion one is faster still. The Omnine AI Readiness Assessment is the readiness read itself, public and scored, and about three minutes with it will tell you which adoption lever is currently setting your ceiling. Run the read tonight, start the audit Monday, and by Friday you will have performed most of an Assess on your own business, which is exactly how the method prefers to be tested.
References
Goodwin, M. R. (2014). A qualitative descriptive multiple-case study: Fortune 500 leaders’ social business platform adoption (Doctoral dissertation, University of Phoenix). ProQuest Dissertations Publishing (UMI No. 3648813).
McKinsey & Company. (2025). The state of AI in 2025: Agents, innovation, and transformation. McKinsey Global Survey on the state of AI.