A management consultancy changes what a company decides. An AI enablement agency changes what its people do on a Tuesday afternoon. That gap is the whole comparison. If nobody can agree on where AI belongs in the business, hire the consultancy. If 200 people hold licenses, the usage dashboard looks respectable, and the weekly work product is identical to last year’s, no strategy document will move it. The AI enablement agency vs consultancy choice turns on where your gap sits: in the decision, or in the behavior.
The case that a signed license is not a capability is made at length in why buying licenses is not the same as building, and the operating model sits in the playbook from license rollout to measurable team capability. This piece is narrower: two vendor types, one procurement decision.
What each one sells
A management consultancy sells a defensible opinion: analysis, benchmarking, option framing, and a recommendation your leadership team can act on and your board can accept. The artifact is a document, and that is the product, not a flaw in it.
An enablement partner sells a changed worker: instruction and repetition against the documents your team already produces. The artifact is a person who reaches for an AI tool at the right moment and gets a usable result. That deliverable walks around the building after the invoice clears.
| Management consultancy | AI enablement agency | |
|---|---|---|
| Core deliverable | Analysis and a recommendation | Fluency in named people |
| Unit of work | The engagement | The cohort |
| Who does the work | The firm’s team | Your staff, on your material |
| How success gets argued | Quality of the reasoning | What the team produces after |
| Billing shape | Advisory time against a rate card | Per cohort, per seat, or a retainer |
| Scales with | Number of consultants deployed | Number of people trained |
| First visible change | A decision, in weeks | Work product, within days |
Read the bottom two rows together and the economics diverge. Consultancy cost tracks the size of the team doing the thinking; enablement cost tracks the heads you want changed. At 50 to 500 people, that is a budget growing with analysis against one growing with coverage.
The AI enablement gap, in numbers
The strongest argument for enablement is the distance between two figures in the same survey. McKinsey’s State of AI survey, fielded May 4 to June 8, 2026 with 1,719 respondents across 97 countries, found 80% say AI has improved their own productivity while 37% say it has moved their organization’s operating profit (EBIT). Six percent clear McKinsey’s bar for high performers.
Individual productivity is close to free now: a capable person with a license and ten minutes of curiosity gets faster at drafting. Organizational return is not free, and most companies have not bought it.
The tell is what separates the 6%. Close to three-quarters of high performers have fundamentally redesigned workflows because of AI, up from 55% a year earlier; about a quarter of everyone else has. Workflow redesign is not a document. It is an argument won in one department, with named people, about one weekly task, and then held long enough that the new way becomes the obvious way.
Why a slide deck cannot close a behavior gap
BCG has put a ratio on this for years, restated in its July 2026 analysis of how AI leaders create competitive advantage: roughly 10% of the effort is technology, 20% is algorithms and data, 70% is people, process and organizational change. Most AI programs invert it and spend their attention on the 10%.
The 70% is the part no deliverable made of slides can execute. A recommendation can say “redesign the proposal workflow”. It cannot sit with four proposal writers and watch three drafts fail, which is where the learning happens.
In the sessions we run, the moment that changes a habit is never the explanation. It is the second attempt, on the person’s own document, when the output is finally good enough to send.
The same analysis puts a skills number beside it: AI leaders carry AI-related skills in about 13% of their workforce against 1% at laggards. Skills density is a slow-moving asset, bought in cohorts over quarters, not decided in a steering committee.
What an AI enablement agency does differently on the ground
Four mechanics separate a real enablement engagement from a product demo wearing a training label.
- It runs on your documents. Exercises use the team’s own proposals, meeting notes, spreadsheets and inbox, not a sample dataset. Anything else teaches the tool rather than the job.
- It is organized by role, not by headcount. A finance analyst and a marketing coordinator need different prompts, different checking habits and different refusals. One all-hands session averages those differences into nothing.
- It is facilitated, not presented. People type during the session. A presenter who talks for 90 minutes has delivered a webinar, and webinars do not change behavior.
- It is measured on output, not completion. Course completion is the easiest metric in corporate learning and the least informative. Ask instead whether a named weekly artifact now takes less time or comes out better.
Worth being plain about scope: what we deliver is hands-on fluency with the mainstream assistants, including ChatGPT, Claude, Gemini and Microsoft Copilot, applied to the team’s real work. Not a tool-stack audit, a custom build, or AI-policy drafting. Vendors who bundle all of it under one word are how buyers end up paying consulting rates for a training outcome. What enablement services include, and what they quietly exclude breaks that line-item question down.
Where the management consultancy is the right hire
Nearly every comparison of this kind online is written by an agency and concludes against the consultancy. A COO who has used a good firm stops reading there. Four situations make the consultancy the correct purchase and enablement premature.
- Capital allocation across business units. When the question is which three of eleven functions get funded, you need comparative analysis, not a workshop.
- Organizational redesign. Role consolidation and headcount planning carry legal and morale risk best held by advisors insured for it.
- A decision that must survive outside scrutiny. Regulators, acquirers and boards sometimes need a third-party opinion with a brand on the cover. No training program produces that artifact.
- Real strategic uncertainty. If leadership disagrees about whether AI changes your product or only your overhead, fluency will not resolve it.
The largest firm in the category reached the same conclusion about its own gap. Accenture launched LearnVantage in March 2024 as part of a $1B reskilling investment, bought Udacity, then bought Ascendient Learning in May 2025 to add instructor-led delivery across a catalog of more than 850 courses. A firm with unlimited access to strategy talent committed a billion dollars to owning the ability to teach people in a room. Treat that as the clearest evidence that these are two separate purchases.
What each model costs, and what the price buys
Start with a denominator. ATD’s 2026 State of the Industry puts average direct learning spend at $846 per employee against 16.7 formal learning hours in 2025, up from 13.7. That is the baseline your finance team already accepts for all workplace learning combined.
Against that baseline the market ranges are easy to sanity-check. Workshop-shaped enablement programs land roughly between $2,000 and $15,000 depending on cohort count and depth. Follow-on custom automation work, once a team knows what it wants built, runs in the $25,000 to $150,000 band. Published prices are rare on either side: Section lists a price for single users but quotes teams privately, which is the norm.
Consultancy pricing works on a different axis. Engagements price from a rate card against deployed hours, so the invoice reflects how many consultants were in the room, not how many of your people changed. Both are legitimate; they answer different questions. Divide either quote by the number of people whose work will be different afterward and the question you are paying for becomes obvious.
One procurement note that saves real money: both models sell readiness assessments, at prices that are not close, for a similar artifact. What readiness assessments measure and what they miss is worth reading before you buy one twice.
Five questions that identify the model in one call
Labels are unreliable. Consultancies list enablement services; training vendors call themselves strategic partners. These five questions sort them in half an hour.
| Ask this | A capability partner answers | Treat as a warning |
|---|---|---|
| Whose documents do we work on in session one? | “Yours. Send three real examples in advance.” | “We have sample exercises for your industry.” |
| Who is in the room, and in what groups? | Named roles, split into cohorts | “All staff, one session, recorded” |
| What does someone do in the first 20 minutes? | Types, with a facilitator watching | “We walk through the platform” |
| What will you measure at week six? | A named weekly artifact, before and after | Completion rates and satisfaction scores |
| What happens if nothing changes? | A specific remedy, written into the agreement | “Adoption depends on internal champions” |
The fifth question separates the field. A partner who has delivered this work has watched a cohort fail to change and has a view on why. One selling it for the first time talks about champions.
Running both without paying twice
Sequencing is the practical answer, and the order matters. Buy the decision only if the decision is open. If leadership already knows which functions matter and the constraint is that nobody in them is good at this yet, skip the strategy phase; it will hand you back what you said in the kickoff.
Where both are warranted, keep the scopes from overlapping. One readiness diagnostic, not two. One named owner of the measurement definition, agreed before either contract is signed, because the common waste here is paying a consultancy to define success and a training vendor to define it differently. At shortlist stage, the enablement options worth comparing for mid-market companies covers who does what credibly.
Frequently asked questions
What is the difference between an AI enablement agency and a management consultancy?
A management consultancy produces a recommendation; an AI enablement agency produces people who work differently. The consultancy’s unit of work is the engagement, priced by deployed hours, and the agency’s is the cohort, priced by coverage. Both sell readiness assessments, so judge them on the end deliverable: a decision document, or changed weekly output.
Is AI enablement just consulting with a new name?
Sometimes. The term is loose enough that firms selling pure strategy have added an enablement page without changing what they deliver. The test is whether the engagement puts your staff’s hands on your own documents during paid hours. If the agenda is interviews, analysis and a readout, it is consulting, whatever the page says.
Who should own AI capability after the engagement ends?
Your team, and the contract should say so in plain terms: the prompts, the checking habits, the internal examples and the session materials stay with you. A vendor who keeps the working method and resells access each quarter is running a managed service. Fine to buy on purpose, bad to buy by accident.
How much should we budget for AI enablement?
Workshop-shaped programs sit roughly in the $2,000 to $15,000 range depending on cohort count and depth, with custom automation work afterward in the $25,000 to $150,000 band. Divide any quote by the heads it covers and compare against ATD’s $846 per employee. A number far above that has to be earned by coverage and measurement, not by brand.
Will a management consultancy train our team?
Many will, through a learning arm or a subcontracted partner. Accenture built LearnVantage and bought two training companies for this. Check who stands in the room: a strategy firm reselling a training partner adds a margin layer to a service you can buy directly.
Our license dashboard looks fine. Why does capability look worse?
Because a dashboard counts sessions, not skill. Logins, message counts and seat activation all climb after rollout without anyone’s weekly output changing, which is the shape of McKinsey’s 80% to 37% gap. The honest measure is a named artifact: take the proposal, report or client email your team writes every week and compare it before and after.
How long before a program changes behavior?
The first visible change lands within days of a well-run session, because people leave with a working method for one real task. Durable change takes a quarter: it needs repetition and a follow-up pass once the easy wins are gone. A vendor promising transformation from one session is selling a webinar.
What if our problem is strategy, not skills?
Then buy the consultancy, and buy it first. The signal is disagreement at the top about what AI changes in your business model, not frustration at how little the team gets from the tools. Different problems, and the second does not respond to analysis.
Can we do this in-house instead?
Sometimes, if you have a credible internal practitioner with protected time, which at 50 to 500 people is rare. What fails in-house is not the teaching; it is the scheduling, the role-specific design and the measurement after week two. If this landed on someone as a second job, assume measurement is what slips.
Where to go next
- Still deciding what to buy: name the one weekly artifact you want different in 90 days. If you cannot name it, the gap is strategic and the consultancy is the first call.
- Already holding quotes: run the five questions against each, then divide every price by the heads whose work will change.
- Building the internal case: McKinsey’s 80% to 37% gap and BCG’s 10-20-70 split are the two figures that move a CFO.
- Ready to compare partners: start with the mid-market enablement options.
If it would help to pressure-test a quote with someone who is not selling you a strategy phase, book a free AI-readiness call. Thirty minutes, your own documents, and a straight answer about which model your next two quarters need.
Arthur Wandzel
