By Andreas Pettersson, Founder, Leaders ADAPT
Every week a CEO of a company between $5M and $50M asks me some version of the same question. Not "should we use AI". That one is settled. The question is "what kind of help do we actually need", because the market now sells four different things under the word consultant, and they cost between a few hundred dollars and a senior salary.
I built an AI company before ChatGPT existed, grew it to 150 people and sold it to Canon. Since then I have sat on the buying side of this question with owners of manufacturing, distribution, construction and professional services companies. This page is the sorting I do with them, written down so you can do it yourself in ten minutes.
Quick answer: Four shapes of AI help exist: a one off consultant, a fractional chief AI officer, a CEO cohort, and doing it yourself. Pick by who has to change. If only the CEO has to change, a cohort or self service works. If the leadership team has to change, you need an owner inside the company for six months. If a board wants documents in sixty days, buy the documents, not a course.
Leaders ADAPT sells three of the four shapes described here. The price ranges quoted for other providers are their own published figures, linked, and the page wins if a figure has changed. No provider on this page pays us, and we resell no platform.
The six question self check
Answer each one honestly. Count your answers at the end.
- Who has to change for AI to pay off in your company? Only me (A). My leadership team (B). The whole company, department by department (C).
- What is the deadline? None, I want to learn (A). A board, bank or owner meeting inside sixty days (B). A specific operational problem that costs money every month (C).
- Who would own AI inside the company after the help leaves? Me, personally (A). Nobody yet, that is part of the problem (B). An operations or finance lead who could run it with direction (C).
- How many hours a week can you personally give it? Two or three (A). Under one (B). It depends on who is doing the building (C).
- What is your revenue? Under $10M (A). $10M to $50M (B). Above $50M (C).
- What do you want in hand at the end? Systems I run myself (A). Documents a board can read and a plan with owners (B). A function that runs without me (C).
Mostly A: a cohort programme or self service. Mostly B: the board documents lane, or a fractional chief AI officer if revenue is above $10M. Mostly C: a fractional chief AI officer, with implementation scoped inside the engagement. A tie between B and C with a board deadline: buy the documents first, then decide on the function.
That is the whole method. The rest of the page explains what each shape actually is, what the providers who publish prices charge, and the mistakes that cost the most.
Shape one: the one off AI consultant
A consultant is hired for a scoped piece of work: an AI readiness assessment, a roadmap, a vendor selection, sometimes a build. The engagement ends when the deliverable is delivered. Prices run from a few hundred dollars for a recorded strategy intensive to five and six figures for an implementation project.
This is the right purchase when you have one specific question and someone inside who will act on the answer. It is the wrong purchase when nobody inside owns AI, because a roadmap with no owner is a PDF.
The question that exposes a consultant quickly: what do I own on day 30 if I stop. A good one answers in one sentence.
Shape two: the fractional chief AI officer
A fractional chief AI officer (also sold as a fractional CAIO, a fractional AI officer or, at Leaders ADAPT, the Fractional AI Executive) joins your leadership team for a fixed number of hours a week over several months and owns the AI function the way a fractional CFO owns finance: discovery, strategy, policy, vendor selection, the first builds and the handover to an internal owner.
Published retainers in 2026 run from about $4,000 to $30,000 a month across tiers, with $7,500 a month and an eight client cap at one US practice, £2,500 to £6,000 a month at a UK one, and $10,000 to $30,000 at a technical practice aimed at funded companies. Same title, a seven fold spread. The spread is hours and whether the person advises or builds.
This is the right purchase above about $10M in revenue when the leadership team has to change and nobody inside can carry it yet. It is the wrong purchase for a company that only needs the CEO to get fluent, because you will pay a senior person to teach you tools.
I wrote a longer comparison of AI adoption consulting against the fractional model and what a fractional chief AI officer actually does in the first ninety days.
Shape three: the CEO cohort programme
A cohort puts the CEO in a room with other CEOs and teaches by building. The CEO leaves with systems running in their own company, not a certificate. Prices are in the low thousands for operator programmes, with peer networks for larger companies priced at $25,000 a year.
At Leaders ADAPT this is the Executive AI Program: eight weeks, one two hour session a week, nine AI systems live inside your company by the end, and a published price on its page. Its sibling, the AI Leadership Mastermind, is for the CEO whose board is asking for a strategy: six months, and you leave with the five documents a board can read.
This is the right purchase when the CEO is the bottleneck and the company is under about $10M, or when a board wants documents and the CEO wants to be the one who understands them. It is the wrong purchase when the CEO has under an hour a week. Cohorts reward the people who do the homework.
Shape four: do it yourself
Plenty of owners do. The ones who succeed do three things in order: write the one page AI policy first, so the team knows what never goes into a tool, pick one workflow that costs money every month, and set a monthly thirty minute review where someone reports what runs now that did not run before.
The ones who fail skip the policy, start five pilots, and six months later have five subscriptions and no system. The AI readiness checklist is the honest test of which group you are in. If you cannot name the internal owner, you are not doing it yourself, you are doing nothing slowly.
What each shape costs, using only public figures
| Shape | Public range in 2026 | What the money buys | Wrong when |
|---|---|---|---|
| One off consultant | A few hundred dollars for an intensive to five or six figures for a build | A deliverable: assessment, roadmap, vendor pick, or a built system | Nobody inside will act on it |
| Fractional chief AI officer | About $4,000 to $30,000 a month | An owner inside the leadership team for months, strategy through adoption | Only the CEO needs to change |
| CEO cohort | Low thousands, with peer networks up to $25,000 a year | Systems the CEO runs, or board documents, plus peers | The CEO has no weekly hours |
| Do it yourself | Tool subscriptions | Whatever the internal owner ships | There is no internal owner |
Two rules make these numbers comparable. Ask for hours inside the fee, because a retainer with eight hours and one with four days can carry the same price. And ask whether the person advises or builds, because building costs more per hour and saves more per month.
The four mistakes that cost the most
Buying the build first. A six figure implementation before the leadership team can run the tools produces a system nobody owns. Train the leaders, build small inside a retainer, then scope the large build when you know which one you need. Slower for a month, faster for the year. I have written about why AI projects fail and this is the most common reason.
Hiring to a title. "Chief AI officer" on a proposal tells you nothing about hours or about who is on the calls. A junior analyst behind a senior name is common at the agency end of the market.
No policy, so no permission. Teams that do not know what they may paste into a tool either paste everything or nothing. The one page policy is a week of work and it is the first artifact in every engagement I run.
Delegating the judgement. AI can draft, summarise, triage and forecast. It should not decide who gets hired, fired or paid, and it should not be the reason a wrong number went to a customer. The list of what not to delegate to AI is short and it does not change with the model version.
Where the shapes connect
The three shapes Leaders ADAPT sells are built as a ladder, not a menu. The Executive AI Program builds the systems in your company. The AI Leadership Mastermind builds the strategy your board signs. The Fractional AI Executive means I run the function inside your company for six months. A CEO who finishes the program and then wants the function owned is the normal path to the fractional engagement, and a company that starts fractional usually puts its operations lead through the program in month three so there is someone to hand the function to.
That is also how to read any provider's ladder. If the steps connect, the provider has done this before. If every step is sold as the only step, you are the first.
Not sure which shape fits? Bring the question to a 30 minute fit call
The call is thirty minutes, no pitch. You bring the question your board or your bank balance is asking. I tell you which of the four shapes fits, including the ones we do not sell, and what it should cost you based on the public ranges above. If it is the fractional engagement, we scope the six months there. If it is do it yourself, you leave with the three things to do first.
I scaled a company to 150 people across Europe, Japan and the US and exited to Canon. I have bought every one of the four shapes for my own companies, and I have been the consultant who delivered a roadmap nobody owned. That is why the questions above are the ones they are.
Which AI help do you need FAQ
What is the difference between an AI consultant and a fractional chief AI officer?
An AI consultant is hired for a scoped piece of work, usually an assessment, a roadmap or a build, and leaves when it is delivered. A fractional chief AI officer joins the leadership team for a set number of hours a week over months, owns the AI function, chooses vendors, writes the policy and stays accountable for adoption. The consultant answers a question. The fractional executive runs a function.
Should a $5M to $50M company hire a full time chief AI officer?
Rarely at that size. A full time chief AI officer costs a senior salary plus a long search, and most companies between $5M and $50M do not yet have a full time workload for one. The usual path is a fractional chief AI officer for six months to set the strategy, policy, vendors and first systems, then a named internal owner who inherits the function. Revisit a full time hire when AI runs across several departments.
How much does AI help cost for a small company?
Public ranges in 2026: a one off intensive or audit from a few hundred to a few thousand dollars, a CEO cohort programme in the low thousands, fractional chief AI officer retainers from about $4,000 to $30,000 a month depending on hours and whether the role is advisory or embedded, and implementation projects priced per build. The providers that publish prices are linked in this article, and the figure that matters is hours inside the fee.
Can a CEO do AI adoption without outside help?
Yes, if the CEO is willing to run the tools personally for a few hours a week and the company has a capable operations lead. The failure mode is a pile of pilots nobody owns. Doing it yourself works when you start with one workflow, write the one page policy first, and set a monthly review. It stops working when the leadership team needs to change, not just the CEO.
What should I ask before hiring any AI advisor?
Six questions: how many hours a month are inside the fee and who delivers them, what you own on day 30 if you stop, which recommendations earn the advisor money from a vendor, one example of a company your size, what should never go into an AI tool at your company, and how the engagement ends. A provider who cannot answer the second and the last in one sentence each is not ready for you.
Which option should a company with a board meeting in sixty days choose?
The one that produces documents a board can read: a one page AI strategy, a ranked use case list with an ROI model, a vendor scorecard and a governance policy. That is the AI Leadership Mastermind lane at Leaders ADAPT, or a fractional engagement if the company is above $10M and wants the function owned rather than planned. A cohort programme for the CEO is the wrong purchase for a board deadline.
Count your answers, then do one thing this week
Mostly A: write the one page policy this week and pick one workflow. Mostly B: list the documents your board needs and the date they need them. Mostly C: name the internal owner who will inherit the function, even if that person does not exist yet, because the job description is the first deliverable.
Which of the six questions did you not want to answer?
Start with the eight week program if the answer was mostly A.
