How to choose an AI consulting firm when you need shipped systems, not another strategy deck
Most companies between ten and five hundred million in revenue do not have an AI strategy problem. They have an AI shipping problem. They already know where the hours go. What they lack is someone who will turn that knowledge into software that runs on Monday. This guide is about picking a consulting firm with that in mind.
the short version
- Decide whether you are buying advice or buying delivery. Most mid-sized companies need delivery with just enough advice to aim it.
- AI consulting firms come in five shapes. Each is good at something different, and the label on the website rarely tells you which one you are talking to.
- Test for shipping ability by asking the firm to scope one small, real workflow and price it. Strategy-only firms will resist that.
- Write the statement of work around outcomes you can see in software, not around workshops and deliverable documents.
01advice or delivery: decide which one you are buying
"AI consulting" is used for two different products. One is advice: an assessment of where AI could help, a roadmap, a governance policy, maybe a training session. The other is delivery: systems designed, built, and running inside your operation. Firms blur the two on purpose, because advice is easier to sell and delivery is harder to do.
Here is a quick way to tell which one you need. If you cannot name three tasks in your business that eat hours every week, you need a little advice first. If you can name them already, and most operators I meet can name ten, you need delivery. Paying for a discovery phase to tell you what your own team already knows is the most common way mid-sized companies waste their first AI budget.
A test I use on myself: could the recommendation be shipped by the same people who made it? If not, the recommendation is a guess about someone else's work.
02the five kinds of AI consulting firm
The label on the website will not tell you which kind you are talking to. The way they answer your first questions will.
| Type | Good at | Watch for |
|---|---|---|
| Large management consultancy | Organization-wide change, board alignment, governance, brand cover | Senior people sell, junior people deliver; building is often subcontracted |
| Systems integrator | Large platform rollouts and enterprise software configuration | Recommends the platforms it already implements |
| Strategy boutique | Fast assessments, executive education, prioritization | Stops at the roadmap; you still need a builder |
| Software reseller with a consulting arm | Getting a specific tool live quickly | Every problem looks like that tool |
| Builder boutique | Scoping and shipping specific systems with a small senior team | Limited capacity; will not run a company-wide change program |
None of these is wrong. A company with thousands of employees and a regulator watching may genuinely need the first type. A company with a few hundred employees and a handful of painful workflows usually needs the last one.
03why mid-sized companies get poorly served
Companies in the ten to five hundred million range sit in an awkward spot. They are too complex for an off-the-shelf tool and a weekend of setup, and too small to be a priority account at a large firm. They have real systems of record, real compliance obligations, and real money at stake, but rarely a dedicated AI team.
That combination means three things for how you choose:
- You need senior attention. At your size, the person scoping the work has to understand both your P&L and your integration constraints. Ask who that person is and how much of their week you get.
- You need integration skill more than model skill. Your value is locked in your CRM, ERP, EHR, or document stores. A firm that cannot connect to those cannot help you, however good its slides on large language models are.
- You need the work to survive the consultants leaving. You probably will not hire a full AI team, so the firm must leave behind systems your existing people can run and a way to get help when something changes.
04how to test a firm for shipping ability
Every firm says it builds. Here is how to find out before you sign anything substantial.
- Hand them one real workflow. Pick something specific: invoice exceptions, intake triage, proposal drafting. Ask them to scope it in writing with a fixed price and a two-week milestone plan. Strategy-only firms will push back toward a broad assessment. Builders will ask for sample data.
- Ask what you will be able to click in the first month. The answer should be software, not a document.
- Ask who writes the code. If the answer is a partner firm, an offshore team you will never meet, or "we will staff it after signature," you are buying a handoff.
- Ask how they will measure quality. A builder will describe a test set made from your examples and a way to catch hallucinated answers. An advisor will describe a governance framework.
- Ask what they recommend against. Good firms talk you out of things. If every idea you raise is a great opportunity, you are hearing a sales process.
05what the first ninety days should look like
A healthy engagement with a builder firm has a shape you can hold them to:
- Week one or two: a short diagnosis that ranks candidate workflows by what they are worth and how hard they are. This should take days, not a quarter.
- A fixed scope for the first build: one workflow, a written definition of done, and a number that does not move unless you change the scope.
- Working software every two weeks: each demo in your environment with your data, so your operators can say "that is wrong" early.
- A first system in production: used by real people on real work, with measurement of the hours or errors it was meant to remove.
- A decision on the next item: made with evidence from the first one, not from the original roadmap.
If a firm proposes ninety days of assessment before anything is built, ask what they would build in the first thirty if you insisted. The answer tells you a lot.
06what to put in the statement of work
The SOW is where a consulting relationship becomes either accountable or vague. Insist on:
- Deliverables described as working software and measurable behavior, not as workshops held or reports delivered.
- Named people, including who builds, with a clause about substitutions.
- A fixed price per phase, and a rule that change requests are priced before work on them begins.
- Ownership of code, prompts, evaluation sets, and cloud accounts by your company from day one.
- A defined period of post-launch support and how it is priced after that.
- Data handling terms that match your obligations, especially if you touch health, financial, or customer records.
07red flags specific to consulting firms
- The maturity assessment. A scored survey of your "AI readiness" is rarely worth what it costs. Your readiness is visible the moment someone tries to build one thing.
- Tool neutrality that is not neutral. Ask how the firm is paid by the software vendors it recommends. Referral and resale arrangements are common and not always disclosed.
- Center of excellence first. Standing up a committee before a single system is in production produces policy for work that does not exist yet.
- Training as the whole engagement. Training matters, and we do it, but training alone leaves people with skills and nothing built to use them on. The best results come when people learn on the systems being built for them.
08where Insomnia Club fits, and where it does not
We are a builder boutique. Business operators and engineers sit on the same team, and I scope and ship every engagement personally. Our AI implementation work starts with the P&L, fixes the budget before work starts, and puts working software in your hands every two weeks. When a client wants their own people to carry the work forward, we pair that with AI training for teams built around the systems we shipped, and for ongoing leadership there is a fractional AI officer option.
We are a poor fit for a company-wide transformation program across dozens of business units, or for a client who wants a roadmap and plans to build it elsewhere. For the first, hire a large firm. For the second, a strategy boutique will be cheaper than us. If you want to compare the boutique path with building an internal team, read in-house AI team vs AI partner.
common questions
What does an AI consulting firm do?
It helps a company decide where AI should be used and, at the better firms, builds and operates the systems that do it. The range is wide: some firms only produce assessments and roadmaps, others write production code, and a few do both with the same people.
Do mid-sized companies need an AI strategy before hiring a consultant?
Not a formal one. You need a short list of expensive, repetitive tasks and a rough idea of what each costs today. A good consulting firm can turn that list into a ranked plan in days, and should be willing to build the first item rather than stopping at the plan.
How is an AI consulting firm different from an AI development company?
In theory, consultants advise and developers build. In practice the useful distinction is whether the people who recommend the work are the people who ship it. If they are not, expect a handoff where context gets lost and the plan gets rewritten.
How do I compare AI consulting proposals?
Send every firm the same one-page brief, ask each to scope the same small workflow with a fixed price, and compare what you would have in your hands after the first month. Proposals that only describe workshops and reports are buying time, not progress.
Should we hire a big consulting firm or a boutique?
Large firms bring brand cover, breadth, and capacity for organization-wide change programs. Boutiques bring senior attention and faster shipping. For a mid-sized company that wants a few systems built and working this year, a boutique that builds is usually the better match.