The problem isn't AI. It's who you're paying to implement it.
The AI consulting market grew faster than firms' ability to deliver real results. Today there are hundreds of consultancies offering "AI transformation" — and most deliver the same thing: an extensive diagnostic, a PowerPoint roadmap, and an MVP that no one ends up using in production.
For a CFO or COO at a company of 50 to 200 people, the cost of getting this wrong isn't only financial. It's your team's time, your internal credibility, and a window for change that closes.
Before signing any contract with an AI consultancy, ask these three questions. The answers will tell you everything you need to know.
Question 1: What is actually running in production when the project ends?
This question seems obvious. It isn't.
Most consultancies deliver documents, workshops, and prototypes. Some deliver MVPs. Very few deliver systems that your team operates autonomously in day-to-day business.
The difference is critical. A prototype demonstrates that something is possible. A system in production generates value. These are not the same thing.
When you ask this question, watch for two signals:
Positive signal: The consultancy describes concrete cases where an agent or automated workflow is running today, with usage metrics and measurable results. They can name the process, the team operating it, and what happened to time or error rates.
Warning signal: The answer centers on deliverables — "a roadmap," "a diagnostic," "a validated prototype." If the final output is a document, you are paying for analysis, not for results.
A mid-size financial services firm, for example, might hire a consultancy to automate its monthly reconciliation process. If after three months the finance team is still doing the process by hand because "the system isn't ready for production," the project failed — regardless of what the final report says.
Question 2: How will my team be equipped to operate and expand what you build?
This is the question most consultancies avoid answering clearly.
There are two business models in the market, and it's important to understand which one you're being sold:
Dependency model: The consultancy builds, the consultancy operates, the consultancy charges for every change. Your team never learns to build. You need the consultancy indefinitely.
Enablement model: The consultancy builds alongside your team, transfers the method and the knowledge, and by the end of the project your team can build without external help.
The second model is less profitable for the consultancy in the short term. That's why few genuinely offer it.
When you ask this question, push for specifics: What exactly does my team learn? At what point in the project? How is it measured that the learning actually happened? What happens if I want to add a new agent six months from now without re-engaging you?
If the answer is vague — "we do a knowledge transfer at the end" — that's a warning signal. End-of-project knowledge transfer is, in practice, almost always insufficient.
Question 3: How do you measure ROI, and when will I see it?
This question has two parts, and both matter.
The first part — how they measure ROI — reveals whether the consultancy thinks in business terms or technology terms. A business-oriented consultancy can tell you, before work begins, which metrics it will move, in which direction, and within what range. Not with surgical precision, but with reasonable hypotheses grounded in comparable cases.
For example: if the goal is to automate the monthly reporting process for a five-person finance team that currently spends between 40 and 60 hours a month consolidating data manually, a serious consultancy can estimate that the process can be reduced to 10–15 hours per month within 6 to 10 weeks, using an agent that consolidates sources, flags anomalies, and generates the report draft. That's between 25 and 45 hours recovered each month — time that can be redirected to higher-value analysis.
The second part — when you'll see it — is equally important. If the consultancy talks about results at 12 or 18 months, ask yourself what is happening in the first six. A well-designed project should show measurable value before the three-month mark. Not the full ROI, but clear evidence that the system works and the team is using it.
If they can't answer either part with specificity, they don't have a method. They have a proposal.
What these three questions reveal together
These questions aren't a technical filter. They're a filter for commercial maturity.
A consultancy that can answer all three clearly — what remains in production, how your team is equipped, and when you see ROI — has a proven method and is oriented toward business outcomes.
A consultancy that responds with generalities, deflects toward technology, or promises results without being able to quantify them is selling you a process, not an outcome.
There are few firms serving companies of 50 to 200 people with this level of specificity. Most available options are designed for large enterprises — with budgets, timelines, and structures that don't apply to your organization.
Before you sign, ask the questions
If you are currently evaluating an AI consultancy, bring these three questions to your next meeting. The answers will tell you more than any commercial proposal.
If you want to compare what you hear against a concrete, results-oriented approach, OuroAI offers a free, no-commitment diagnostic. In 15 minutes, you can assess whether your case is a fit for our model — and if it isn't, we'll tell you that too.
Request the diagnostic through the form on this page.
