How to Evaluate AI Consulting Firms Before You Sign Anything

You have had three calls with AI consulting firms this month. Each one had a polished deck, a partner who asked good questions, and a proposal that arrived two days later with a scope that somehow cost more than your original budget. You are not sure any of them actually understood what you need.

This article is not about what AI consulting is. For a full breakdown of how engagements are scoped, priced and delivered, read our guide to AI consulting services. That covers the overview. This piece goes one level deeper: what questions to ask, what proof to demand, and what answers should send you toward the door.

Who should be reading this

This is for an operations director, COO or founder at a company doing somewhere between $5M and $100M in revenue. You are losing real hours to manual work. Someone on your team is copy pasting between systems, chasing approvals by email, or building reports by hand every week. You have budget to fix it. You are not evaluating AI out of curiosity. You need it working inside 90 days, not inside a roadmap.

This is not for you if you are still deciding whether AI is relevant to your business. Start with a smaller experiment first. And it is not for you if you need a system that touches hundreds of users across multiple countries on day one. That is a different kind of engagement and a different kind of firm.

What separates firms that ship from firms that advise

Most of what gets called AI consulting is strategy work: workshops, assessments, technology roadmaps, vendor selection reports. That work has its place. But if your problem is a manual process burning 30 hours a week, a roadmap does not save those hours. A running system does.

The first question to ask any firm is simple: can you show me something you built that is in production today?

Not a demo. Not a prototype. A system a real team uses for real work. Ask what the system replaced, what changed in the first 60 days, and whether you can speak to the team that runs it. If the answer is a case study PDF with a logo and a quote, that is not proof. That is marketing.

Firms that build keep those answers short, because the work speaks for itself. Firms that advise keep those answers long, because they are filling space where proof should be.

The five questions that expose the difference

Bring these into every evaluation call.

1. What is the first thing we would build, and how do we measure whether it worked? A firm that knows what it is doing can name a specific process, a specific output, and a specific number that changes. If the answer is "that depends on our discovery phase," ask how long discovery takes and what it produces. Weeks of discovery before anything is scoped is a warning sign.

2. What does your team actually build versus what do you hand off to a third party platform? Many firms resell automation tools with a setup fee attached. That is not implementation. Ask whether they write code, own the deployment, and can modify the system after go live without billing you for a change order every time.

3. What projects have you turned down, and why? This is the question most buyers never ask. A firm that has turned down work, because the client did not actually need it, is a firm you can trust with your budget. A firm that has never turned anything down is a firm that sells before it thinks.

4. What does the first two weeks look like, from our side? Ask what you will need to provide: access, documentation, time from your team. Vague answers here mean the firm does not yet know what it is building. Good firms can describe the first sprint before the contract is signed.

5. What does the system look like if we part ways after six months? You own it. You can hand it to your own developers. It does not vanish if the firm's pricing changes. If those answers are not immediate and clear, you are about to rent, not build.

What real proof looks like in practice

AlbTech has shipped AI agents into production across several industries. One built for a pharmaceutical distribution operation handles the full order intake flow: customer orders arrive on WhatsApp at any hour, the agent reads the message, resolves it against the product catalogue, confirms ambiguities with the customer, and writes the finished order into the ERP. Before it ran, four people shared that work. Thirty hours of manual entry per week came out of the operation.

That is what proof looks like. A named process, a measurable before, a measurable after, and a system still running. Ask for that standard from every firm you evaluate.

If you want to see how this kind of engagement is scoped before you commit to a call, book a free system review and we will tell you inside 30 minutes whether there is a fit.

The objection worth addressing directly: what if the firm is based outside the US or UK?

Many buyers in the US and UK reflexively filter for local firms. The reasoning is understandable: time zones, accountability, and the assumption that proximity equals quality.

Here is the honest version of that concern. Time zones are solved by async communication and clear delivery agreements. Accountability comes from contracts with defined outputs and payment tied to delivery, not from a firm being reachable by train. Quality comes from the work.

The practical question is not where the firm is headquartered. It is whether your system will be delivered by people who built it themselves, who own it, and who will pick up a problem at 9am your time. Ask for those answers directly. If you get them, the postcode does not matter.

How to score your shortlist

What happens when you reach out to AlbTech

When you message us, the first conversation is a system review: 30 minutes, no deck, no upsell. We ask you to describe the most repetitive process in your operation. We tell you whether AI addresses it, what we would build first, how we would measure it, and what it realistically takes from your team to get there.

If there is no fit, we say so. We have done it before.

If there is a fit, we send a scoped proposal within five business days. The scope names one bottleneck, one measurable output, and a timeline. You decide from there.

Message DEMO to book that review. It costs you 30 minutes and tells you exactly where you stand.