What should you ask before signing with an AI consultant?

A person signing a printed contract with a pen across a desk

AI CONSULTING · VENDOR CHECKLIST · DUE DILIGENCE

AI consultants are currently selling faster than most companies can evaluate. A short list of plain questions separates the ones worth paying from the ones worth a polite no. We sell AI services ourselves, so this is the checklist we would want a buyer to use on us.

Before signing, ask five things: what exactly will exist when you leave, who owns it, where our data goes, what happens when the underlying model changes, and what this costs in year two. A good consultant answers in writing, with names and numbers. Vague answers are the answer.

Scope and ownership

What will concretely exist when the engagement ends - a running system, documentation, trained people, or a slide deck? Who owns the prompts, workflows and code? Can your team operate it without them? If the answer to the last question is no, you are not buying a solution, you are renting their presence.

Data and compliance

Where does your data go, and which models and subprocessors touch it? If your company is in Switzerland and personal data is involved, Swiss data protection law applies to you regardless of what the contract says. And if you sell into the EU, ask which EU AI Act duties apply to your setup: its transparency rules for certain AI systems - such as telling people they are interacting with an AI - started applying on 2 August 2026, and the Act can also reach companies outside the EU. "We take care of compliance" is not an answer; ask them to show you how, in the contract.

People and change

Who actually does the work - the partner who pitched, or someone you have not met? And what happens when the underlying model or vendor changes, which it will? A consultant who has no answer to that is selling you a snapshot.

Money

Fixed price or time and materials? What is included, and what is billed as extra? What does it cost in year two, after the project becomes operations? The year-two question is the one that separates honest quotes from landing strategies.

The takeaway: a good consultant answers in writing, with names and numbers. Vague answers are the answer.

Sources

European Commission - AI Act transparency obligations, applying from 2 August 2026

SECO - AI gains ground among Swiss SMEs (5 November 2025)

Last updated: 24 October 2026.

TL;DR: Ask what will exist when the consultant leaves, who owns it, where your data goes, what happens when the model changes, and what year two costs. Good consultants answer in writing with names and numbers - vague answers are the answer.

Reading time: ~3 min

Want to try these questions on us? We are happy to answer every one in writing before you sign anything. Discuss a project: eli5a.com/contact

Previous
Previous

How do you turn your company documents into an AI knowledge base?

Next
Next

What is RAG, and why does it matter for your company data?