How do you tell if a consultancy is really AI-native?

Almost every supplier now says they are AI-native. The phrase costs nothing to print, and few mean the same thing by it. The honest test is simple: look at the delivery model, not the tool list. A real one has changed how the work gets estimated, built, priced and owned. A bolted-on one has bought a few AI seats and kept everything else the same. What follows is how to tell them apart before you sign: the questions worth putting in your RFP, and how to read the answers.

AI-native is a delivery model, not a tool

Buying a few AI seats for the team is not being AI-native. The question is whether AI has changed how the work is estimated, built, reviewed and priced. Or whether it just sits on top of the same six-person team and the same six-month plan.

The distinction matters because the tools on their own do not move your outcome. In 2025, controlled studies found experienced developers were often slower with AI, not faster, even while they felt faster. So “we use AI” tells you very little. What you are testing for is whether the model behind the work has actually changed. The questions below are built to surface that.

Can you show me how your own team uses AI?

Ask them to show it, live, on their own work. Not a client demo, their own estimating, their own code review, their own testing.

A strong answer is a screen share. They walk you through where agents do the typing and where a senior person makes the call, because they work this way every day. A red flag is a supplier who can talk fluently about AI for clients but cannot show it in their own delivery. If it is real, it is in front of you in minutes.

How small is the team, and how soon will I see working software?

An AI-native model shows up as smaller teams and weeks to first value, not months. That is the whole point of the model, so it should be visible in the shape of the proposal.

A strong answer is two or three senior people and working software you can use within weeks. A red flag is the same six-to-eight-person team and a long discovery phase before anything runs. If the team and the timeline look like a project from three years ago, the AI has not reached the part that affects you.

If AI makes you faster, who keeps the saving?

This is the question most pitches dodge, so ask it plainly. If AI lets a supplier do in three weeks what used to take three months, where does that gain go: to you, or to their invoice?

A strong answer is a price fixed to the outcome up front. Faster delivery is then their reward for being good at the work, and the certainty on cost is yours. A red flag is a supplier who cannot say how any of the saving reaches you. If their speed only ever lands as their margin, you are paying for efficiency you never see.

Walk me through a recent build in plain terms

Pick one of their projects and ask them to explain it without jargon: the problem, the calls they made, what they would do differently.

A strong answer is plain and specific. They know why they built what they built, and they are honest about what went wrong. A red flag is vagueness, buzzwords, or senior names in the pitch with junior hands on the actual work. If they cannot explain a project simply, they either did not understand it or would rather you did not.

What have you stopped doing because of AI?

A real change in the model means some old things have gone. Ask what.

A strong answer names specifics: work they no longer do by hand, steps they have retired, roles that have changed. A red flag is “AI is additive, it just makes us better.” That usually means the marketing has moved and the model has not. Genuine change leaves a trail of things you used to do and no longer need to.

Who owns the outcome, and who do I hold to account?

A computer cannot be held accountable, so a person has to be. Find out who, by name, before the work starts.

A strong answer is a named senior person who owns the result, working to a fixed price for a defined piece of software. If a call turns out wrong, it is theirs to fix, not yours to fund. A red flag is accountability spread thin across a tool and a rotating team, where no single person carries the outcome. You want someone on your side of the table you can hold to it.

Are you still billing time and materials, on the old rate card?

Ask how they price the work, and whether the rate card has moved at all. This is where AI either reaches the commercial model or quietly does not.

A strong answer is a model that has moved: pricing tied to the outcome, or rate cards that have changed to reflect what a smaller AI-native team now delivers. A red flag is the same time-and-materials model on the same day rates as a few years ago. Think it through. A team that is genuinely faster, still billing by the hour at the old rate, earns less for the same result. The billable hour rewards delay. If nothing about the pricing has changed, the model behind it probably has not either.

Do they talk about responsible AI, or just the shiny demo?

Ask what could go wrong, and how they handle it. A supplier who only shows you how clever their AI is, and never how they keep it safe, has told you something.

A strong answer covers the unglamorous side: where your data goes and what is logged, and how they check AI-written code before it ships. It also covers whether a decision the software makes can be explained, rather than just trusted. That last part, explainable AI, matters the moment the software touches money, people, or anything regulated. A red flag is all capability and no care: no policy on client data, no review step, and a shrug when you ask why the system did what it did. Confidence about the upside and silence on the risk is not maturity, it is a sales pitch.

A quick checklist for the shortlist

Run each supplier through these. A no is where the work, or the awkward conversation, needs to be.

• Can they show you AI inside their own delivery, not just in the pitch?
• Is the team smaller, and the first working software weeks away, not months?
• Will the saving from their speed reach you, or only their margin?
• Can they explain a recent project in plain terms, with the same people who will do your work?
• Can they name something they have stopped doing because of AI?
• Is there a named senior person who owns the result?
• Has the rate card or billing model moved, or is it the same time and materials as before?
• Do they talk about responsible AI and explaining its decisions, not just how clever it is?

Where that leaves you

Most of these questions have nothing to do with AI as a feature. They test one thing: has the supplier actually rebuilt how they deliver, or have they painted the word over the old way of working.

If a supplier clears them, you are hiring a model built for the outcome, not a team built for the timesheet. That is the work we do: smaller senior teams, working software in weeks, a fixed price, and a name against the result.

If that is what you are looking for, start a conversation →.