What an AI business app actually costs

Two hosts walk through the four pricing tiers — audit, prototype, pilot, production — what separates them, why the audit is the only one with a published figure, and what a client owns at handover.

A: "Contact us for a quote" is a fine answer when the scope is genuinely unknown. For an AI business app, scope is basically the whole question.

B: Right — two businesses asking for what sounds like the same system can be miles apart in price once you know how many systems it touches and how much of the logic is actually custom.

A: So here's what we do publish. Four tiers, and one number that doesn't depend on your workflow at all. Workflow Audit — free, credited against the build if it proceeds.

B: Prototype Sprint, Pilot Build, Production Build — all contact for pricing. Not because we're being cagey. Because a published band for those three would either be wide enough to be useless or narrow enough to be wrong for most people reading it.

A: Ongoing support is the same story — priced at handover, because it depends entirely on what got built and how it needs to keep running.

B: What actually separates the tiers, though? It's not ambition.

A: It's scope and risk. A prototype answers "does this work at all, for one real case." A pilot answers "does a real team actually use this, and what does the run log show over real weeks." A production build is what you commission once the pilot's already answered those questions and the workflow needs to hold up at volume.

B: Most enquiries should start at the smallest tier that actually answers the question in front of them. Starting at Production Build before a prototype's ever been in front of a real user is usually buying certainty you haven't earned yet.

A: Why's the audit free and disclosed, when nothing else is?

B: Because it's the step most likely to end a project honestly — to conclude a workflow doesn't need building yet, or needs something smaller than what was asked for. Charging for the step most likely to tell you not to spend money creates a bad incentive.

A: Free, credited against the build if one happens, removes that incentive. And it's the one number that's true no matter whose workflow we're looking at — which is exactly what the other three tiers aren't, until the audit's actually happened.

B: What does a client actually own at handover?

A: The code. The data exports. The credentials. The documentation. Access is removed. That's not a special arrangement negotiated per client — it's the default on every engagement.

B: A vendor with permanent access to a system it no longer maintains is a liability nobody wants to be.

A: A couple of engagement details worth naming too — an NDA before scoping if the client wants one, a written SOW per project, an advance up front followed by milestone payments. Working hours shift per project to overlap the client's day, for both US and UAE clients — there's no fixed support window.

B: And one honest limit, stated plainly — Arabic support is model-level only, and hasn't been tested specifically against Gulf Arabic. It can be scoped as a pilot deliverable with accuracy testing and an English fallback. It's not, today, a proven strength.

A: One thing to check this week.

B: If you've got a workflow you've been meaning to ask about, the free audit is the actual first step — not a sales call, an honest scoping of whether it's worth building at all. The number comes after that, once it's real.

A: Written version's at atinnovators.in, under insights. That's the season, as far as this batch goes.