Is a paid discovery or audit phase standard before an AI automation build?

Is this a real question, or a sales objection dressed up?

It’s a real question, and buyers are asking it before they’ve talked to anyone who’d benefit from the answer. On community.n8n.io, a buyer running CRM syncs, lead routing, and AI agent workflows asked exactly this: is a paid discovery or audit phase before a build normal, and if so, what should it cost and produce? Twenty replies followed, mostly from contractors, which is worth flagging up front: the people answering had a reason to say yes.

That doesn’t make the answer wrong. It means the answer needs to stand on what it produces, not on who’s willing to say it’s normal.

What did the replies actually say it should cost?

Two anchors showed up in that thread. One contractor scoped it as “a paid, fixed deliverable, typically 5-15 hours.” Another was more specific: “4-8 hours of work, $300-$800,” ending in a dependency map and a list of every silent-failure surface, the places a workflow breaks quietly instead of loudly.

Treat those numbers as what contractors currently quote for this kind of work, not as a settled market price. Nobody in that thread was a buyer confirming they’d paid it. But the shape is consistent: a few hundred dollars, a handful of hours, a written map of what’s fragile.

So why does SimplyCubed’s Audit cost $1,500?

Because it’s a different scope, not the same thing at a markup. A $300-$800 mapping pass tells you what workflows exist and roughly where they’d break. It doesn’t tell you which one to build first, what it would cost, or whether your data and access controls are ready for an agent to touch them.

The $1,500 AI Automation Audit reviews five dimensions against your real stack:

Dimension	What it answers
Manual-work cost	Which workflows, their volume, and the annual dollar cost of running them by hand
Automation readiness	Whether your stack is API-friendly and your data is clean enough to automate
Security and governance	What can safely be handed to an agent, and how, reviewed through a security lens
First-workflow fit	Which candidate scores best on impact, feasibility, risk, and speed to value
Build vs. buy vs. partner	For the top opportunity, the honest path: build it, buy off the shelf, or hire it out

It ends in a readiness heatmap, a prioritized gaps-and-fixes list, a roadmap across your candidate workflows, and a build-ready blueprint for the one worth building first: trigger, steps, systems, a guardrails plan, success metrics, and what a build would cost and take. That’s a plan you can execute with anyone, not a list of what to worry about.

Does the fee just disappear if I don’t build with you?

No. You keep the heatmap, the fix list, the roadmap, and the blueprint either way. The $1,500 is credited in full toward an AI Automation Sprint only if you proceed within 30 days. If you don’t, the audit still stands on its own: you paid for a diagnosis, and you got one, whoever ends up building it.

That structure only works if the diagnosis is honest about which workflow is actually worth building, including telling you when the answer is “not yet,” or “build this yourself, it’s a weekend of no-code.”

When is the cheaper option the right one?

When the job is genuinely small. If a vendor can look at one well-understood workflow and scope it honestly in a short call, with a fixed price and a written success metric, that’s a real answer, and it’s cheaper than either option above. A paid discovery phase, at either price point, earns its cost when the number of workflows, the number of systems involved, or the state of your data is genuinely unclear, because that’s exactly what a 4-8 hour mapping pass isn’t scoped to resolve.

The failure mode to watch for isn’t paying for discovery. It’s paying for discovery that never produces anything you can act on, or worse, a “free” discovery call that was never discovery at all.

How do you tell a real discovery phase from a sales call wearing its name?

One test: does it end in something written down? A scope, a price, a success metric, a list of what’s risky and why. A free call that ends in “let’s get on another call to talk about next steps” wasn’t discovery. It was a funnel with extra steps, and it cost you the same 30 minutes a paid one would have, minus the deliverable.

Ask directly, before you book anything: what do I walk away with at the end of this, in writing? If the answer is vague, the call is a pitch regardless of what it’s called on the invoice.

What should you actually check before paying for one?

Four questions, whatever the vendor and whatever the price:

1. What’s the fixed scope, in hours or in deliverables? “We’ll take a look” isn’t a scope. “5-15 hours, ending in a dependency map” is.
2. What do I own at the end, regardless of what I decide next? A real discovery phase hands you something portable: a document, a map, a blueprint. If the output only exists in the vendor’s head or their proposal deck, it isn’t a deliverable.
3. Does the fee credit toward a build, and under what condition? Ours credits in full within 30 days. Ask what the window is and what happens if you miss it.
4. Does the cost match the scope? A five-hour mapping pass and a full stack, security, and blueprint review are different jobs. The price should track which one you’re actually buying.

If a vendor can’t answer all four in the same call where they’re pitching you, that’s the more useful signal than the price itself.