The One Question That Separates Good Agencies From Bad Ones

By the time you're choosing between three AI agencies, you've read three proposals, three sets of case studies, and three websites that look about as good as each other. None of that predicts what you're going to get.

That's not a complaint about agencies. It's a description of what those artifacts measure. A proposal tests the quality of a sales process and the people who write for it. A website tests how seriously a firm treats client acquisition. Clutch reviews capture how a client felt at the close of an engagement, which is the moment furthest from knowing whether the work held up. Platform certifications test platform knowledge, which is real and which almost never turns out to be the thing that failed.

Every one of these is a proxy. Each measures something adjacent to delivery and correlates with it loosely enough that a firm can score well on all four and still hand you a document nobody uses.

There's a faster test. It takes about thirty seconds to ask.

The question

Ask it on the first call, after the pitch is finished and before anyone starts talking about timelines:

"Walk me through what your engagement produces at the end of week four. Not what you'll have built. What decisions we'll be able to make that we can't make now."

Then stop talking.

What the answers sound like

A firm with real methodology answers immediately, and the answer has nouns in it. Something close to: at week four you'll have three to five workflows ranked by estimated return, each with an implementation complexity score and a note on whether your current data supports it, and a recommendation for which one to build first with the reasoning written down.

That answer is specific because they've produced the artifact before. They know what's in section three because they've written section three eleven times.

A firm without real methodology answers in one of two ways.

The first is activity. "At week four we'll be wrapping up discovery and moving into scoping for the first deliverable." That tells you they'll be busy. It doesn't tell you what you'll know.

The second is an output that isn't a decision. "You'll have a full audit of your current workflows." A document is not a decision. Plenty of companies have paid for an audit, received it, read the executive summary, and done nothing, because the audit described the situation without ranking anything or recommending anything or telling anyone what to stop doing. The document exists. The decision doesn't.

Both answers are common. Neither is dishonest. They're what you get from a firm that has documented its methodology as a sequence of phases rather than as a set of outputs, which describes most of the market.

Push once

If the answer is soft, ask one follow-up: "What specifically is in that document? Walk me through the sections."

The response to that follow-up is the actual test, and it's more reliable than the first answer, because the first answer can be rehearsed.

Firms that have produced the thing get more precise under pressure. They'll tell you section three is the return estimate with three scenarios, that the conservative case assumes no headcount change, that there's an appendix listing the workflows they looked at and rejected and why. The detail keeps arriving because it's memory, not invention.

Firms that haven't get vaguer. The answer retreats into process language: we'd work with your team to align on priorities, we'd tailor the output to your needs, it depends on what we find. Every one of those sentences is true. None of them describes a page.

The retreat is the answer. You don't need to catch anyone out, and you shouldn't try to. You just need to notice which direction the specificity moves.

Why this particular question

Most methodology is written from the inside. Steps, phases, deliverables, a diagram with arrows. That's the natural way to document how you work, and it's how nearly every agency describes itself, because the document was written for the people doing the work.

The week-four question forces a translation into the client's side of the table. Not what we do, but what you'll be able to decide. A firm that can make that translation fluently has already designed the engagement in those terms. A firm that can't has designed it around its own delivery process and is hoping the two overlap.

They often do overlap. That's the uncomfortable part. Plenty of competent firms would fail this question and still do fine work, because a good team can produce a useful outcome from a badly framed engagement. The question doesn't identify the bad agencies with any precision. What it identifies is which firms have thought about the thing you're actually buying, which is not four weeks of activity but the ability to make a decision you can't make today.

That's a smaller group than it should be. Ask ten firms and you'll get maybe two answers with nouns in them.

The part worth sitting with is what that ratio says about the market rather than about any one agency. A category where most sellers can't describe the decision their first month produces is a category where most buyers haven't asked. The question works right now because it's unexpected. It will stop working the moment enough people ask it and the answer becomes something to prepare.

---

If you want that question answered for your specific situation, the Forge Playbook does it. Answer a few questions about your business and we'll put together a tailored outline of which workflows are worth automating and what a realistic budget looks like for each. Free, no obligation, takes about three minutes.

Get your free Forge Playbook →