A founder who runs an AI agency for small businesses posted something this week that's worth keeping near your desk.
He'd spent months pitching restaurants, gyms, real estate offices, and clinics. Every meeting felt promising. Genuine questions. Nods. "Can you send more information?"
Then silence.
After enough ghosting, he figured out what was happening. He was selling the engine. His clients wanted the destination.
The pitch that got him meetings was about architecture, models, automation workflows. The pitch that got him clients was about time, money, and problems that hadn't been solved.
He put it plainly: "These people are doing payroll at midnight and answering customer emails on weekends. They do not have the bandwidth to care what RAG stands for. Nor should they."
He's right. And if you're a small business owner sitting across from an AI vendor right now — or about to be — you should know this is happening on a massive scale. A lot of pitches are being given in a language you never agreed to learn.
Here's how to stop that conversation in its tracks and find out fast whether you're talking to someone worth your time.
The Three Questions
1. "What specific problem does this solve for a business like mine?"
Not "what does your tool do." Not "what features does it have." Specifically: what goes wrong for a business in your industry or at your stage, and how does this tool stop that from happening?
If the answer involves any of the following without further translation, you have your answer:
- "Multi-agent workflows"
- "LLM-powered automation"
- "RAG pipelines"
- "Fine-tuned vertical models"
- "API integrations"
These are tech descriptions. They are not problem descriptions. A vendor who can only speak in tech terms has not done the work of understanding your business. That's not a technology gap — that's a preparation gap.
The vendors worth talking to will answer this question with something you recognize from your own week. "It catches the customer emails that fall through the cracks." "It drafts your follow-up messages so you don't have to." "It flags when you're about to run low on inventory before it becomes a crisis."
If they can't say what the problem is in plain language, they haven't earned the right to pitch you a solution.
2. "Can you tell me what it won't do?"
This question does two things.
First, it tests honesty. A good vendor — one who has actually worked with businesses like yours — knows the limits of their tool. They'll tell you: "It's great for X, but it struggles with Y, and if you have Z in your workflow you'll need to handle that differently."
A vendor who answers this question with silence, deflection, or another round of features is one who either hasn't been asked this before (they've only talked to early adopters who didn't push back) or is more interested in closing than in whether the tool actually fits.
Second, it gives you the real picture. Every AI tool right now has real limitations. Context windows, integration gaps, accuracy on certain types of content, latency, cost at scale. The ones that work for small businesses are the ones where the limitations don't land on your specific use case.
You can only figure that out if someone tells you what they are.
3. "What does a win look like in 90 days, and how do we measure it?"
This is the question that separates a pitch from a commitment.
Any tool worth paying for should come with a definition of success. Not vague success ("you'll save time," "your team will work more efficiently") — specific success. "Your average response time to customer inquiries drops below four hours." "You process invoices in half the time." "You're generating three times as many social posts with the same team."
If a vendor can't answer this question, it's not because success is hard to define. It's because they haven't thought about your success — only their sale.
This question also gives you something important: an exit condition. If 90 days pass and the agreed measure hasn't moved, you have a clear, unemotional reason to stop. No awkwardness. No "well, we just need more time." You agreed on a number. You check the number.
What This Is Really About
The AI vendor-to-small-business conversation is broken right now because one side is speaking tech and the other is speaking pain.
You're doing payroll at midnight. You're managing customer complaints before your coffee gets cold. You're watching margins. You don't have time for a demo that makes you feel dumb for not knowing what "fine-tuning" means.
The good news: the vendors who've figured this out are genuinely good. They've translated their technology into your language. They show up with a specific problem in mind, a clear sense of whether it fits your situation, honest limits, and a definition of what winning looks like.
The ones who haven't figured it out aren't bad people. They're just too close to the tech and too far from the business.
These three questions will sort them in about ten minutes.
One More Thing
A vendor who's prepared will welcome these questions. Someone who's been in the room with businesses like yours will light up when you ask "what does a win look like in 90 days?" — because they've had that conversation before and they know the answer.
Someone who hasn't will get uncomfortable. Not because the questions are hard. Because they've never been asked to leave the pitch deck behind and talk about your actual business.
That discomfort is information.
Use it.
Published Sunday, April 26, 2026. The Useful Daily covers AI for people running real businesses.