Strategy · October 27, 2026

Should You Hire an AI Agency? 9 Questions to Ask First

There are now thousands of AI agencies, and they sell very different things. Nine questions to ask before you hire one, what good answers sound like, and the warning signs.

A few years ago there were almost no AI agencies. Now there are thousands, from solo freelancers to large consultancies. They use similar words, like agents, automation and AI transformation, but they sell very different things.

Before you hire one, it helps to know what you are actually buying. These nine questions will tell you most of what you need to know, including whether you need an agency at all.

First: what do you want AI to do?

Most AI agencies fall into one of three groups:

Automation agencies make the work you already do faster and cheaper: email handling, document processing, workflows, internal assistants.
Build agencies build an app or product to your specification, often quickly with AI coding tools.
Product partners help decide what to build, build it, and stay involved in making it succeed.

If your goal is cutting costs, the first group is usually right. If you know exactly what you want built, the second. If you want to grow with something new and are not sure what it should be, the third. We compare them in detail on our comparison pages.

The nine questions

1. "What will this change for my business, in numbers?"

A good agency talks about outcomes: hours saved, enquiries converted, clients kept, revenue earned. Be wary of answers that only describe technology.

2. "Who usually hires you?"

An agency that mostly builds MVPs for startups works differently from one that works with established businesses. Ask for examples similar to your business, and ask whether you can speak to a past client.

3. "What happens before I pay?"

Some agencies start with a paid discovery phase. Some start building right away. Some show you a working prototype first. Know which, and what you get for your money at each step.

4. "How do you get paid, and does any of it depend on results?"

Fixed price, hourly, per sprint, retainer or a share of results. None is wrong, but they create different incentives. An agency paid only for time does equally well whether your project succeeds or not.

5. "Who owns what you build?"

You should own your product, your data and ideally the code. Get it in writing.

6. "How do you handle client data and security?"

If the project touches customer or client data, ask where it is stored, who can access it, and how it is tested. Vague answers are a warning sign. Our security checklist gives you good follow-up questions.

7. "What happens after launch?"

Who fixes things when they break? Who improves the product? How fast do they respond, and what does it cost? Many agencies are gone the day the project ships.

8. "What would you tell me not to do?"

A good partner says no sometimes, including "you do not need us for this". If everything you suggest is a great idea, be careful.

9. "Can I end the relationship easily?"

Long contracts and hard-to-leave setups protect the agency, not you. Ask about notice periods, and what you take with you if you leave.

Warning signs

They promise to "transform your business with AI" without asking much about your business.
Everything is described in tools and features, nothing in results.
They cannot show anything similar they have built.
The price is fixed before anyone understands the problem.
Security and data questions get vague answers.
They will not put ownership in writing.

Do you need an agency at all?

Not always. If you want a simple internal tool or a prototype to test an idea, you can often build it yourself with tools like Lovable. See Veyra vs building it yourself. If you need someone permanent, compare with hiring in-house.

How we answer these questions

Since we asked them, here are our own answers in short:

Who hires us: established service businesses with revenue and customers.
Before you pay: a working prototype, free.
How we are paid: one fixed build fee agreed after the prototype, then a share of what the product earns.
Ownership: you own what we build.
After launch: we stay on as head of product.
Ending it: you can end the revenue share whenever you want. It runs for three more months, then stops.