AI agency startup checklist: what you need before your first client

Before selling to your first client you need four things: a legal entity that can invoice, a service you can demonstrate working, a price you have decided in advance, and a list of businesses to approach. Everything else — refined branding, a complete service range, reporting dashboards — can wait, and waiting for it is the most common reason people never start. This checklist covers what genuinely has to exist, in the order it should exist.

It is written to be worked through rather than read. Items are grouped by whether they block your first sale, your first delivery, or neither.

The short version

• Four things block a first sale: an entity, a demonstration, a price, and a prospect list.
• A second group blocks delivery rather than selling, and can be arranged in parallel.
• Most of what people prepare first blocks nothing at all.
• Decide pricing before the conversation, not during it.
• Know what your monthly fee includes before a client asks.

What blocks your first sale?

A legal entity that can invoice

A registered company or sole trader status, a business bank account and a way to raise an invoice. Formation takes about a day. You can hold exploratory conversations before this exists, but you cannot take money.

A service you can demonstrate working

Not described — working. A prospect needs to hear the receptionist answer, or watch the chatbot handle a question. This is the single highest-leverage item and the one people delay longest, usually by preparing other things first.

One service demonstrated confidently beats three explained vaguely. What an AI automation agency actually sells covers choosing which.

A price you decided before the conversation

A setup fee, a monthly fee, and a clear position on what happens if usage is heavy. Deciding these live in front of a prospect produces numbers you regret, and raising a price on an existing client is much harder than setting it correctly. Pricing AI services for local businesses covers arriving at them.

A list of businesses to approach

Specific names, not a category. "Plumbers in Leeds" is a category; forty named businesses with contact details is a list. The difference determines whether outreach happens on a given morning.

What blocks delivery rather than selling?

These can be arranged while you are already having conversations, and holding up outreach for them is a mistake.

• An onboarding process. What you need from a client to configure their service: opening hours, services, what counts as urgent, where leads should go.
• A written scope. What the monthly fee includes, how many changes, what counts as new work. This prevents the slow slide into unpaid support.
• Data protection basics. You will be handling personal data on a client's behalf. Establish your position with the ICO and have the client arrangement documented before the first deployment rather than after.
• A support route. How a client reaches you and what response they should expect.
• Insurance. Professional indemnity is not legally required but business clients increasingly ask.

What blocks nothing?

This list matters as much as the others, because these are where preparation time disappears.

A credible site matters — a prospect will look you up after a conversation. A perfect one does not, and the gap between the two is weeks.

What should you have ready for the first conversation?

Five things, all of which fit on one page:

1. A one-sentence description of what you do, in terms of the client's problem rather than the technology.
2. Three questions that establish whether they have that problem — what happens when a call is missed, how quotes get followed up, what a job is worth.
3. The demonstration, tested and reliable, that runs in under two minutes.
4. Your prices, and the reasoning if asked.
5. A clear next step — not "I'll send some information", but a specific thing happening on a specific day.

Handling objections to AI services covers what tends to come back, and it is worth having positions on those before you need them.

What should you have ready before the first client goes live?

Once someone says yes, a different set matters. Test the whole path end to end before it faces real customers — a caller reaching the receptionist, the details arriving where they should, the client seeing them. A service that fails in front of a real customer costs more than the client is worth.

Agree what happens when something goes wrong, before it does. Which is not pessimism: every service has an off day, and the difference between a client who tolerates it and one who cancels is whether they knew who to call.

Set a review date. The most common cause of early cancellation is a service configured once and never revisited. Booking a check-in at week four costs you an hour and is the single most effective retention step available. Building recurring service packages covers why.

What order should you do all this in?

Roughly: decide who you are approaching, get a demonstration working, decide prices, build the list, start conversations. Arrange the delivery items while those conversations are happening, and ignore the third list until something in it starts blocking you.

The instinct is to complete everything before speaking to anyone. It feels responsible and it is the slowest possible route, because the things you learn from ten conversations change what you would have built anyway.

If you have not chosen an entry route, buying versus building determines how long the demonstration takes to exist, and how long it takes to start covers the two timelines. The live demo shows what a working demonstration looks like, and what's included lists what an assembled agency provides against this checklist.