Custom GPTs for small business: what they actually do.

Everyone's heard "custom GPT" by now. Plenty of people still aren't sure what one is. Here's the short version.

A custom GPT is a version of ChatGPT (or Claude, or Gemini) that you've pre-loaded with your specific information and instructions. Same underlying AI everyone else is using. But instead of starting every conversation from zero, it already knows your business.

Think of it like training a new hire

When you hire someone new, you spend the first week explaining how your business works. Your pricing. Your tone. Your most common customer questions. Your services. Your "we don't do that" list.

A custom GPT is the same thing, but you only have to do it once. You upload a few documents — your brand voice guide, your FAQ, your pricing PDF, past proposals, past emails — and write a short set of instructions. From that moment on, every conversation with the GPT happens with that context loaded.

Ongoing cost: usually just the ChatGPT or Claude subscription you may already pay for. How much time it saves depends entirely on which task you point it at, which is why picking the task matters more than building the tool.

Three examples of what one can do

These are examples of the kind of GPT that can be built, not client stories. They're common shapes.

1. A "new inquiry" triage assistant

Say a small-business owner gets ten to fifteen inbound emails a week from prospects. Some are tire-kickers, some are real, and sorting them eats a morning.

A triage GPT takes a pasted inquiry, reads it, sorts it (hot lead, curious, wrong fit), drafts a reply in her voice, and flags anything urgent. She reviews, tweaks, and sends. The judgment stays hers. The blank page doesn't.

2. A proposal drafter

Picture a coach who writes every proposal from scratch. Same six paragraphs, slightly tuned each time.

A proposal GPT loaded with her past proposals, her packages, and her pricing can write a first draft from a three-line brief: client name, what they want, budget. She edits for tone and specifics instead of starting cold.

3. A "what did we decide" lookup

Imagine a nonprofit director with three years of board minutes scattered across Google Drive. Every time she needs to remember what the board decided about a policy, she digs.

Load those minutes into a GPT and she can ask "what did we decide about grant reporting in 2024?" and get a direct answer that names the meeting, so she can check the source.

What they're not

Custom GPTs aren't autonomous agents. They don't send emails on their own. They don't make decisions without you. They don't replace a human — they make one human faster at the same tasks they were already doing.

They also aren't magic. If your business is a mess of undocumented processes in your head, the GPT will reflect that. Garbage in, garbage out still applies. Part of the setup is getting your processes written down for the first time — which is often the biggest win, independent of the AI.

What it costs to have me build one

It starts with the half-day AI workflow audit ($750). That's a written plan, not a build: the three to five automations worth doing first, what each should save, what each costs to set up and run, and the order to build them in.

Builds are scoped after the audit. If a custom GPT makes the list, the build-out is a fixed price I quote in writing, typically $1,500 to $3,500 depending on what's in it. After that, the AI coaching retainer ($500/month, month to month) covers refining the GPTs you have, training your staff, new-tool guidance, and two 30-minute office-hour calls a month.

The real value isn't in the GPT. It's in figuring out which task to automate in the first place. If you want a head start on that, the free AI Time-Leak Checklist is a good place to start.

---

Curious whether AI could save you time in your specific business? Book a free call and I'll tell you straight whether it's worth the investment.