I built a working CRM just by talking to my AI (prompt included)

You can build a working CRM just by talking to your AI agent. In AgentDocs you describe the people and details you want to track, and your agent creates a real Database it can read and update for you, with no setup screen, no spreadsheet formulas and no per-seat fee. The exact prompt that does it is at the end of this post.

What is a CRM, really?

Strip away the sales jargon and a CRM is a short answer to two questions: who do I know, and what did I last say to them? For most freelancers, founders and side-hustlers that's it. You don't need pipelines, lead scoring or a 12-step onboarding flow. You need a tidy list of people, a note on the last thing that happened, and a nudge when someone's gone quiet.

The problem isn't the idea, it's the upkeep. A list is only useful if it's current, and keeping it current by hand is exactly the chore everyone abandons by week three.

Why not just use a spreadsheet?

A spreadsheet is a fine place to keep a contact list. It's a tiring place to maintain one. Every update is manual: open the file, find the row, type the note, fix the date. A Notion- or Coda-style workspace gives you nicer views, but you're still the one doing the data entry.

With AgentDocs you stop being the data-entry person. You tell your agent what happened in a sentence, and it does the row-finding and typing. The table stays current because keeping it current is now a one-line message, not a task.

How do you build a CRM in AgentDocs?

Three steps, all in plain language:

  1. Connect AgentDocs to your AI agent. It works in Claude and ChatGPT over a single link, with no install and no card.
  2. Describe the CRM you want. Tell your agent the kind of people you're tracking and the few details that matter (name, company, how you met, last contact, next step). It creates a Database with those fields: a real structured table, not a wall of text.
  3. Start talking to it. "Add Marcus from the podcast, we spoke today, he wants the pricing deck next week." Your agent files it in the right columns.

You never open a settings panel. The structure comes from the conversation.

What can your agent do with it day to day?

This is where it earns its place. Once the Database exists, the everyday moves are all sentences:

  • Log a contact: "I just got off a call with Priya, moving to a paid trial, follow up in two weeks."
  • Find the gaps: "Who haven't I spoken to in over a month?"
  • Prep your day: "Who am I meant to be following up with this week, and what was the last thing we discussed?"
  • Keep it clean: "Merge the two entries for the Hamilton account."

Because your agent both reads and writes the same Database, you can also pull the list into a Sheet for a quick export, or have it draft a follow-up message straight from the notes, with no copy-paste between tools.

Does the data stay put?

Yes. It lives in a real Database inside AgentDocs, built on Stash, with the same Google sign-in as the rest of your account. It's not trapped in a chat thread. Ask Claude today and ChatGPT tomorrow and they're both looking at the same up-to-date list.

What does it cost?

AgentDocs has a free tier with no per-seat fees. You connect over one link, with no install and no card, and there's a Pro tier with generous fair-use limits for heavier use. Billing isn't switched on yet, so it's a low-friction way to try the idea on your own contacts. It's a new product, so pricing may change as it settles.

The exact prompt

Paste this to your AI agent once AgentDocs is connected. Edit the fields in brackets to match how you work:

Build me a simple CRM in AgentDocs.

Create a Database called "CRM" with these columns:
- Name
- Company / context
- How we met
- Status [e.g. New, In conversation, Customer, Dormant]
- Last contact (date)
- Last thing discussed
- Next step
- Next step due (date)

Rules for keeping it current:
- When I tell you about a conversation, find the right person (or add them),
  and update Last contact, Last thing discussed, Status and Next step.
- If I name someone who isn't in the table yet, add them.
- Keep one row per person, and merge duplicates if you spot them.

Then confirm the table is set up and ask me for my first few contacts.

That's the whole thing. No template to download, no fields to wire up. You describe it, your agent builds it, and from then on you just talk.

This is the first in our "Built in AgentDocs" series, real things built by talking to an AI agent, with the prompt included every time. Next up: a content calendar that fills itself in.

FAQ

Can an AI agent build a CRM for me?

Yes. In AgentDocs you describe the contacts and fields you want in plain language, and your AI agent (Claude or ChatGPT) creates a real Database you can both read and update. There's no setup screen and no spreadsheet formulas to write.

Do I need to know spreadsheets or databases to do this?

No. You talk to your agent the way you'd brief an assistant ('add Sarah from the design studio, we spoke Tuesday'), and it keeps the table tidy for you.

How much does it cost?

AgentDocs has a free tier with no per-seat fees, and you connect your AI agent over a single link, with no install and no card. A Pro tier exists for heavier use.

Where does the data live?

In a real structured Database inside AgentDocs, built on Stash. Your agent reads and writes it directly, so the same data is there whether you ask Claude or ChatGPT.

Try it free. Connect AgentDocs to Claude or ChatGPT over one link — no install, no card.

Add to Claude →