How to stop re-explaining your company to ChatGPT and Claude

Every new chat starts from zero. So your team types the same story in again. Who the client is. What tone you use. What you decided last week.

Skip it and the answer comes back generic. Do it and the real question waits behind a wall of pasting.

Here is why it happens, and a simple way to stop it across the team.

Every new chat starts from zero

A project manager described it well on Reddit:

"Every time I open a new chat I have to explain the project again. Why we're doing it, who's involved, what happened last week, why we made a certain decision. And if I don't, the output is just... generic."

Agency owners describe the same routine. One founder wrote that their team would "open ChatGPT, copy-paste their brand voice, paste the email they're replying to, paste the SOP for how they handle this situation, then ask it to help."

Now picture that for every person on your team, every day.

Why the AI keeps forgetting

The AI has nothing to start from, for four reasons.

First, it does not keep what you told it. MIT's NANDA group studied how businesses use AI. Its finding: "Most GenAI systems do not retain feedback, adapt to context, or improve over time." (MIT NANDA, 2025) The sample was small and the report is marked preliminary. But anyone who uses AI daily will recognise it.

Second, built-in memory is personal. ChatGPT and Claude can remember things for one person. That helps that person. It does not help the colleague sitting next to them. Your new hire's AI knows nothing your senior person taught theirs.

Third, the context is spread out. Some of it is in email. Some in your project tool. A lot of it is in people's heads. As one PM put it: "that context isn't even in one place."

Fourth, facts change. A client changes their invoice date. Prices go up. If the old fact sits in ten people's notes, nine of them are now wrong.

Atlassian surveyed 12,000 knowledge workers. It found that "leaders and teams waste 25% of their time just searching for answers." (Atlassian, 2025) Atlassian sells team software, so treat it as one view. Still, a lot of that searching ends in a paste into an AI chat.

Write the context down once

You can do this today with no new tool. Write the background once, keep one copy, and make it easy to reach.

1. Write a one-page company brief. What you sell. Who you sell to. Your tone of voice. Words you never use. Keep it to one page.
2. Write a short note per main client. Who they are. Who you talk to there. What they care about. Any fixed rules, like "invoices go out on the 1st."
3. Write down your house rules. The few rules the AI must never break. For example: "refunds over $200 go to a person."
4. Log decisions as you make them. One line each, with a date. "We stopped offering free shipping under $50." This is the part people forget.
5. Keep one copy, with one owner. Put it in a shared folder. One person keeps it current. Old facts get deleted, not left next to new ones.

Then show the team how to use it. Paste the company brief at the start of a chat. Paste the client note when you work on that client.

Make it easy to use every day

Writing it down is half the job. The other half is making sure people use it.

• Keep each note short. A one-page brief gets pasted. A ten-page one gets skipped.
• Use project features where you have them. ChatGPT and Claude both let one person save background in a project. Put the company brief there.
• Review it once a month. Remove what changed. Add last month's decisions.
• Ask new people to read it on day one.

This gets you a long way. But it still depends on people. Someone has to paste the note in. Someone has to update ten personal projects when a fact changes. And each AI app keeps its own copy.

Where Homla fits

Homla gives your team one memory that every AI on the team reads. It removes the pasting and the copies.

• Ask your AI to remember a decision, a client detail or a fix. It is saved for the whole team.
• Ask about a client and your AI looks in the team's memory first.
• Save the new fact and it replaces the old one. The change is logged, so you can see where a fact came from.
• Write a rule like "refunds over $200 go to a person" once. Every AI on the team gets it before it acts.
• Freelancers only read memory if you let them.

It works in every AI app your team connects: Claude, ChatGPT, Cursor and more. Memory only holds what your team asks it to save. It does not read your calls or meetings. You can read more on the Memory page.

How the team does a job, like a refund reply, fits better as a shared playbook. We covered that in why your team's AI work stays in one person's chat.

Start with one page

Write the one-page company brief this week. Share it with the team. Ask everyone to start their next AI chat with it.

Compare the answers with the ones you got before. Then add your top three clients, one note each.