Client projects with AI

Client projects with AI: keep every client's context, files and feedback where the work happens.

Freelancers and small agencies lose hours to the same thing: re-explaining a client to an AI, hunting for the file the client sent in March, and redoing a revision because the feedback from last time was in an email thread. A Project per client fixes all three. This page is the practical setup: what goes in, how revisions work, how to handle feedback, and how to hand a client to a colleague without a handover document.

Free to start with one client. More Projects when there are more clients.

The structure

What a client Project contains, and why each part earns its place

PartWhat goes inWhat it prevents
InstructionsClient in one page: business, audience, voice with examples, rules, who approves.Re-briefing every model every time; off-brand drafts.
FilesBrand guide, past work, product info, data, the files they email you. Upload, or import from Drive where enabled.Hunting through email for the March file.
ScopeThe proposal, the agreed deliverables, the dates, the price.Unpaid work that crept in because nobody could point to the scope.
DeliverablesEach piece of work as a file, named by date and version."Which version did we send?"
Feedback logEvery comment the client made, next to the version it was about.Repeating a mistake the client already flagged.
DecisionsWhat was agreed in calls: tone changes, priorities, things they rejected.Relitigating settled questions.
Revisions

A workflow where the third revision is better than the first, not just different

  1. Paste the feedback, verbatim

    Client comments go into the Project as they were written, with the date and the version they refer to. Paraphrasing loses the thing they actually meant.

  2. Ask for the change and the conflicts

    "Apply this feedback. List what you changed. Flag anything that conflicts with the brand rules or with earlier feedback." The flags are where you earn your fee.

  3. Have a second model read the result as the client

    "Read this as [client]. What would they still object to?" Catches the revision that fixed the words but not the tone.

  4. Send with a change list

    Clients approve faster when they can see what changed. The list comes from step two.

  5. Log what they accepted

    Accepted feedback becomes a rule in the instructions, so every future draft starts from it.

Scope and handover

Two things that go wrong without a Project

Scope

Answer "is this included?" in ten seconds

The proposal is in the Project. When the client asks for something new, ask the AI whether it is in scope and, if not, for a polite note and a price. Scope creep dies when it is visible.

Handover

A colleague is useful on day one

They read the instructions, the feedback log and the decisions. They ask the Project what the client cares about. No handover deck, no lost week. On team plans the Project is shared directly.

Return

A client comes back after a year

You still know their voice, their rules and what they rejected last time. That is the difference between a returning client and a new one.

Free is the start. The latest models are the difference.

Revisions are where clients judge you. Use the models that get them right.

Applying feedback while respecting the brand rules, earlier feedback and the client's tone is a reasoning task, and the latest models do it with far fewer misses than older ones. Fewer misses means fewer rounds, and fewer rounds is the whole economics of client work. With more than one client, the paid plan's extra Projects are simply how the work is organised.

Newer models hold your whole Project in view instead of the last few messages, reason through trade-offs instead of picking the first plausible answer, and are wrong less often and with less confidence. For the questions on this page, that is the difference between advice that sounds right and advice you can act on.

Pro from $20 a month, cancel any time. The free plan stays free.

Free plan
  • Your first Project and file uploads
  • The major AI families to try the workflow
  • A small daily message allowance
Paid plans
  • The latest ChatGPT, Claude, Gemini, Grok and Perplexity models, all on the same Project
  • Far more messages a day, so a working session does not stop halfway
  • Modes where several models draft, challenge and verify each other's answers
  • More Projects: one per client
  • Larger uploads for brand guides, decks and data files
  • Shared Projects on team plans for handover without documents
Start here

First messages for a client Project

Add the client's files, three voice examples and the proposal. Then start here.

BriefFrom these files, write the client instructions: business, audience, voice with rules and examples, what they never say, who approves.
DraftDraft [deliverable] for this client from the instructions and files. Then list the three places you were unsure of their preference.
ReviseApply this feedback verbatim. List what changed. Flag conflicts with brand rules or earlier feedback in this Project.
Client readRead this as the client. What would they still object to? Be specific.
ScopeThe client asked for [request]. Is it in the agreed scope? If not, draft the note and a fair price.
HandoverWrite the one-page handover for a colleague taking this client: who they are, what they care about, what to avoid, what is in progress.
Honest answers

What people ask

Is client data safe in a Project?

Files and conversations stay in your account, are encrypted, and are never used to train any model. One Project per client keeps everything separate. See security.

Can I share a client Project with a colleague?

Yes, on team plans, which include shared Projects, chats, prompts and files with seat management. See team projects.

Can the client see the Project?

No, unless you give them a seat. You deliver files and reports; the working context is yours.

How do I handle a client with a very specific voice?

Three to five examples in the instructions, plus explicit rules ("never uses exclamation marks", "always British spelling"). Then a second model checks every draft against them.

What if the client sends feedback by phone?

Write it down immediately, in their words as far as you can, and paste it into the feedback log with the date. Unlogged feedback is the one you repeat.

What is free?

One Project, file uploads and the major model families. More Projects, larger uploads and team sharing are on paid plans. See plans.

Set up the next client as a Project before the first draft.

Instructions, files, scope. Ten minutes now saves the March-file hunt forever. Free to start.

Create a client project

Continue learning

Open MultipleChat