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.
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.
| Part | What goes in | What it prevents |
|---|---|---|
| Instructions | Client in one page: business, audience, voice with examples, rules, who approves. | Re-briefing every model every time; off-brand drafts. |
| Files | Brand 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. |
| Scope | The proposal, the agreed deliverables, the dates, the price. | Unpaid work that crept in because nobody could point to the scope. |
| Deliverables | Each piece of work as a file, named by date and version. | "Which version did we send?" |
| Feedback log | Every comment the client made, next to the version it was about. | Repeating a mistake the client already flagged. |
| Decisions | What was agreed in calls: tone changes, priorities, things they rejected. | Relitigating settled questions. |
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.
"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.
"Read this as [client]. What would they still object to?" Catches the revision that fixed the words but not the tone.
Clients approve faster when they can see what changed. The list comes from step two.
Accepted feedback becomes a rule in the instructions, so every future draft starts from it.
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.
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.
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.
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.
Add the client's files, three voice examples and the proposal. Then start here.
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.
Yes, on team plans, which include shared Projects, chats, prompts and files with seat management. See team projects.
No, unless you give them a seat. You deliver files and reports; the working context is yours.
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.
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.
One Project, file uploads and the major model families. More Projects, larger uploads and team sharing are on paid plans. See plans.
Instructions, files, scope. Ten minutes now saves the March-file hunt forever. Free to start.
Also: the agency workspace · start an AI agency · analyse files with AI · team projects
Continue learning
Run the same prompt through ChatGPT, Claude, Gemini and Grok before trusting one answer.
Core featureLet several models draft, challenge and verify each other instead of trusting one answer.
ProjectsKeep files, instructions, code and chats together so every model works from the same context.
FeaturesModels, Projects, collaboration, verification, Studios and images in one workspace.
AgencyOne Project per client; every deliverable from the same context.
AgencyNiche, offer, delivery system, first clients, then the product.
FilesContracts, spreadsheets and notes questioned by several models.
TeamsShared context, every model, one invoice.
MarketingTwo channels, a weekly routine, one number.