Why generic prompts fail agencies
A prompt like "write a LinkedIn post about our new service" gives polite, generic copy for every account. The model is not the problem. It knows nothing about the client, so it fills the gaps with the safest wording it has.
When you manage many brands, that sameness shows up fast. Clients notice when their feed reads like everyone else's, and your team ends up rewriting most of each draft by hand.
Give the model the client context once
Good drafts start from facts about the client, not from a clever prompt. Collect the same short set of inputs for every brand and keep them in one place.
- The client website, so the offer and wording come from the source.
- Who the audience is and what problem they are trying to solve.
- The tone, in a few plain words, with examples of posts the client liked.
- Competitors and anything the client never wants to say.
- Brand guides, product sheets and other PDFs the client already has.
Keep each client separate
Mixing clients is the quiet risk of using AI at an agency. If one chat or one document holds several clients, tone and facts leak between them. Give each brand its own workspace, its own context and its own history, so a draft for one client can only draw on that client's material.
Review before anything goes out
Treat every AI draft as a starting point. A short review checklist keeps the quality steady across the team.
- Are the facts, names and offer correct for this client?
- Does it sound like the client or like a template?
- Is there any claim, number or promise the client has not given you?
- Does it follow the platform format you are posting to?
Where Impret fits
Impret gives each client brand its own workspace with its own tone, audience, positioning, competitors and content history. You add the website and any PDFs once, then generate three draft variations to compare and refine. You can start free with one brand.
You can try the same idea on one page right now with the free demo: paste a client page and get three short drafts, no account needed.
Frequently asked questions
Can AI write in a client's voice?
It gets close when it has the client's own website, tone notes and examples to work from. It will not get it right from a one-line prompt. Plan on a quick human edit either way.
How much context does each client need?
Less than you might think. The website, a short description of the audience and tone, and one brand guide cover most cases. Add more only when drafts keep missing the mark.
Should the model see one client's material when drafting for another?
No. Keep each brand's context separate so a draft only uses that client's own material.