The situation
Meet Marcus, an illustrative example: he runs a one-person management consultancy, good at the actual work, but spending hours a week writing proposals, follow-up emails and summary documents that follow a similar pattern every time. He's started using ChatGPT for a first draft, but isn't always sure if that's even allowed with client information, and there's no clear policy to check against.
Meanwhile, proposals that go out sometimes get no reply, and he's not quite sure if that's because of the pitch or because nobody followed up a week later.
"I spend my evenings writing proposals and following up on leads from LinkedIn," says Marcus. "I want help drafting proposals and a simple way to track who I've spoken to."
He knows AI could help with some of this. What's missing isn't enthusiasm, it's a clear, safe way to actually use it, and a proper look at where it would help most rather than guessing.
What we'd set up
- An AI opportunities review, looking at where Marcus's time actually goes and finding the two or three places AI would genuinely help, rather than adding it everywhere for the sake of it.
- An AI email and document assistant, drafting first versions of proposals, follow-ups and summaries inside the tools Marcus already uses, ready for a human to check and send.
- A safe AI usage policy, setting out plainly which tools are approved, what client information should never be typed into them, and simple rules he (and any associate he brings in) can follow.
- An enquiry pipeline, so a proposal that's gone quiet gets a reminder to follow up, rather than depending on someone remembering.
How the project would run
Week 1: a free call and an AI opportunities review, looking at real proposals, emails and documents from the last few months to find where drafting actually eats the most time.
Weeks 2-3: the AI email and document assistant is built and tested against real examples, drafting proposals and follow-ups in the firm's own tone rather than generic AI phrasing.
Week 4: the safe AI usage policy is drafted with Marcus, covering what's fine to draft with AI and what client information should stay out of it entirely.
Week 5: the enquiry pipeline is set up so unanswered proposals get a follow-up reminder, followed by a short training session and a written guide to the whole setup.
What changes day to day
A proposal that used to take an hour to draft from scratch starts as a first version in a few minutes, built from the brief and the firm's own past proposals, ready for a consultant to check, adjust and send. A routine follow-up email gets drafted automatically rather than written fresh every time. And when a client's gone quiet after a proposal, a reminder appears instead of the enquiry just fading out. The thinking and advice stay entirely human. What changes is how much of the writing around it needs doing from a blank page.
What you get back
Drafting a proposal from scratch stops taking up an evening, because a first version is ready in minutes for Marcus to check, adjust and send. Writing a routine follow-up email stops being a task in its own right, because it drafts itself. And chasing a client who's gone quiet after a proposal stops relying on memory, because a reminder appears when it's time to follow up.
If this sounds familiar, start my project like this.
Tools involved
A project like this typically works inside Microsoft 365 or Google Workspace, using Copilot or a similar assistant alongside Marcus's own templates and past documents, so drafts sound like him rather than a generic chatbot. The enquiry pipeline usually sits in a lightweight CRM such as Pipedrive or HubSpot, connected to however proposals are currently sent and tracked.
Related
If this sounds familiar, see We know we should use AI but don't know where to start and Admin eats our evenings and weekends. Further reading: Practical AI for small businesses: where it actually helps and Using ChatGPT and Copilot safely: a UK GDPR guide. For the wider picture on this sector, see professional services.