Skip to content
The GrowthCred field guide

Use AI for proposals and follow-ups without losing control.

Start by using AI to draft from approved information. Keep prices, commitments and sending under human review. Measure the whole process, including corrections, before deciding whether automation helps.

By Phila Ngwenya · Reviewed

Gather the inputs

You need the customer's request, the agreed scope, approved pricing and your usual terms. Missing information should remain a question, not turn into a confident assumption. Use redacted examples while testing.

Separate client-specific notes from your reusable business brief. A previous customer's confidential details do not belong in a template used for everyone.

Try this briefing structure

Ask: Draft a proposal using only the supplied scope, approved pricing and meeting notes. Use our house style. Include deliverables, exclusions, timeline and next step. Mark missing facts as questions. Do not invent dates, discounts or payment terms.

For a follow-up, ask for a short message summarising the agreed next step. Request a list of uncertain facts separately so they cannot accidentally appear as commitments in the email.

A fictional worked example

A fictional consultant has a meeting note saying the client wants a reporting template. The note contains no agreed deadline or price. A useful draft describes the proposed deliverable and asks to confirm timing and cost. An unsafe draft invents a Friday deadline and a discount.

Both drafts may sound polished. The difference is whether the output respects what is actually known. This example illustrates a review rule; it is not a customer result.

Review before sending

Check the recipient, scope, figures, dates, exclusions, attachments and tone. Recalculate totals outside the prose draft. Make one person responsible for approval and keep the final version where the team can find it.

  • Does every commercial claim have an approved source?
  • Is anything from another customer included?
  • Does the next step match the actual conversation?
  • Are draft notes and unanswered questions resolved?

Measure the complete task

Record how long your usual process takes. Then measure preparation, AI drafting, correction and final review on comparable tasks. Use several examples, including messy ones. Count errors as well as minutes.

Only automate the repeatable parts once the workflow is dependable. Automatic sending is a separate decision that needs controls for exceptions, duplicate messages and changes in customer context.