Skip to main content

A follow-up email has one job: move a real conversation one step forward. Most bad ones fail because they forget what was actually said. They open with “Just checking in,” pretend to a friendship that does not exist, or invent a deadline to create pressure. This guide shows how to write a follow up email with AI so the draft is built from the enquiry, call or proposal you already have, and then checked against it before anyone presses send.

The method is the same for every situation: collect the context, give the model a strict brief, and verify the draft line by line. Below you will find a reusable prompt, four complete examples for an illustrative business, and a pre-send checklist. One boundary up front: this is about replying to people who contacted you or are already talking to you. Cold bulk outreach is a different activity with different rules, and nothing here makes it a better idea.

Step 1: Gather what was actually said

The model knows nothing about your customer except what you paste in. Before you open any AI tool, put together a short context pack.

Field What to include
Thread The original enquiry, your last reply, or call notes. Paste the text, not your memory of it.
Their question or need In their words where possible.
Commitments already made Prices quoted, dates promised, what you said you would send.
Allowed facts Availability, policies, attachments. Anything not listed here stays out of the email.
The next step One concrete action for them or for you.
Do not mention Discounts not on offer, internal problems, competitor names.

If you have read our beginner’s guide to prompting for small businesses, this is the same task, audience, facts and constraints structure applied to one specific message. The “commitments already made” row matters most. A follow-up that quotes a different price from the one in your last email does more damage than no follow-up at all.

Step 2: Use one base prompt and add the situation

Keep one base prompt and add a short situation line per email. The base prompt carries the rules that never change; the situation line explains what this particular email needs to do. Clear do and don’t constraints work better than vague tone words, which is the point of our piece on writing AI instructions as dos and don’ts.

You are drafting a follow-up email for [business name].
Write in this voice: [2-3 lines from your voice notes].

Use only the facts in CONTEXT. Do not add prices, dates,
promises, discounts or personal details that are not there.
Do not claim a previous conversation happened unless CONTEXT shows it.
Keep any commitment already made exactly as stated.
End with one specific next step. No urgency that is not real.
Under 150 words. Plain text. Include a subject line.

SITUATION: [one line from the examples below]
CONTEXT:
[paste the context pack]

The voice lines are worth getting right once. If you have no written voice yet, our guide to AI prompts for brand voice and identity will get you a usable first version.

Step 3: Draft for the situation in front of you

The four examples below use an illustrative business: Fernhill Upholstery, a two-person workshop that recovers chairs and builds seating for cafés. The customers, prices and emails are invented for this guide. Each email is the edited version after the checks in Step 4, and the first-draft errors mentioned are typical examples, not logged test results.

The enquiry that went quiet

Inputs: Mara asked about recovering two armchairs. Fernhill replied asking for photos and fabric preferences. Six days, no answer. Nothing was quoted.

Situation line: They have not replied to our questions. Make it easy to answer. Do not quote a price.

Subject: Your armchairs: two quick questions

Hi Mara,

Following up on your enquiry about recovering two armchairs. To give you a price, I need a photo of each chair from the front and side, plus a rough idea of the fabric you have in mind (plain, patterned, or “no idea yet” is a perfectly fine answer).

If it is easier, reply with the photos and we can talk fabric afterwards. If you have decided not to go ahead, no problem at all, a one-line reply saves us both a reminder.

Best,
Jonas, Fernhill Upholstery

Why it fits: it repeats the two things needed to move forward, lowers the effort of replying, and gives a polite exit. The first AI draft said “Hope you’re having a great week!” and offered a “free consultation” the workshop does not run. Both went.

After a discovery call

Inputs: a 30-minute call with Dev, who owns a café, about banquette seating along one wall. Agreed on the call: Fernhill sends two heavy-use fabric options with physical samples by the end of next week, and Dev sends wall measurements. Price not discussed.

Situation line: Recap what was agreed on the call, who does what, and nothing else.

Subject: Banquette seating: notes from our call

Hi Dev,

Thanks for your time today. Here is what we agreed so nothing slips:

From us: two fabric options suitable for heavy café use, with samples you can handle, by the end of next week.
From you: the length of the wall and the height to the window ledge.

Once we have the measurements, we will put together a proposal with a layout sketch and price. If anything in this summary does not match your memory of the call, tell me and I will correct it.

Best,
Jonas

Why it fits: a call recap works as a lightweight record. The draft initially added “and we’ll have it installed before Christmas”, which nobody said. If your calls are transcribed, our guide to turning a meeting transcript into decisions and next actions produces exactly the inputs this email needs.

The proposal with no reply

Inputs: proposal for a two-section banquette in Dev’s preferred fabric, sent nine days ago, quoted at an illustrative £2,400 including fitting, valid for 30 days. No response.

Situation line: Check the proposal arrived and invite questions. Restate the price and validity exactly. No pressure.

Subject: Your banquette proposal

Hi Dev,

Checking that the proposal reached you safely. It covers the two-section banquette in the fabric you preferred, at £2,400 including fitting, and the price holds for 30 days from the date we sent it.

If something in it does not work, whether that is the layout, the fabric or the budget, I would rather hear it now and adjust than guess. A 15-minute call is fine too; just reply with a time that suits you.

Best,
Jonas

Why it fits: it restates the commitment word for word and asks for objections, which are the usual reason proposals go quiet. The model’s first attempt said “prices may rise soon”. Nothing in the context said so, which makes it fake urgency.

Sending information they asked for

Inputs: a customer, Ruth, asked for care instructions and the fabric supplier’s test certificate for her sofa, recovered in a wool blend. Both are attached.

Situation line: Deliver what was requested, say what each attachment is, and stop.

Subject: Care instructions and fabric certificate for your sofa

Hi Ruth,

As requested, attached are two documents: our care instructions for the wool blend on your sofa, and the fabric supplier’s test certificate for that fabric.

The short version of the care sheet: vacuum on low suction and blot spills rather than rubbing. Anything stubborn, send us a photo before trying a cleaner.

Best,
Jonas

Why it fits: the shortest of the four, because the reader asked for documents and nothing else. Resist letting the model add an upsell paragraph about cushion refills. The care summary must be copied from your actual care sheet, never generated.

Step 4: Check the draft against the source

Run every draft through the same review, because first drafts rarely pass it untouched. The common failures are predictable: invented familiarity (“lovely to see you again” to someone you have never met), new commitments, softened prices, and pressure that does not exist. Run this before sending.

Check Compare against Pass when
Names and spelling Their last email Name, company and product match exactly
Prices and figures Quote or proposal Identical, including VAT and validity
Dates and deadlines Your calendar and prior emails Every date was already agreed or is genuinely yours to give
Commitments Call notes or thread Nothing promised that was not agreed
Relationship claims Thread history No “great chatting” unless you chatted
Urgency Real constraints Any deadline is true and stated plainly
Attachments and links The email itself Attached, correct version, links open
Next step The context pack One clear action, easy to take

When a draft fails, fix the context pack or the situation line rather than only editing the email. If the model keeps inventing a free consultation, add “We do not offer free consultations” to the do-not-mention row and that error stops recurring.

Where to run this

Any chat assistant works with the prompt above. If you would rather draft inside your inbox, Gmail’s Help me write can generate or refine a draft from a prompt, including from a reply; Google lists it for eligible Google Workspace or Google AI plans, with personal accounts limited to the US. In Outlook, Copilot can draft a message from a prompt and lets you adjust length and tone before you keep it. Either way, the context pack still does the real work. (Features checked 5 November 2026.)

If you are tempted to scale this into automated messages to people who never contacted you, read the ICO’s guidance on direct marketing by electronic mail first (UK rules; other countries have their own). A good follow-up is a reply in an existing conversation, and that is where AI drafting earns its keep.

Direktorium publishes practical AI workflows like this one, one task at a time. If you want a follow-up process set up for your team, with voice notes, templates and review steps, BBDirector can build it with you.

AI to build a brand in the fitness industry
AI Prompts: ChatGPT for Brand Voice and Identity

AI Prompts: ChatGPT for Brand Voice and Identity

ViktorViktorAugust 26, 2026
workflow automation for small businesses
Workflow Automation for Small Business: 5 Processes to Start With

Workflow Automation for Small Business: 5 Processes to Start With

ViktorViktorOctober 16, 2026
How SMEs Can Use AI to Understand What Their Agritech Logo Should Communicate
How SMEs Can Use AI to Understand What Their Agritech Logo Should CommunicatePrompt Engineering

How SMEs Can Use AI to Understand What Their Agritech Logo Should Communicate

ViktorViktorAugust 26, 2026