A follow-up can sound polished and still be misleading. “I’ll send the revised file tomorrow” is a plan, not evidence that the file has been sent; “we expect approval this week” is an estimate, not an approval. When you use AI to draft a reply, the important job is to keep each statement tied to what the email thread actually establishes. Start with the original messages, identify what has and has not happened, then check the draft before sending. The aim is a useful update that moves the conversation forward without quietly upgrading uncertainty into completion.
The useful bits
- Separate completed actions from plans, estimates, and open tasks before prompting AI.
- Ask for a draft based only on the thread, and mark unknowns for your review.
- Check every status claim against the original email before sending.
Make a small status ledger first
Before asking for a draft, scan the relevant messages and write down the status of each item. Use plain labels such as completed, planned, estimated, waiting, or unknown. Include the message or date that supports a completed claim, if that will help you verify it later.
A promise to do something is not proof it happened. A proposed date is not a confirmed deadline, and silence from another person is not approval. If a thread does not say whether an action was completed, record it as unknown rather than filling the gap with a likely-sounding answer.
Give the assistant boundaries, not just a goal
If AI assistance is available for your account, provide the relevant thread and specify what the draft may claim. Ask it to use only facts stated in the messages, preserve uncertainty, and leave placeholders for missing details. You can also name the next step you want the recipient to take, such as confirming a date or sharing a document.
Avoid prompts like “tell them it’s all done” unless the thread establishes that. A more careful instruction is: “Draft a concise follow-up using only the status below. Do not describe planned or pending work as complete. Mark any unsupported detail in brackets for me to check.” AI can still misread a thread or introduce a mistake, so a careful prompt is a starting point, not verification.
Spot status words that change the meaning
Read the draft for verbs and time cues. “Sent,” “approved,” “fixed,” and “completed” state that something happened. “Will send,” “expect,” “aim to,” and “waiting for” describe a future action, estimate, or dependency. A draft can become inaccurate through a single verb, even if the rest of its summary is correct.
Also check who owns the next action. “We are waiting for your confirmation” is different from “we will confirm.” If responsibility or timing is unclear in the thread, keep it open in the draft rather than assigning it to someone by assumption.
A practical thread-to-draft example
Suppose the thread says that you plan to revise a proposal, expect to send it Thursday, and are waiting for a colleague to confirm one price. Unless another message confirms those events, the revision and price are still pending. A safe draft can report the plan and dependency without implying delivery or approval.
Prompt: Draft a brief follow-up from this thread. The proposal revision is planned, not complete. Thursday is an estimate, not a confirmed delivery date. One price is awaiting confirmation. Do not say the proposal has been sent or approved; use brackets for anything I need to verify. Draft to review: Hi [Name], I’m working on the proposal revision and am aiming to send it Thursday. I’m still waiting to confirm one price, so I’ll update that detail before sending the revised version. Please let me know if your timing has changed. Best, [Your name]
Verify against the original, not the summary
Compare each factual sentence in the draft with the actual messages. Confirm names, dates, recipients, amounts, attachments, and promises against the originals. In particular, look for details added by the draft that sound plausible but were never stated. If an attachment is mentioned, check that you have the right file and that it is actually attached before sending.
If a fact cannot be verified, remove it, qualify it, or leave a bracketed question for yourself. Do not send bracketed placeholders or wording that still needs confirmation. Review the recipient field as well as the message body; a correct status sent to the wrong person is still a problem.
Confirm the message says what you intended
Before sending, read the draft once as the recipient: would they understand what is finished, what remains open, and what happens next? Then check the sent message in your mail so you can confirm the version you intended was sent to the intended recipient. If the system does not show a sent copy or the send appears to have failed, do not assume the follow-up went out; check its status and resolve that before describing it as sent.
BITTU Mail offers plan-dependent AI assistance, and its results can be wrong. Review the original email and the final draft yourself. That check is what distinguishes a useful update from an AI-generated claim that turns a plan, estimate, or pending task into a completed result.
About this guide
Prepared with AI assistance for the BITTU Mail editorial library. Check important details against your original messages and current account settings. Product access depends on your role, plan and provisioning state. Provider interfaces and requirements can change.
View current plans · Security information · Suggest a correction