Draft a Candidate Follow-Up With AI at the Cursor

A recruiter can speak four verified facts (the role, the requirement discussed, the agreed action, and the date) and ask AI to shape them into a structured follow-up draft at the cursor. The recruiter then checks that draft against their notes and sends it personally. This article walks through the workflow with SpeakLingo using a fictional example, and adds a reusable spoken template, a review checklist, and a method for measuring time to reviewed output.

The workflow is designed for compatible focused Windows text fields, such as an email compose window or a notes box. Compatibility is not universal, so try your own tools first. SpeakLingo places text in the focused field, and every draft needs your review before it goes anywhere. An optional Enter-after-delivery behavior can send a message in some apps, so the setup below tells you to turn it off for this task.

A follow-up after a screening call or interview is short, yet it carries several obligations at once:

• It confirms what was discussed, so the candidate can see the conversation was understood.
• It states the agreed next step and who owns it.
• It includes a date or timeframe only where one was actually agreed.
• It keeps role details accurate, since the candidate may rely on them when deciding what to do next.
• It avoids promises that nobody made.

The writing itself is usually simple. The friction comes from the facts being spread across your memory, scattered notes, and a call you finished minutes ago. Turning them into a clean message means structuring them by hand. If you also use a separate AI chat window, you add a copy-and-paste step in each direction.

The cursor workflow targets that friction. You speak the verified inputs and a task instruction, and the AI request produces a draft in the text field where you were already working. Your effort then goes into checking facts and tone. For other tasks built around the same idea, see the use cases and features pages.

Syping is the term SpeakLingo uses for speaking text into the place where you would normally type. You can read the explanation on the Syping page.

For a candidate follow-up, the sequence looks like this:

1. Turn off automatic Enter/send behavior. Before you start, check the delivery settings and switch off any option that presses Enter after text is inserted. That way the draft stays in the field until you decide what happens next.
2. Collect the verified inputs. Check your own notes and confirm four fields: role, discussed requirement, agreed action, and date. If something is uncertain, say so in the request and leave the guess out.
3. Focus the right text field. Click into the email draft or note field where the message belongs, and confirm it is a field SpeakLingo works with in your setup.
4. Speak the request. Give the instruction and the inputs aloud. Online transcription converts the speech to text, and AI contextual formatting shapes the result into a structured message.
5. Let the draft appear at the cursor. AI requests need an internet connection, an enabled service, your consent, and the required access. Audio and text, plus optional screen context if you opt in, may be sent to remote services to produce the result.
6. Review, edit, and send it yourself. Read every line against your notes. Send only when you are satisfied with the message.

The order matters. Verifying facts before you speak gives the AI less room to fill gaps, and reviewing after drafting keeps the decision with you.

One boundary applies to every request. Use the supported AI Task workflow, and say the instruction and the facts the draft needs together in that one request. Dictating alone does not turn AI on. Earlier dictations, text already in the email field, and what is visible on your screen are not automatically available to the AI. If the recruiter wants the AI to use the role, requirement, action, and date, those must be spoken in the request itself. Screen context is a separate opt-in and may be sent to remote services, so leave it off for candidate details you would not want to leave your device.

The scenario below is fictional and uses non-sensitive details. The draft shown is illustrative, not observed output from a test, and shows the kind of result this workflow is meant to produce. A real draft will vary with your wording and context.

Setting: A recruiter at a fictional company, Harbourline Logistics, has just finished a screening call with a fictional candidate named Priya for a Senior Data Analyst role.

Before: raw notes

Here is what the recruiter has in a notes file:

priya sr data analyst call today 12 june. she's done a lot of dashboard work for ops teams, we talked about stakeholder reporting being the main thing for the role. she'll send a portfolio sample, said early next week, I think monday. I need to check with the hiring manager about panel times. didn't get to talk about salary. also mention team is hybrid?? not sure that's confirmed

Writing a clean message from this means deciding what is firm, what is uncertain, and how to order it. It also means avoiding an accidental promise about the hybrid arrangement.

The spoken request

The recruiter focuses the email body and speaks, adapting the standard prompt:

“Using only these verified inputs, draft a candidate follow-up. Inputs: Role is Senior Data Analyst at Harbourline Logistics. Discussed requirement was stakeholder reporting for operations teams. Agreed action: Priya will send a portfolio sample, and I will check panel availability with the hiring manager. Date of call was 12 June. The portfolio sample is expected early next week, possibly Monday, which is approximate and still to be confirmed. Use a clear professional structure. Flag missing facts; do not invent them. Do not mention salary or working arrangements.”

Three choices in that request are worth noting. The unconfirmed hybrid arrangement is excluded, and the AI is told not to mention it. The portfolio timing is marked approximate. Salary, which was never covered, is kept out of the message.

After: the draft at the cursor (illustrative)

Subject: Following up on our call about the Senior Data Analyst role

Hi Priya,

Thank you for speaking with me on 12 June about the Senior Data Analyst position at Harbourline Logistics. I enjoyed hearing about your dashboard work with operations teams.

We discussed stakeholder reporting as a central part of the role, and I would like to keep that in focus as we move forward.

Next steps - You: send a portfolio sample, expected early next week. - Me: check panel availability with the hiring manager and come back to you with options.

If the timing on the portfolio sample needs to shift, just let me know.

Best regards, [Name]

Missing facts flagged: Confirmed date for the portfolio sample; panel dates; interview format.

What the recruiter reviews and changes

The structure arrived formatted, and the flagged gaps are useful. The recruiter still checks the draft against the notes and makes several decisions:

• Timing. Confirms that “early next week” matches what Priya said, or replaces it with a firm date once agreed.
• Added wording. Notices that “I enjoyed hearing about your dashboard work” and “come back to you with options” are phrases the recruiter never spoke. They keep, reword, or delete them.
• Flag block. Removes the “Missing facts flagged” block before sending, since it is addressed to the recruiter only.
• Unconfirmed claims. Checks that hybrid working and salary are absent, as intended.
• Signature. Adds the real signature and sends the message personally.

The draft is a starting point. Responsibility for accuracy and for the candidate relationship stays with the recruiter.

Copy this into your own notes and fill in the brackets aloud. Keep the constraints, since they do much of the work.

“Using only these verified inputs, draft a candidate follow-up. Inputs: Role: [role and company]. Discussed requirement: [what we covered]. Agreed action: [who does what]. Date: [call date and any agreed deadline, marked firm or approximate]. Use a clear professional structure with a short thank-you, a summary of the discussion, and a next-steps list showing who owns each action. Flag missing facts; do not invent them. Do not mention: [topics not confirmed or not to be raised].”

Practical notes on using it:

• Mark certainty. Say “firm” or “approximate” beside each date, which gives the AI a signal for how to phrase it.
• Name exclusions. If compensation, location arrangements, or start dates were not agreed, list them under “do not mention.”
• Keep one task per request. Ask for the follow-up draft alone. A rejection message or a scheduling note deserves its own request.
• Adjust tone with one phrase. Add “warm but concise” or “formal” to the structure line.
• Keep your voice. If the same stock phrases keep appearing in drafts, add a short style instruction such as “plain wording, no enthusiasm phrases I have not dictated.”

The template also works if you never use SpeakLingo. Typed into any drafting tool your organization approves, the same structure and constraints give you a consistent starting point.

The review boundary for this workflow is simple. The AI drafts wording from facts you provide. It is not used to score, rank, or assess candidates, and sensitive data stays out of the request.

Run this before you send anything. It is a writing practice, not a substitute for your organization's own policies or your professional judgment.

Facts - [ ] Role title and company name match the job posting. - [ ] The requirement described matches what was actually discussed. - [ ] Each agreed action names the right owner. - [ ] Every date is correct and is marked firm or approximate as appropriate. - [ ] No salary, location, start date, or benefits statement appears unless it was confirmed.

Commitments - [ ] The message promises nothing you cannot deliver, such as a decision date you do not control. - [ ] Panel, interview format, and timeline statements are accurate or clearly marked as pending.

Content boundaries - [ ] No candidate scoring, ranking, or evaluative language slipped in. - [ ] No sensitive personal data appears in the draft or was spoken into the request. - [ ] Phrases you did not dictate have been checked, kept deliberately, or removed. - [ ] The flagged-missing-facts block is removed before sending, or its items are resolved.

Tone and delivery - [ ] The wording sounds like you and suits the candidate's seniority and the relationship so far. - [ ] The name, spelling, and recipient address are right. - [ ] Automatic Enter/send behavior was off, and you press send yourself.

Online transcription supports 99+ languages, so you can speak in a supported language that suits you and your candidate. That statement covers transcription only. It is not a translation feature, and it does not guarantee AI output in every language. If you write to candidates in more than one language, test a sample draft in each and read the result closely, or ask a fluent colleague to check it.

Connectivity affects the AI step. AI requests need the internet and an enabled service. If your connection drops, AI drafting stays unavailable until you are back online. Transcription has its own mode, setup, and access requirements, so AI being unavailable does not show that transcription is available. Offline English is only a fallback, and only where you have already configured it with a local runtime. In that situation, write the follow-up yourself, using the template above as a skeleton for structure.

Fast text appearing in the field says little on its own. A message is finished only once you have checked it, so measure time to reviewed output: the span from the end of the call to a message you are ready to send.

A simple way to run the comparison for yourself:

1. Pick a baseline set. Over a week, note the start and end time for each follow-up you write your usual way. Start the clock when the call ends and stop it when the message is reviewed and ready.
2. Record the same span for the cursor workflow. Include verifying inputs, speaking the request, reading the draft, and making edits.
3. Log the edits. Note what you changed: dates, tone, missing details, removed lines. Frequent corrections in one area suggest a change to your spoken template.
4. Compare like with like. Match similar message types, since a first-contact follow-up differs from a post-interview one.
5. Note the side effects. Track whether details were missed less often, or whether structuring felt easier, alongside the clock time.

Your own records will show whether the workflow suits your day. This article makes no speed claim, and you should not assume one.