A sales AI agent template is a reusable set of instructions that tells an agent when to run, what to read, what to look for, and how to report back. The five below cover post-call follow-up, MEDDIC gap checks, deal risk, pipeline hygiene, and forecast call prep. Each one is ready to paste, with its trigger, inputs, and what good output looks like.
If your agents keep producing summaries nobody reads, you are not alone. Gartner predicts that by 2028 AI agents will outnumber human sellers tenfold, yet fewer than 40% of sellers will say agents improved their productivity. The gap is rarely the model. It is the instructions.
What makes a sales agent template work?
A template works when it removes every decision the agent would otherwise guess at. In practice that means four parts: scope, criteria, evidence, and output.
Scope is which calls, deals, or owners the agent covers and over what window (“open deals closing this quarter,” “the last 7 days”). Criteria define what counts (“no buyer activity in 14 days,” “no economic buyer named”). Evidence is what each finding must cite, such as a quote, a date, or a CRM field, so a manager can trust it without reopening the call. Output is a fixed shape with a cap (“top 5, as a table”), so this week’s run reads like last week’s.
Salesforce’s 2026 State of Sales report found that 54% of sellers have already used agents, and that 51% of sales leaders with AI say disconnected systems are slowing their initiatives. Both point the same way. Agents are easy to start and hard to make useful, and the difference is whether the agent can see clean call and CRM data and has been told exactly what to do with it.
Every template below follows the same layout: what it does, when it runs, the instructions, the inputs it needs, and what good output looks like.
Template 1: How do you automate post-call follow-up emails?
What it does. Drafts a follow-up email for the rep within minutes of the call ending, built from what the buyer actually said rather than a generic recap.
When it runs. After each call, limited to external calls tagged Discovery, Demo, or Negotiation.
You are drafting a follow-up email for the rep who ran {{call_title}}.
Read the transcript for call {{call_id}}. Then write an email that:
1. Opens with one sentence on the buyer's main problem, in their words.
2. Lists the agreed next steps, each with an owner and a date.
3. Answers any open question the buyer asked that the rep promised
to follow up on. If the answer is not in the call, write
[REP TO CONFIRM] instead of guessing.
4. Ends with one clear ask that matches the agreed next step.
Keep it under 150 words, professional and plain. No bullet points
in the opening line. Do not invent pricing, dates, or commitments
that were not said on the call.Inputs. The call transcript and attendee list. CRM deal stage is useful for tone but not required.
Good output. An email the rep would send with one or two small edits. Next steps have names and dates, and anything uncertain is flagged rather than made up.
Template 2: How do you find MEDDIC gaps after a discovery call?
What it does. Checks each qualifying call against MEDDIC and tells the rep which elements are still missing, with a suggested question for the next conversation.
When it runs. After each call tagged Discovery or Demo.
Review call {{call_id}} against MEDDIC: Metrics, Economic Buyer,
Decision Criteria, Decision Process, Identify Pain, and Champion.
For each element, return one of: Confirmed, Partial, or Missing.
- Confirmed or Partial: quote the line from the call that supports it,
with the timestamp.
- Missing: write one open question the rep can ask on the next call
to fill the gap.
Then list the two gaps that put this deal most at risk, and why,
in one sentence each. Return the result as a table followed by
the two risks. Do not mark anything Confirmed without a quote.Inputs. The call transcript, plus the deal’s existing qualification fields so the agent does not re-ask what an earlier call already answered.
Good output. A six-row table where every Confirmed has a quote and every Missing has a usable question. If you want the fields written back to the CRM, see the MEDDIC Analyzer agent.
Template 3: How do you run a deal risk review with an AI agent?
What it does. Scans open late-stage deals daily and flags the ones showing the quiet signals that come before a slip: a buyer who has gone silent, a single thread, a close date that keeps moving, or no confirmed decision maker.
When it runs. Weekdays at 8am, delivered to each deal owner with individual output, plus a rollup to the manager.
Review every open deal I can access that is in Proposal stage or
later, with a close date in the current quarter.
Flag a deal only if at least one of these is true:
- No call or email with the buyer in the last 14 days.
- Only one buyer contact has joined a call in the last 60 days.
- The close date moved more than once in the last 30 days.
- No economic buyer has been identified on any call.
For each flagged deal, return: deal name, owner, close date, the
signal(s) triggered, the evidence (a date, a quote, or a CRM field),
and one recommended next action. Sort by close date. Return at most
10 deals. If nothing is flagged, say so in one line.Inputs. Deal stage, close date and its change history, call and email activity, and attendee lists.
Good output. A short list a rep can act on before their first meeting, where each flag has a date or quote behind it. Airspeed ships a ready-made version of this as the Deal Risk Agent.
Template 4: How do you keep pipeline data clean automatically?
What it does. Sweeps the pipeline for the records that quietly break forecasting: missing next steps, close dates in the past, stale stages, and amounts that no call supports.
When it runs. Weekdays, early morning, with individual output so each rep only sees their own records.
Check every open deal owned by the recipient for these hygiene issues:
1. Close date is in the past.
2. Next step field is empty or older than 14 days.
3. Deal has been in the same stage for more than 30 days.
4. Amount is empty, or no call has discussed budget or pricing.
5. Stage is later than Discovery but no discovery call is on record.
Group the results by issue. For each deal, give the deal name, the
current value, and a suggested correct value based on the most recent
call, with the call date as evidence. Cap the list at 15 items, most
urgent first. Finish with a one-line count of clean deals.Inputs. CRM fields for each open deal, and recent call transcripts so the agent can suggest values rather than only point out blanks.
Good output. A to-do list a rep can clear in ten minutes, with suggested values instead of bare warnings. Salesforce’s same report found 79% of high performers prioritize data hygiene, against 54% of underperformers, which is why this agent tends to pay for the others.
Template 5: How do you prepare for a weekly forecast call with an agent?
What it does. Gives a sales manager a one-page brief before the forecast call: what changed since last week, which commit deals lack evidence, and which questions to ask each rep.
When it runs. Weekly, the morning of the forecast meeting, with previous-run context turned on so it can compare week over week.
Prepare my forecast call brief for my team's open deals closing this
quarter.
1. Summary: total open pipeline by forecast category, and what changed
since your previous run (new, slipped, pulled in, closed).
2. Commit check: for each Commit deal, state whether the last 30 days
of calls show a confirmed economic buyer, an agreed decision
process, and a mutual close plan. Quote the evidence or mark it
missing.
3. Risks: the five deals most likely to slip, with one sentence of
evidence each.
4. Questions: one specific question to ask each rep, tied to a deal.
Format: a three-sentence executive summary, then a table for each
section. Keep the whole brief to one screen.Inputs. Forecast categories, deal history, the last 30 days of calls, and the previous run’s output.
Good output. A manager walks in knowing which three deals to spend the meeting on and what to ask. For the coaching side of this, the agents worth building by role covers rep snapshots and pipeline review prep.
How do you roll these templates out without creating noise?
Roll out one template at a time, starting with the one whose output is easiest to judge. The post-call follow-up is usually first, because every rep can tell within a day whether the draft is usable.
Send early runs to an in-app history rather than Slack or email, read ten outputs, and tighten the instructions where the agent guessed. The most common fixes are a smaller cap, a stricter definition (“silent” means no buyer reply, not no rep email), and a requirement to cite evidence. Narrow triggers early too. An after-call agent that runs on internal syncs will train your team to ignore it.
Then check the data underneath. An agent can only flag a missing economic buyer if calls are recorded and attendees are captured, and it can only spot a moved close date if the CRM keeps history. If you are still deciding how much to build yourself, the guide on how to build AI agents for your sales team covers build versus buy and the data foundation agents need.
Where Airspeed fits
Every template above runs as written in Airspeed Agents. Airspeed already captures your calls, transcripts, and CRM fields, so the agent starts with the inputs each template needs instead of a blank context window. You pick the trigger (a schedule, a finished call, a deal stage change, or a web trigger), paste the instructions, and choose whether output goes to Slack, email, or in-app history, either as one shared digest or scoped to each rep’s own deals. Turn on previous-run context and the weekly agents report what changed, not just what is. Airspeed does the work on every call and every deal, so your reps get the follow-up and your managers get the brief without anyone running a prompt by hand.
If you want to see these templates running on your own calls and pipeline, book a demo.
Frequently asked questions
What is a sales AI agent template?
A sales AI agent template is a reusable set of instructions that tells an agent when to run, what data to read, what to look for, and how to format the result. A good one names a trigger (a schedule, a finished call, or a deal stage change), a time window, the evidence it must cite, and a fixed output format, so every run is comparable to the last.
What should every sales agent prompt include?
Four things. The scope, meaning which calls, deals, or owners it covers and over what time window. The criteria, meaning exactly what counts as a risk, a gap, or a change. The evidence, meaning quotes, dates, or CRM fields that back each finding. And the output, meaning a capped list or table in the same shape every time.
Which sales agent should a team build first?
Start with the post-call follow-up agent. It runs on every call, the output is easy to judge (would the rep send this email?), and reps feel the time saving on day one. Once trust is there, add a weekly manager digest or a deal risk review, which depend on the same clean call and CRM data.
Should a sales agent run on a schedule or after each call?
Match the trigger to when the action is due. Follow-up emails and qualification gap checks belong right after the call, while the conversation is fresh. Pipeline hygiene, deal risk, and forecast prep work better on a daily or weekly schedule, because they look across many deals and the value is in the pattern, not the single conversation.
How do you stop a sales agent from producing noisy output?
Cap the number of items, require evidence for every claim, and narrow the trigger. An after-call agent that only runs on calls tagged Discovery or Demo sends far fewer messages than one that runs on every internal sync. During setup, send output to an in-app history first and only route it to Slack or email once the format is stable.
