Cookbook

Each recipe below is a working shape: what triggers it, what to tell the builder, where to deliver, and what goes wrong. Adapt the scope and the destination to your team.

1. Auto CRM Update

Trigger — event: a call ends. Delivery — CRM write-back, plus a Slack confirmation to the deal owner.

After every external sales call, update the deal in the CRM:
- write the MEDDPICC fields you can support from the transcript
- log the next steps that were actually agreed, with owners
- update the close date if the call changed it, and record the reason from the call
- flag anything that contradicts what the rep logged last week

Don't invent fields you can't support from the transcript. If the call was internal
or non-sales, skip it.

Failure modes. Runs on internal syncs (tighten the exclusion). Writes optimistic fields the transcript doesn't support (say so explicitly, as above). Fails to write because the CRM connection is read-only (delivery).

Watch it for the first week (validation and review). CRM fields are what everything else reads.

2. Meeting Prep

Trigger — schedule: weekdays, early, in the rep's time zone. Delivery — Slack, the channel the sales team actually uses.

Each weekday at 7am, find today's external calls on the sales team's calendars.
For each one, deliver a brief:
- who's attending, their role, and what we know about them
- our history with the account: last three calls, open threads, unanswered questions
- any competitors mentioned in past calls, and the current objection handling
- recent company news worth referencing
- three discovery questions specific to this deal's stage

Keep each brief under 200 words. Skip internal meetings and interviews.

Failure modes. Too long — nobody reads a 900-word brief at 7am. Fires before the calendar is settled for the day. Includes internal meetings.

3. Competitive watchlist

Trigger — schedule: weekly, or daily for a fast-moving market. Delivery — Slack, an intel channel; battlecard updates reviewed by a human before they ship.

Every Monday at 8am, check each competitor on our watchlist for changes since the
last run:
- pricing and packaging pages
- homepage and product messaging
- news and funding
- LinkedIn posts and job postings

Report only what changed since the previous run, with a link and a one-line
"why this matters for us". Call out hiring clusters separately — a group of GTM
roles or a new location is a signal about where they're investing.

If a pricing or packaging change affects an objection we handle, propose the
rewritten battlecard section.

Failure modes. Re-listing everything instead of diffing (say "since the previous run" explicitly). Reporting noise — a blog post is not a signal. Rewriting battlecards unattended: gate that part.

4. Deal intelligence sweep

Trigger — schedule: weekly, before the forecast meeting. Delivery — Slack to the sales lead; per-deal notes to each owner.

Every Tuesday at 6am, review open deals over $25k in the current quarter and flag:
- single-threaded deals — only one stakeholder engaged across all calls
- deals with no next step agreed on the most recent call
- deals where the close date has moved twice or more
- deals where the buyer raised an objection that was never resolved on a later call

For each flag, include the evidence — which call, which moment — and one specific
play the owner should run this week.

Failure modes. Flagging everything, which is the same as flagging nothing (raise the threshold). Generic plays ("multi-thread the deal") instead of specific ones ("ask for the CFO by name; they were mentioned on the 12 June call").

5. Win/loss follow-through

Trigger — event: a deal is marked lost. Delivery — interview program, then Slack to the rep and to PMM.

When a deal over $25k is marked lost:
- queue a buyer interview with the primary contact
- queue an interview with the rep who lost it
- when both are complete, put the two accounts side by side and report where they
  diverge
- if the loss was competitive, propose the battlecard update and a roleplay drill
  built from the objections the buyer actually raised

The gap between what your rep believes lost the deal and what the buyer says lost it is the useful output. See interview types and from interview to action.

Failure modes. Reaching out before the deal is genuinely dead. Interviewing the wrong contact — the champion who liked you rather than the person who decided. Sending a coaching note to the rep before the interview is read by a human (validation and review).

Patterns across all five

• Name the exclusions. Every one of these recipes spends a line on what to skip. That's what makes the output readable.
• Ask for evidence. "Include which call and which moment" turns an assertion into something a manager can check.
• One destination, one audience. If two audiences need different detail, that's two agents.
• Test before you save, and read the first three real runs.