When should you use it?
- An owner or office lead spends time hunting through today's board just to see what needs a decision
- You can export or paste the day's jobs, including status, technician, notes and callback flags
- You want a short internal digest, not a message to customers
Not for: Sending the digest to customers or posting it anywhere public; Inventing a tidy outcome when a technician has not written notes; Live, minute-by-minute tracking; this is a briefing from a snapshot or export; Assuming a live software connection; a spreadsheet export is enough
What do you need, and what do you get?
You need
- A day's job or call export with job id, customer label, area, technician, status, times, notes and callback flags
- The period covered and the internal recipients
- Optional: the sections the owner wants, and any total the export claims
You get
- A header with unique job count, incomplete count, callback count and conflict count
- Decisions needed and callbacks at the top
- A conflicts and incomplete-records list, with blanks left blank
- One line per unique job, marked as an internal briefing not sent to customers
How does it work, step by step?
- Agree the sectionsDecide what belongs in the briefing, such as decisions, callbacks, incomplete records, jobs run and open issues.
- Export the dayDownload or copy the period's jobs from your field-service system. A spreadsheet is enough.
- Run the briefingPaste the rows with the prompt and the internal recipient list.
- Check counts and flagsConfirm the unique job count, read conflicts and incomplete rows, and do not treat a blank note as a successful close.
- Keep it internalShare the briefing only with the named internal recipients. Use it to assign follow-up; do not send it to customers.
The prompt
Our own prompt, inspired by Tyler Griffin (@TylerGriffin) and written to work in Grok Bot, Claude, ChatGPT and others. Values in lime are placeholders.
Build an internal daily briefing from a field-service job or call export. Send it only to the named internal recipients. Do not contact customers. Inputs: - JOBS: rows with job_id, customer label, address or area, technician, status, scheduled or completed times, notes, callback flag, and any completeness flags - TODAY, and the SHIFT or PERIOD covered - RECIPIENTS: internal people only - BRIEFING_SECTIONS the owner wants, default: decisions needed, callbacks, incomplete records, jobs run, open issues, notable upsell notes Rules: 1. Deduplicate by job_id. If the same job_id appears more than once, merge into one row. Prefer the later timestamp. If two rows for the same id disagree on status or notes, list that under CONFLICTS and do not invent a single story. 2. Count unique job_ids, and state that count. Do not count duplicates as extra jobs. 3. A record is incomplete if status, technician, or notes are missing, or if status is still open or in progress with no end note after the scheduled window. List these under INCOMPLETE. Do not fill gaps or write that the job went fine when notes are missing. 4. Put decisions needed and callbacks at the top. 5. Only report what the rows support. If a field is blank, say it is blank. 6. If the export looks truncated (for example a technician is mentioned in a note but has no job row, or a stated total does not match unique ids), say the export may be incomplete. Output: - Header: period, unique job count, incomplete count, callback count, conflict count. - Decisions needed - Callbacks - Conflicts and incomplete records - Jobs run (unique, one line each) - Open issues - Marked INTERNAL BRIEFING — NOT SENT TO CUSTOMERS Hard limits: - Do not email or text anyone except by preparing a draft for the listed internal recipients. - Do not invent job outcomes. - Do not include payment card data if it appears in notes; replace it with [redacted].
Inspired by Tyler Griffin (@TylerGriffin): X Article: how an HVAC and plumbing owner actually uses an AI assistant. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Counting duplicate rows as extra jobs. Unique job ids are the count that matters.
- Letting two disagreeing rows for the same job collapse into a single tidy story.
- Trusting an export that claims a total the unique ids do not match, or that names a technician who has no job row.
- Leaving customer addresses in a briefing that is then forwarded too widely.
Our take
A daily digest is only useful if you can trust what it leaves out. Griffin's account points to a simple need: stay current on the day's calls without living in the board. The hard part, which the article does not specify, is messy data. Duplicate rows, conflicting statuses and missing notes are common in field-service exports. Our prompt treats those as first-class output: deduplicate, flag disagreements, leave blanks visible and say when the export itself looks truncated. We also kept the briefing internal, because these snapshots often include customer names and addresses. We did not test this on a live company account or a vendor API. It will work on a reasonably complete export you can paste; it will not repair a board that was never updated. In our rule-engine dry run on synthetic fixtures, duplicates collapsed, conflicting rows were not silently merged, incomplete notes stayed blank, counts matched unique ids, and a request to send the briefing to customers was refused.
Where the time goes back: Reading every job on the board to find the few that need a decision is a poor use of an owner's late afternoon. A briefing that already counts unique jobs and lifts callbacks, conflicts and missing notes to the top shortens that pass.
FAQ
How often should this run?
What if notes are empty?
Can this update the official job record?
Source and credit
Inspired by Tyler Griffin (@TylerGriffin). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- X Article: how an HVAC and plumbing owner actually uses an AI assistant
- Creator
- Tyler Griffin (@TylerGriffin)
- Published on
- X
- Original date
- Nov 14, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Grok Bot
- AI dry run
- Nov 16, 2026: passed. Dry-run tested by our AI test agent on sample input. Dedupe by id, conflicts kept, gaps visible, internal only.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Nov 16, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.