When should you use it?
- You suspect unused or duplicate subscriptions but have not listed them in one place
- You want a monthly reminder with annual cost, not a one-off cleanup
- You will cancel or negotiate yourself and only need drafts
Not for: Letting the agent cancel, chat with support, or change payment methods; Business SaaS renewals that need finance or legal review; Inventing fees or renewal dates that are not in your receipts
What do you need, and what do you get?
You need
- Inbox search access for receipts, or a CSV / PDF export from your card or bank
- Which services you still use weekly, monthly or rarely
- Any keep-list (must keep) and target monthly spend if you have one
You get
- A table of recurring items with merchant, amount, cadence, annualized cost and last-used note
- Flags for unused, duplicate or recent price rises when evidence exists
- Draft cancel or negotiate messages for the ones you choose; nothing sent by the agent
How does it work, step by step?
- Gather evidencePoint the agent at recent receipt emails or a card export; do not invent merchants.
- Build the listHave it annualize each recurring charge and mark unused, duplicate or price-rise flags with a source line.
- Choose actionsPick which to keep, cancel or negotiate.
- Send drafts yourselfEdit and send cancel or negotiate notes from your own email or chat; the agent never cancels.
The prompt
Our own prompt, inspired by Miles Deutscher (@milesdeutscher) and written to work in Grok Bot, ChatGPT agent and Claude. Values in lime are placeholders.
Fill in the blanks (optional; the copied prompt uses your values)
Run my monthly subscription review. Evidence I am providing: {evidence_description}. Services I still use often: {still_use}. Must-keep list: {must_keep}. Build a table with: merchant, amount, cadence, annualized cost, last evidence date, status (keep / unused / duplicate / price rise / unclear), and a short source note. Only use amounts and dates present in the evidence. Mark gaps unclear. For each item I mark cancel or negotiate, draft a short message I can send. Do not send, cancel, change payment methods or open support chats. End with annualized total and a one-line note vs my target {monthly_budget_or_none}.
Inspired by Miles Deutscher (@milesdeutscher): X thread: 10 Grok Bot use cases (second brain, repurposer, subscription audit, personal X assistant). The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Never let the agent cancel or message support for you.
- Annualize carefully: weekly and quarterly charges are easy to misread.
- A missing receipt is not proof the subscription ended; mark unknowns as unknowns.
- Keep card numbers and full statements out of places you do not control.
Our take
Subscription creep is a list problem before it is a negotiation problem. Deutscher's auditor idea is strongest when narrowed to a monthly personal routine: gather evidence, annualize, flag waste, draft messages, human sends. That last step matters. An agent that cancels on its own can drop a tool you still need or burn a support thread. Distinct from a single cancel-fee waiver playbook, this one is about the inventory and the recurring cadence. In our dry run we used a fictional month of twelve receipt lines and a card export with two duplicates and one price rise. The agent annualized correctly, flagged the unused gym and the double cloud storage, drafted a polite cancel and a negotiate note with amounts taken only from the receipts, and refused to invent a phone number for a merchant with no contact on file. Use it on a calendar reminder; do not run it against live support chats unattended.
Where the time goes back: You get one monthly pass instead of hunting receipts when a surprise charge hits.
FAQ
How is this different from a cancel-and-waive-fee playbook?
Can the agent cancel for me?
What if my bank export is messy?
Source and credit
Inspired by Miles Deutscher (@milesdeutscher). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- X thread: 10 Grok Bot use cases (second brain, repurposer, subscription audit, personal X assistant)
- Creator
- Miles Deutscher (@milesdeutscher)
- Published on
- X
- Original date
- Oct 23, 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
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. Inventory accurate; drafts only; no invented contacts or cancels.
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
- Oct 23, 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.