Automation playbooks
A playbook is a piece of company work that an agent runs end to end on your real tools. Each one below is a working standard: what starts it, what it does, where a person steps in, and what comes out the other side. Pick one from the menu to read it in full.
Ready to run
- Inbound Tender & RFP Turnaround
- Client Intake & Verification
- Weekly Operational Reconciliation
- Inquiry Triage & Assisted Resolution
How a playbook is put together
- Playbooks The end-to-end standard operating procedure. Defines the business trigger, the handoffs between systems and teams, the approval gates, and the final deliverable.
- Skills The specific competencies agents draw upon to perform the work. Modular and testable units of intelligence and document understanding.
- Blueprints The battle-tested runtime scaffolding that executes the work safely. Handles sandboxed execution, tool connections, state, and audit logs.
Reading the properties
Every playbook page opens with the same five properties. Trigger is the event that starts a run. Autonomy says how much a person is involved: supervised playbooks always wait for sign-off, exception-based ones only ask when something looks wrong, and autonomous ones read and report without changing anything. Turnaround is how quickly the work comes back, and Blueprint names the Cloudflare OS runtime pieces the playbook is built on.
Inbound Tender & RFP Turnaround
Analyzes multi-page tenders against past winning proposals, margin schedules, and compliance rules to assemble a citation-backed draft quote.
- Category
- Commercial
- Trigger
- Inbound tender notification or document upload
- Autonomy
- Supervised
- Turnaround
- Hours instead of 4 days
- Blueprint
- Document Sandbox + Vector Index + Mail Connector
Overview
Tenders arrive as long documents with requirements buried in the text. This playbook reads the whole pack, pulls out every requirement and deadline, and matches each one to the proposals you have already won. It prices the work from your own margin schedules and assembles a draft response where every claim points back to its source, so the bid team starts from a complete first draft rather than a blank page.
Skills used
- Document & requirement parsing
- Historical proposal semantic search
- Cost & margin modeling
- Draft response compilation
How a run works
- Parse the tender pack Every requirement, deadline and evaluation criterion is extracted into a structured checklist.
- Find what you have said before Past winning proposals are searched for matching answers, case studies and evidence.
- Price from your schedules Costs and margins are modelled from your rate cards and margin rules, not guessed.
- Compile the draft A citation-backed response is assembled in your template and handed to the bid lead for review.
Where people come in
Outcome
Accelerates bid response velocity while preserving commercial pricing governance.
Client Intake & Verification
Extracts company details and beneficial ownership from intake documents, verifies records against official registers, and flags anomalies.
- Category
- Compliance
- Trigger
- New client registration or form submission
- Autonomy
- Exception-based
- Turnaround
- Instant verification
- Blueprint
- Webhook Event Loop + Verification API Adapter + D1 Audit Trail
Overview
Client intake is mostly checking. Details on a form have to match certificates, registers and screening lists, and most of the time they do. This playbook does the matching on every new client, checks the official records itself, and only asks a person to look at the cases that are actually unusual. Clean onboarding completes straight away and every check is written to an audit trail.
Skills used
- Identity & certificate OCR
- Companies House register lookup
- Sanction & PEP screening
- Discrepancy risk scoring
How a run works
- Read the intake documents Company details, directors and beneficial owners are extracted from forms and certificates.
- Verify against the registers Extracted details are checked against Companies House and the verification services you already use.
- Screen and score Sanctions and PEP screening runs, and any mismatch between documents and records is scored for risk.
- Complete or escalate Clean cases are approved and recorded. Anything flagged goes to compliance with the evidence attached.
Where people come in
Outcome
Passes clean onboarding immediately; surfaces genuine edge cases directly to compliance.
Weekly Operational Reconciliation
Pulls delivery milestones and billing items from multiple operational tools every Monday, checks for mismatched records, and publishes an executive briefing.
- Category
- Operations
- Trigger
- Scheduled cron (Every Monday 06:00)
- Autonomy
- Autonomous
- Turnaround
- Delivered before Monday standup
- Blueprint
- Scheduled Worker + MCP Connector Suite + Slack/Email Publisher
Overview
Most companies keep delivery in one tool and billing in another, and someone spends Monday morning working out where they disagree. This playbook runs on a schedule, reads both sides, and lists every place a milestone and an invoice do not line up. The result is a short briefing in Slack or email before the standup, with each discrepancy linked to the records behind it.
Skills used
- Multi-database cross-referencing
- Discrepancy detection
- Milestone variance analysis
- Executive digest synthesis
How a run works
- Pull the week's records Delivery milestones and billing items are read from each operational tool through approved connectors.
- Cross-reference Records are matched across systems by project, client and period.
- Find the variances Missing invoices, unbilled milestones and mismatched values are detected and ranked by size.
- Publish the briefing An executive digest is written and posted to the channel or inbox you choose.
Where people come in
Outcome
Eliminates 3+ hours of manual data hunting across disconnected internal dashboards.
Inquiry Triage & Assisted Resolution
Reads inbound customer queries, checks verified account records using approved tools, and drafts accurate responses with full source provenance.
- Category
- Customer Experience
- Trigger
- Incoming ticket via email, portal, or chat
- Autonomy
- Supervised / Tiered
- Turnaround
- Sub-minute first draft
- Blueprint
- Webhook Ingestion + Protected MCP Gateway + Token Spend Guardrails
Overview
Inbound queries are a mix of the routine and the genuinely difficult, and both wait in the same queue. This playbook reads each one, works out what is being asked and how urgent it is, looks up the verified account record through approved tools, and drafts a reply with every fact traced to its source. Routine questions get an accurate answer in under a minute. Difficult ones reach a person with the context already gathered.
Skills used
- Intent & urgency classification
- Customer record retrieval
- Policy search & clause verification
- Resolution drafting with handover notes
How a run works
- Classify the query Intent and urgency are identified from the message and any attachments.
- Retrieve the account The customer record is fetched through the governed tool gateway, within the permissions you set.
- Check the policy The relevant policy clauses are found and verified before anything is promised.
- Draft and route A sourced reply is drafted. Routine cases are answered; complex ones go to an agent with handover notes.
Where people come in
Outcome
Resolves routine inquiries instantly while equipping agents with context-rich drafts for complex issues.