Operations
Turn SOPs into skills, map processes, review vendors, analyse inventory with real calculations, schedule with a solver, keep clean incident logs, run a weekly ops review, and wire AI into n8n, Zapier, Make or Copilot Studio with approvals.
For Operations managers and COOs, supply chain and procurement leads, office and facilities managers, business operations analysts
Run the weekly ops cycle with fewer manual steps and a person on every irreversible action
- Each core process written as an SOP skill an assistant follows step by step
- The weekly ops review pack drafted and waiting before Monday's meeting
- Reorder points and ABC classes computed from your data, not estimated
- Every incident logged with timeline, impact, causes and owners within 24 hours
- Automations where every order, payment and customer message waits for a person
On this page
- Where AI helps in this job
- Start in 15 minutes
- The tools people in this role use
- Members only: Workflow: turn an SOP into a skill
- Members only: Workflow: map a process and find the waste
- Members only: Workflow: vendor review and scorecard
- Members only: Workflow: inventory and supply analysis
- Members only: Scheduling: let a solver do the hard part
- Members only: Incident logs that teach you something
- Members only: Automation tools that exist, and where AI fits
- Members only: Data quality first
- Members only: A skill for operations
- Members only: Guardrails and hooks
- Members only: Routines that run without you
- Members only: Prompt ladder for operations
- Members only: Measure it
- Members only: Risks and what not to automate
Operations is a loop that never stops: processes that must run the same way every time, suppliers to manage, stock to balance, rotas to fill, incidents to learn from and a review every Monday. AI helps with the writing and the analysis around that loop. It is bad at arithmetic in prose, bad at constraint puzzles, and it acts on bad data as confidently as on good data. So on this page, SOPs become skills, numbers come from code, schedules come from a solver, and anything that spends money or reaches a customer waits for a person.
Where AI helps in this job#
| Task | What AI does well | What you still own |
|---|---|---|
| SOPs | Writes and updates SOPs from a walkthrough, then turns them into skills | Whether the process is right, and the sign-off |
| Process mapping | Turns a description into a flowchart, counts handoffs and waits | Walking the process with the people who do it |
| Vendor management | Summarises contracts, computes spend and SLA trends, drafts QBR questions | Negotiation and the renew or exit decision |
| Inventory and supply | Computes ABC classes, safety stock, reorder points and days of cover with code | Service-level targets and buying decisions |
| Scheduling | Writes the rules and a solver model, explains what could not be met | Fairness calls and exceptions |
| Incident logs | Turns threads and notes into a structured, blameless record | Confirming causes and owning the actions |
| Weekly ops review | Pulls KPIs, flags exceptions and drafts the pack | Priorities and decisions |
| Automation | Designs flows in n8n, Zapier, Make or Copilot Studio and writes the AI steps | Approvals, credentials and monitoring |
| Data quality | Finds duplicates, blanks and impossible values, writes validation checks | Fixing master data at the source |
AI is weak in four places that matter in operations. It does arithmetic badly when it writes numbers in prose instead of running code. It cannot reliably solve constraint problems such as a rota with skills, rest rules and leave. It does not know how your floor, warehouse or office actually runs. And when a lead time is missing, it may fill one in. Give it code execution, a solver and clean data, and keep the decisions.
Start in 15 minutes#
These work in Claude, ChatGPT, Gemini or Microsoft Copilot (formerly Microsoft 365 Copilot). For the third one, use an assistant that can run code on a file.
Prompt
Turn rough notes into an SOP
Turn these notes into a standard operating procedure with these sections: purpose, scope, trigger, roles, numbered steps with one action each and the check that ends each step, exceptions and who to call, and records to keep. Keep my wording for system names, thresholds and roles. Wherever my notes do not say something, write [MISSING] instead of guessing. Notes: [PASTE NOTES]
A good SOP has one action per step and shows its gaps as [MISSING]. Ask the person who does the job to follow it once, out loud, and fix every place they hesitate.
Prompt
Draw a process as a flowchart
Draw this process as a Mermaid flowchart. Use one box per step with the role in it, a diamond for each decision, and label each arrow with its condition, like this example:
flowchart TD
A[Customer: submits return request] --> B{Agent: within 30 days?}
B -- yes --> C[Agent: sends return label]
B -- no --> D[Agent: escalates to team lead]
After the diagram, list every handoff between roles and every step where work waits.
Process: [DESCRIBE THE PROCESS]Paste the code into anything that renders Mermaid: GitHub does in Markdown files, issues and pull requests. Check the map with the people who do the work before you trust it.
Prompt
Find data problems in an export
Check the attached export for data problems before I analyse it. Use code and do not change or drop any data. Report: duplicate IDs in [ID COLUMN], blanks in [REQUIRED COLUMNS], impossible values (negative quantities, future dates, lead times over 365 days), inconsistent units or spellings in [COLUMN], and values more than 3 standard deviations from the mean in [COLUMN]. For each problem, give the count, up to five example rows and the likely cause, like this example: "Duplicate sku: 14 rows. A101 appears on lines 3 and 4 with different unit costs. Likely cause: two suppliers entered under one item."
Check two or three of the counts with a filter in your spreadsheet. A good report tells you what to fix at the source, not just what is wrong.
The tools people in this role use#
- An assistant that runs code. For any number, use an assistant that computes with code on your file. Claude for Excel, PowerPoint and Word are generally available on all paid Claude plans if you prefer to work inside the workbook.
- Automation platforms with AI steps. n8n has an AI Agent node and a Human review step for tool calls, and you can self-host it under its Sustainable Use License. Zapier MCP lets Claude, ChatGPT, Cursor and other AI clients act in more than 9,000 apps, and each MCP tool call uses two tasks of your plan. Make offers AI Agents on all plans, across more than 3,500 integrations, and an MCP server. Microsoft Copilot Studio agent flows combine AI actions, approvals and schedule triggers in deterministic flows.
- Anthropic's operations plugin. Free and open source, for Claude Cowork and Claude Code. Its skills include process-doc, runbook, vendor-review, change-request, capacity-plan, status-report, risk-assessment, compliance-tracking and process-optimization. The plugins guide shows how to install and adapt one.
- A solver for scheduling. Google's OR-Tools (Apache 2.0) includes the CP-SAT solver and a worked employee scheduling example.
Workflow: turn an SOP into a skill
Members only below this line
Get the full Operations playbook free
What you get on this page:
- 4 step-by-step workflows
- 13 prompts
- 1 skill
- 1 hook
- 2 routines
Also in the members part
- Workflow: map a process and find the waste
- Workflow: vendor review and scorecard
- Workflow: inventory and supply analysis
- Scheduling: let a solver do the hard part
- Incident logs that teach you something
- Automation tools that exist, and where AI fits
- Data quality first
- A skill for operations
- Guardrails and hooks
- Routines that run without you
- Prompt ladder for operations
- Measure it
- Risks and what not to automate
Sources
Checked against these on 13 Oct 2026.
- Anthropic knowledge-work-plugins: operations plugin (opens in a new tab)github.com
- n8n docs: AI Agent node (opens in a new tab)docs.n8n.io
- n8n docs: Human-in-the-loop for AI tool calls (opens in a new tab)docs.n8n.io
- n8n licence (Sustainable Use License) (opens in a new tab)github.com
- Zapier MCP (opens in a new tab)zapier.com
- Make: AI Agents (opens in a new tab)make.com
- Microsoft Learn: Agent flows overview (Copilot Studio) (opens in a new tab)learn.microsoft.com
- Google OR-Tools: Employee scheduling with CP-SAT (opens in a new tab)developers.google.com
- Google SRE book: Postmortem culture, learning from failure (opens in a new tab)sre.google
- GitHub Docs: Creating diagrams (Mermaid) (opens in a new tab)docs.github.com
- Claude for Excel, PowerPoint and Word availability (opens in a new tab)claude.com
- Claude Help Center: Use skills in Claude (opens in a new tab)support.claude.com
- Claude Help Center: Use connectors (tool permissions) (opens in a new tab)support.claude.com
- Claude Help Center: Schedule recurring tasks in Claude Cowork (opens in a new tab)support.claude.com
- Claude Code docs: Hooks reference (opens in a new tab)code.claude.com
- Claude Code docs: Routines (opens in a new tab)code.claude.com
- Microsoft Support: Schedule your most used Copilot prompts (opens in a new tab)support.microsoft.com
- Gemini Apps Help: Schedule actions in Gemini Apps (opens in a new tab)support.google.com