An operations manager can use AI without writing code by giving chat assistants such as ChatGPT or Claude the reading and drafting work that fills an operations week: first drafts of SOPs, sorting inbound requests, turning a week of numbers into a summary, comparing vendor quotes, and stress-testing a forecast, with a person checking every output before it is used. Add one simple no-code automation once those habits are steady. The skill is not the tool. It is choosing the right task, keeping sensitive data out, and checking the result.
This piece is for people who run operations and want time back. If you want to build automations as a job, what an AI automation engineer does covers that route.
Where does AI actually help in operations?
Wherever the work is reading, writing or summarising text, and a person can check the result quickly. It helps least where the answer depends on facts it cannot see, such as your live stock levels, or where a mistake is expensive and hard to spot.
| Task | What AI does well | What you still do |
|---|---|---|
| SOPs and documentation | Turns rough notes or a recorded walkthrough into a clear, numbered procedure | Check every step against how the work is really done |
| Inbound request triage | Sorts emails or tickets into categories and drafts first replies | Set the categories, approve replies, handle exceptions |
| Reporting summaries | Turns a table of weekly figures into a plain-language summary | Check every number it quotes against the source |
| Vendor and contract review | Compares quotes side by side and lists clauses worth a closer look | Read the contract; send legal questions to someone qualified |
| Forecasting support | Explains a trend, lists assumptions, suggests scenarios to test | Own the forecast and the numbers in it |
How do you use AI to write SOPs?
Start from what already exists, not a blank page. Record yourself or a colleague doing the task and talking through it, or paste in rough notes, and ask the assistant to turn it into a procedure with a purpose, numbered steps, the exceptions, and who to ask when something goes wrong. Then give the draft to someone who does the task and have them follow it literally. Every place they hesitate is a gap.
AI is good at structure and plain language, and bad at knowing the unwritten step, such as "check the second label because the printer sometimes duplicates". That is why the check is not optional.
How can AI triage inbound requests?
Write down your categories first: for example, order change, delivery problem, invoice query, supplier update, other. Paste a batch of requests (with personal details removed, see below) and ask the assistant to label each one, with a short reason, and to draft a reply for the common categories. Review the labels: where it disagrees with you, your categories are unclear or the request is ambiguous, and both are worth knowing.
In a chat window this saves reading time; inside a no-code automation it becomes a queue that sorts itself while a person approves what goes out.
Can AI help with vendors, contracts and forecasts?
Yes, with care. For vendors, AI can lay three quotes side by side in one table, flag differences in delivery terms or minimum orders, and draft questions to ask each supplier. For contracts, it can summarise and point to clauses worth reading closely, such as auto-renewal, liability or termination. It is not legal advice, it can miss or misread a clause, and anything you would normally send to a lawyer still goes to one. Check whether your company allows contracts to be pasted into an AI tool at all.
For forecasting, use AI as a thinking partner, not a calculator. Ask it to explain what might drive a pattern in your demand history, list the assumptions behind your forecast, and suggest scenarios to test. Do the arithmetic in your spreadsheet or planning tool, where you can see every formula, and write the caveats on the forecast itself.
What is a good first no-code automation?
One small, boring, repeated task with a clear trigger and a person approving the output. A good first automation:
- Trigger: a new email arrives in the shared operations inbox.
- AI step: label it with one of your categories and draft a reply.
- Output: the label and draft go to a sheet or a chat channel.
- Approval: a person reads the draft and sends it, or not.
- Log: every item and decision is recorded.
Tools such as n8n, Make and Zapier do this without code. Have it write to a sheet only, run it alongside your usual process for a week, and compare before it does anything that leaves the building.
What should you never put into an AI tool?
Follow your company's AI policy first; if there is none, ask for one. Without one, a safe default is to keep these out of any AI tool that your company has not approved for them:
- Customer or staff personal details: names with addresses, phone numbers, identity documents, health or payroll information.
- Passwords, access codes and system credentials.
- Confidential contract terms, pricing agreements and commercial information you are bound not to share.
- Unreleased financial results.
Before pasting a batch, strip the names and replace them with references such as "Customer A". Prefer an approved business version of a tool, which usually has clearer data terms. How to write an AI usage policy covers what a policy should say.
Who should approve what AI produces?
Match the approval to the risk: the person who does the task approves the SOP, whoever signs the purchase order approves the vendor pick. A rule that works: AI may draft anything, and a named person approves anything that reaches a customer, a supplier or a system of record. Human-in-the-loop oversight goes further.
How do you know it is saving time?
Measure hours returned, not tasks tried. Before you start, time one task for a week: how long SOP updates, inbox sorting or the weekly report take by hand. Then time the same task with AI, including the checking. If the checking eats the saving, drop it or change the approach. Keep a one-line log per task: time before, time after, errors caught in review. Two or three honest numbers beat a list of tools.
Where should you start this week?
- Pick one task you do every week, and time it.
- Check your company's AI policy and which tools are approved.
- Rewrite one SOP from rough notes with AI, and have a colleague follow it literally.
- Summarise last week's report with AI, and check every number against the source.
- Write down three candidates for a first no-code automation, with who would approve the output.
Square 1's AI for Operations is an on-demand course, recorded by an instructor and graded by Nova, the AI tutor: four modules and about 6.5 hours covering AI in the operations week (SOPs, checking, handovers), forecasting and planning, and procurement and vendors, ending in an operations playbook. Its two graded projects are an SOP set and the playbook, and it asks only that you run operations; no code. Agents with n8n and Make is the next step for automation, also no code, with three processes automated, and the twelve-week AI Automation Engineer Bootcamp is live on Zoom for those who want to build automations as a job. All are on a waitlist today. The operations manager role page shows how these skills fit an operations career, and the free AI for operations skill check tells you where to start.