AI Prompts for Operations

What is in this pack

AI prompts for operations cover the repeatable, process-heavy work that keeps a team running: writing SOPs, drafting status updates, running retrospectives, tracking blockers, and turning meeting notes into action items. This page is for operations managers, chiefs of staff, and team leads who want consistent outputs without rewriting the same prompt from scratch every time.

The pack contains 20+ prompts across five categories: process documentation, meeting and communication, project tracking, vendor and procurement, and team reporting. Each prompt is ready to copy into ChatGPT, Claude, Copilot, or Gemini with no modification beyond swapping the bracketed variables.

A few things this pack does not do well: it will not replace judgment calls on headcount, budget approval, or anything that requires knowing your organization's internal politics. Use these prompts to handle the drafting and structuring work, then apply your own context before anything goes out.

If you want to see how similar packs are built for other functions, the full prompt library by role covers everything from marketing to finance.

The prompts, by category

These 20 prompts cover the most common operations tasks: process documentation, vendor management, meeting efficiency, resource planning, and incident response. Each one is written to run as-is, with variables you swap for your own context.

The table below shows how they map to common ops roles.

One honest caveat before you start: these prompts work best when you already know what good output looks like. If you have never written a vendor scorecard or a runbook before, the AI will give you something plausible that may miss your organization's real requirements. Use a colleague's existing doc as a reference point when you can.

Process documentation

Write a standard operating procedure from rough notes

When to use it: You have a process in your head or in messy notes and need a clean, followable SOP. What to change first: Replace [PROCESS NAME] and paste your notes into the designated field.

[NEEDS REAL OUTPUT] What to notice: How the model handles ambiguous steps, and whether the "common errors" section reflects your actual pain points or defaults to generic warnings.

Convert a process diagram description into a written procedure

When to use it: You have a flowchart or whiteboard photo description but need a text version for a handbook. What to change first: Swap [DIAGRAM DESCRIPTION] with your own narrated walkthrough of the diagram.

Identify gaps in an existing SOP

When to use it: You have an SOP that gets ignored or generates repeated questions. You want to know why. What to change first: Paste the full SOP text into [EXISTING SOP].

Write a process change announcement

When to use it: A process is changing and you need to tell the team clearly without a wall of text. What to change first: Fill in [OLD PROCESS], [NEW PROCESS], and [EFFECTIVE DATE].

Vendor management

Draft a vendor evaluation scorecard

When to use it: You are assessing two or more vendors and want a consistent framework. What to change first: Replace [VENDOR TYPE] and [CRITERIA LIST] with your actual priorities.

[NEEDS REAL OUTPUT] What to notice: Whether the model's "5/5" descriptors are specific enough to use, or too vague to distinguish between vendors in practice.

Write a vendor performance review summary

When to use it: You have raw data or notes from a quarterly vendor review and need a clean summary for stakeholders. What to change first: Paste your notes or data into [PERFORMANCE DATA].

Draft a vendor contract renewal checklist

When to use it: A vendor contract is coming up for renewal and you want to make sure nothing gets missed. What to change first: Replace [VENDOR TYPE] with the specific service category.

Write a vendor off-boarding notice

When to use it: You are ending a vendor relationship and need a professional, clear written notice. What to change first: Check [NOTICE PERIOD] against the actual contract before sending.

Meeting efficiency

Write a meeting agenda from a one-line goal

When to use it: You have a meeting on the calendar but no agenda, and you need one fast. What to change first: Replace [MEETING GOAL] with a single, honest sentence about what the meeting needs to decide or produce.

Summarize a meeting transcript into decisions and actions

When to use it: You have a raw transcript or rough notes and need a clean record people will actually read. What to change first: Paste the transcript into [TRANSCRIPT].

[NEEDS REAL OUTPUT] What to notice: Whether the model correctly distinguishes between "we discussed doing X" and "we decided to do X." That distinction matters for accountability.

Write a follow-up email after a meeting

When to use it: You want to send a clear post-meeting summary without writing it from scratch. What to change first: Paste your action list into [ACTIONS] so the email reflects real commitments.

Create a recurring meeting effectiveness review

When to use it: A standing meeting feels like a waste of time and you want a structured way to audit it. What to change first: Replace [MEETING NAME] and [ATTENDEE COUNT] with the specific meeting you are questioning.

Resource planning

Draft a capacity planning summary for a quarter

When to use it: You are heading into a planning cycle and need to present team capacity clearly. What to change first: Replace [TEAM DATA] with your headcount, confirmed absences, and known project load.

[NEEDS REAL OUTPUT] What to notice: Whether the model flags the right constraints. If the output looks balanced but your gut says otherwise, it is probably because [TEAM DATA] was too sparse.

Write a resource request justification

When to use it: You need to make the case for a new hire, contractor, or tool to someone who will say no by default. What to change first: Replace [BUSINESS IMPACT] with something measurable, not a description of how busy you are.

Create a RACI matrix from a project description

When to use it: A project has started but ownership is unclear. You want a RACI before the confusion compounds. What to change first: Paste a real description of the project into [PROJECT DESCRIPTION], not a generic one.

Incident response

Write an incident report template

When to use it: You need a consistent format for post-incident documentation across your team. What to change first: Replace [INCIDENT TYPE] with the category most relevant to your ops context (IT outage, supply chain disruption, compliance breach, etc.).

Draft a stakeholder communication during an incident

When to use it: Something has gone wrong and you need to tell stakeholders clearly without creating more panic. What to change first: Be specific in [WHAT HAPPENED] and honest about [CURRENT STATUS]; vague updates erode trust faster than bad news.

[NEEDS REAL OUTPUT] What to notice: Whether the model softens the facts to sound reassuring. If it does, edit the first paragraph to be more direct. Stakeholders notice when the update is rosier than the reality.

Run a blameless post-mortem facilitation guide

When to use it: An incident is resolved and you want to run a post-mortem that produces useful changes, not finger-pointing. What to change first: Replace [INCIDENT SUMMARY] with the facts, not your interpretation of what went wrong.

Write a risk register entry from a near-miss description

When to use it: Something almost went wrong and you want to log it properly before it is forgotten. What to change first: Replace [NEAR-MISS DESCRIPTION] with what actually happened, including how close it

Real example outputs

Five of the prompts above have been run against real models so you can see what the output actually looks like. Those sections are marked [NEEDS REAL OUTPUT] pending capture from a live session. Once outputs are added, each will include a "what to notice" line pointing to the specific detail that makes the result useful, whether that is the way the model surfaces a hidden dependency, flags an ambiguous owner, or structures a decision log entry you would actually keep.

One honest caveat: AI models do well at drafting SOPs and summarizing meeting notes, but they will confidently omit steps they were never told about. Any output that describes a physical process or a compliance-sensitive workflow needs a human review before it goes live. Do not skip that step.

How to adapt these for your team

The prompts above work as starting points. Most teams get better results by making two or three changes before saving a prompt as a shared template.

Swap in your own terminology. If your team calls them "incidents" not "issues", or "workstreams" not "projects", edit the prompt before saving it. Consistent language across prompts reduces the chance of a model returning generic output.

Add a constraint that reflects your actual situation. A prompt asking for a root-cause summary hits differently when you add "our team has no engineering resource on Fridays" or "decisions must be reversible within 24 hours." Context like that shapes the output in ways that role instructions alone do not.

Remove steps that don't apply. Several prompts here include optional sub-tasks. Cut what your team won't use. A shorter, precise prompt is harder to misuse than a long one with escape hatches.

One honest caveat: these prompts don't know your org chart, your existing tooling, or your team's current priorities. The first run often surfaces gaps in the prompt itself, not problems with the model.

Save this folder to your workspace

These prompts work best when your whole team is pulling from the same source. If everyone saves their own version and edits it locally, you end up with drift: one person's vendor-review prompt asks for cost breakdowns, another's doesn't, and the outputs stop being comparable.

Convergence's shared folders for teams let you store this pack in one place and push updates to everyone at once. When you refine a prompt based on real output, the improvement is immediately available to the whole team rather than sitting in one person's chat history.

A few practical notes before you do:

• Rename any prompt that references a process your team has already retired.
• Add your company's terminology as a variable at the top of each prompt, so no one has to remember to include it manually.
• Assign one person to own the folder and review it quarterly.

The full prompt library by role has companion packs for other functions if you want to build this out further.

Frequently asked questions

Do these prompts work in ChatGPT, Claude, and Copilot equally well?

Yes, every prompt in this pack is written to work across all major models. Slight differences in output formatting are normal, particularly with tables and numbered lists, but the substance holds.

Can I use these prompts without any AI experience?

You can start using them immediately. Copy a prompt, fill in the bracketed variables, and paste it into whichever tool your team already uses.

What if the output is too long or too formal?

Add a line at the end of the prompt: "Keep it under [X] words, plain language, no jargon." That single instruction fixes most formatting complaints.

Are these prompts suitable for regulated industries?

Use caution before pasting confidential data into any public AI tool. Review your organization's data-handling policy first, and substitute placeholder names or figures when testing new prompts.

How often should I update my prompt library?

Review your saved prompts whenever a process changes significantly. Quarterly is a reasonable default for most teams.