Operator notes7 min read
How to use AI in marketing operations
A practical guide to how to use AI in marketing operations: sort your weekly tasks, start three builds, and know what to keep human.

How to use AI in marketing operations starts with sorting your recurring tasks into three buckets: what AI does alone, what it does with you, and what stays yours. Then build three small automations for the repeatable, checkable tasks. Keep a person on anything a client or customer sees before it goes out.
Marketing is a sensible place to start. In McKinsey's global survey, revenue increases from AI were most commonly reported in marketing and sales, strategy and corporate finance, and product and service development. That shows where people are finding value, not what will happen on your team.
The method underneath is the Automation Audit we run in Week 1 of every cohort in Delta Camp's AI training. I explain the full version in what tasks to automate first, so this post applies it to one role. I haven't run every build below inside your stack, so test each one on a small sample first.
A typical week in marketing operations
These hours are typical, not measured. They're my rough picture of the recurring work in the role, so swap in your own numbers.
- Build the weekly performance report from ad, email and web numbers (about 2 to 3 hours).
- Turn campaign requests into briefs and chase the missing details (about 2 hours).
- Set up and check campaigns, including links, UTM tags and audiences (about 3 hours).
- Keep the content calendar current and chase approvals (about 2 hours).
- Repurpose one piece of content into posts, emails and snippets (about 3 hours).
- Tidy lead and contact lists before they reach the CRM (about 2 hours).
- Write meeting notes and status updates for stakeholders (about 2 hours).
Write your own version down with a time against each line. The audit starts there because you can't sort a task you haven't counted.
Which marketing operations tasks can AI do alone?
Sort each task by one question: how fast can you tell the output is wrong? Fixed input, fixed output and a quick check means AI-only.
AI-only
- Tag incoming campaign requests by type and channel, because the categories are fixed and a wrong tag is obvious at a glance.
- Build UTM tags from your naming convention, because the rule is fixed and a bad tag fails a quick check.
- Reformat exported numbers into your report template, because the layout never changes and you can compare it with the source.
Human + AI
In Microsoft's survey of knowledge workers who use AI across ten markets, 86% said they treat AI output as a starting point, and half named quality control of AI output as a human skill that matters more.
- Draft the commentary on the weekly report with AI, then add the reason the numbers moved, because only you know what changed.
- Write campaign briefs from a request with AI, then fill the gaps with the requester.
- Turn one piece of content into posts and email snippets, then edit for voice and claims.
Human-only
- Decide which channels and campaigns get budget, because it's a trade-off AI can't see from a spreadsheet.
- Sign off any claim, offer or price before it reaches a customer, because a wrong one is hard to take back.
- Manage agency and stakeholder relationships, because trust is built between people.
How to use AI in marketing operations: three builds for this week
Each build has a trigger, steps and tools. Start with the first, because it repeats every week.
Build 1: the Monday report draft
Trigger: a schedule. In Make, one scenario can run at several set times from one Daily schedule setup, so set it for early Monday.
- Pull last week's numbers from your sheet or exports.
- Send them to ChatGPT or Claude with a fixed prompt: summarise what changed under your usual headings, and say so when a figure is missing.
- Save the draft to a doc and notify yourself.
- Check it against the source, then add the why.
Tools: Make, a spreadsheet, ChatGPT or Claude.
Build 2: request intake
Trigger: a new form response. In Zapier, a trigger is the event that starts a Zap, and Zapier watches for it once the Zap is on.
- Add an AI step (ChatGPT or Claude) that tags the request type and channel.
- Have it draft a brief skeleton and a list of missing details.
- Post both to your team channel and tag the owner.
- Keep it internal, so nothing goes to the requester until a person has read it.
Tools: Zapier, your form tool, ChatGPT or Claude.
Build 3: content repurposing with an approval step
Trigger: a content calendar row marked ready.
- Have an n8n AI Agent draft three social posts and one email snippet from the finished piece.
- Put the posting tool behind approval, since n8n documents human review for tools with higher risk, such as sending messages, modifying records or deleting data.
- Edit the drafts and approve only what you'd sign yourself.
Tools: n8n, ChatGPT or Claude.
What should you not automate in marketing operations?
In McKinsey's survey, 51% of respondents from organisations using AI reported at least one negative consequence, and inaccuracy was the most common. In marketing, the cost lands when the error reaches a customer. The cheapest fix is a person reading before anything leaves.
Your team is probably using AI already. In BCG's survey of employees across nine APAC markets, 78% used AI at least weekly and 35% said they would bypass company restrictions to use it. Write your rules before the habits set.
- Keep customer and lead personal data out of any tool until you know where it goes and who approved it.
- Check any audience or exclusion rule for bias before AI applies it, because a skewed audience is hard to spot afterwards.
- Read every email, ad or reply a customer will see before it goes out.
- Leave spend, claims, prices and legal wording to a person.
- Skip anything you'd have to read line by line to trust.
For a team, add a fourth column to your audit. Flag any task that touches sensitive data, anything a customer sees before a person checks it, or a regulated process.
A flag means a person stays in the loop, not that the task can never be automated. The four-hour tax covers what checking AI output costs you.
Map your own week
Open a sheet, list your recurring tasks with a time against each, then take the free AI Maturity Audit. It takes about three minutes on the individual track and about six on the business one. The business track puts a risk flag on each opportunity it finds.
This week, sort your most frequent task into one of the three buckets before you open any tool.
Frequently asked questions
What is AI used for in marketing operations?
Mostly the repeating work around campaigns: report drafts, request triage, UTM tagging and turning one piece of content into several. A person still checks anything that reaches a customer. The best starting point is the task you do most often.
Which tools do I need to automate marketing operations with AI?
Make, Zapier or n8n move data between your apps, and ChatGPT, Claude or Copilot do the AI step. Pick the one that fits the stack you already use. You don't need to code, but you do need to describe the task precisely and check the output.
Is it safe to put customer data into AI tools?
Not by default. Check where the data goes and who approved the tool, and use sample or synthetic data while you build. If a task touches sensitive data, keep a person in the loop and decide in writing how output is checked.
How do I decide what to automate first in marketing operations?
List your recurring tasks, put a time against each, then sort them into AI-only, Human + AI and Human-only. Start with the most frequent task that is quick to check, because frequency beats size almost every time. I walk through the method in the audit-before-tools post.
Where the numbers come from
- McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation, November 2025.
- Microsoft WorkLab, 2026 Work Trend Index report: Agents, human agency, and opportunity, May 2026.
- Boston Consulting Group, Asia Pacific Leads the World in AI Adoption While Grappling with Job Fears, BCG Survey Finds, October 2025.
- Zapier Help Center, Set up your Zap trigger, May 2026.
- Make Help Center, Schedule a scenario - Help Center, 2026.
- n8n Docs, Human-in-the-loop for tools, 2026.

