AI task automation guide: how to give 30% of your job to AI

AI task automation works best when it starts from your own week. Luis Clark and James Ferrier of monday.com showed how in a hands-on workshop at aiblLIVE London 2026. The room built an agent that checks incoming work requests.

Their premise: around 30% of most jobs is chasing, reporting and reformatting that never needed a person. This free guide shows you how to find it and hand it over.

How do you decide which tasks to hand to AI?

List everything your team does in a week and sort it into two buckets. Bucket one is repeatable heavy lifting, patterns and waiting on others. Bucket two is judgement, taste, relationships and risky decisions. If a task fits neither, Luis Clark’s test is simple: does a human make this better? If not, hand it over.

They just need a process, and somewhere along the way, you’ve become that process.

Luis Clark, AI Transformation and Adoption Lead, monday.com

How do you know whether a task needs an automation or an agent?

Use an automation when the task follows a fixed rule and should run the same way every time. Use an agent when it needs context: reading, interpreting, deciding and writing. James Ferrier asks this question first for every process, because the answer changes what you build and how you test it.

How do you set up your first AI agent?

Hire it like a new starter. Write a job description covering its name, role, what it owns, what it uses, who it reports to and who signs off. Then give it four building blocks: instructions, a trigger, knowledge and tools. Grant only the access the job needs, and widen it as the agent proves itself.

How do you test an AI agent before relying on it?

Run it on a bad case and a good case, then give feedback. In the workshop, a vague “urgent, ASAP” request scored 2 out of 10 and went back for clarification. A detailed launch request passed. When the requester added detail, the agent reassessed it.

An agent is only as good as the data you actually point it at.

James Ferrier, Manager, Solution Engineering EMEA, monday.com

Key takeaways

• Sort one real week. Ask of every task whether a human makes it better.
• Write a job description first. Name, role, ownership, tools, reporting line and sign-off.
• Fix your intake. An agent that checks requests saves time downstream.
• Decide where the time goes. Reinvest it in thinking, collaboration, communication and creativity.

About the workshop

Luis Clark is AI Transformation and Adoption Lead at monday.com, and spent many years in business transformation before joining. James Ferrier is Manager, Solution Engineering EMEA at monday.com. They ran this workshop on the Efficiency track at aiblLIVE London on 20 October 2026.

What’s in the full guide

If your team’s week is full of work that never needed a person, this is a method you can start on Monday. The full PDF goes step by step.

• An eight-step method from sorting your week to testing your first agent.
• The worked example from the room: an agent that scores every incoming request.
• Templates and a checklist for the two-bucket sort, the agent job description and its instructions.
• Five actions to take back to your own organisation.

For team leads, heads of operations, PMOs and marketers. Seven pages, about a nine-minute read.