Agentic AI for Professionals
Use AI agents to improve your own work, and stay responsible for the quality of what they produce.
Is this your path?
This path is for people who do knowledge work and want agents to take on the repeat parts of it: analysts, executives' assistants, HR, finance, sales, marketing, operations and administration roles.
If you run a team's process, see the Managers path. If you build technical systems, see the Builders path.
- Spot where AI agents help in your own role, and where they do not
- Build simple assistants and workflows for your repeat tasks
- Check an agent's output and decide to use, verify, reject or escalate it
Your three levels
Not sure where to begin? Start at Level 1 if agents are new to you. If you already use AI at work every day, Level 2 is a good start. You do not need to take all three.
Level 1 · 8 hours · Understand and chooseAgentic AI Essentials
Understand what AI agents are, find the right opportunities in your own role, and learn to use them safely.
A personal AI opportunity map, a safe-use checklist and one small task chosen for you to practise on.
- What AI, generative AI and agents are, and how assistants differ from automation
- What agents can and cannot do, and which tools are approved for your work
- Privacy, intellectual property and safe use
- Listing your tasks to find repetitive work, delays and avoidable rework
- A simple test: is an agent right for this task? A quick check of value and risk
Level 2 · 16 hours · Apply and buildAgentic AI Work Assistants
Build simple assistants and workflows for your real tasks, and check what they produce. Includes a short refresher on generative AI.
A verified work product, a reusable set of instructions, a simple personal workflow and a plan for using it at work.
- Writing clear instructions: goal, audience, context, sources and format
- Using agents for research, documents, analysis, meetings, email, presentations and reports, including files and data
- Building reusable instruction libraries and simple personal assistants and automations
- Using Microsoft Copilot where it fits
- Verifying results, and comparing time, quality and rework against doing it by hand
Level 3 · 8 hours · Test and controlAgentic AI Quality and Guardrails
Judge what an agent gives you, set the limits it works within, and decide when to trust it and when to step in.
A judgment record: the evidence you checked, the uncertainty, the options you weighed and your final decision.
- Telling fluent output from real evidence, and checking sources
- Spotting made-up facts, missing context, bias and over-reliance
- Framing the problem, weighing alternatives and handling uncertainty
- Setting boundaries for what an agent may do on its own
- Deciding to use, correct, verify further, reject or escalate
Function and sector editions
The same method, applied to one industry with its own examples and controls.