The shift the Microsoft AI Learning Summit made clear
One of the more striking moments at the summit: 80% of the global workforce reports lacking the time or energy to meet the demands placed on them. AI is starting to separate organisations that have found a way to do more with the same headcount from those that have not. If you are still doing the same administrative work manually that you did two years ago, the gap is widening.
The question to ask yourself: what proportion of your week is spent on tasks that follow a pattern — gathering, summarising, formatting, sending? That proportion is your automation opportunity.
Three levels of AI at work
The summit framed AI adoption in three levels. Most professionals are at level one — which is fine, as long as you know where to go next.
Most professionals are at Level 1. Moving to Level 3 doesn't require coding.
What agents actually do
In my time at Microsoft, Expedia, and Citi, senior professionals spent significant time on the same preparation tasks every week: scanning emails before a meeting, pulling project updates on Monday morning, tracking competitor news across multiple sites. An agent handles each of those on a schedule — you describe the task once, clearly, and it runs without you. The pattern is always the same: identify something you repeat, write down the steps as you would explain them to a new colleague, and that description becomes your agent's instruction.
Identify something you repeat. Write down the steps as you would explain them to a new colleague. That description becomes your agent's instruction. No coding required — just clarity.
The prompting insight
The quality of what AI produces is almost entirely determined by the quality of the instruction. That is a professional skills insight, not a technical one. Four elements that make the biggest difference:
- Give it a role. "Act as a senior communications advisor" produces better output than "write something for me." The role shapes tone, expertise level, and assumptions.
- State what you actually need. "Draft an email that gets my director to approve the budget before Friday, direct but not pushy" is specific enough. "Draft an email" is not.
- Give it the source material. The most accurate outputs are grounded in real documents and data, not general knowledge. If you want a summary of a report, give it the report.
- Iterate. The first output is a draft. Ask it to be more concise, adjust the tone, or take a different angle. Treat it as a collaboration, not a transaction.
Take any prompt you use regularly. Add a role ("Act as a senior analyst"), a specific outcome, and the source document. Run it and compare the output to what you got before. The difference will be immediate.
Want to put agents to work without writing code?
Make.com is one of the tools I recommend for building automated workflows visually — no developers needed. It connects your apps and lets you set up the kind of multi-step processes described above. There is a free plan to start.
Try Make.com free →This post contains an affiliate link. I only recommend tools I use or would use myself.
Your next step
Three things to try this week
- Identify the one task you repeat most that involves gathering, summarising, or formatting information. Write down the steps. That is your first automation candidate.
- Before your next important meeting, give AI a prep brief — describe the meeting, who will be there, and what you need to know. Share any relevant documents or emails with it.
- After that meeting, paste the notes into an AI tool and ask for: decisions made, action items with owners, open questions. Compare what it produces to what you would have written manually.
The professionals who get the most from AI start with one task, get it working, then build from there. You do not need a corporate programme — just a deliberate habit. For a deeper look at how agents work, see the previous post on AI assistants vs agents.
Frequently asked questions
An AI agent is a system that can carry out a sequence of actions on your behalf — without you doing each step manually. Instead of asking AI a question and reading the answer, you give an agent a goal and it acts: gathering data, making decisions, triggering actions in other tools, and reporting back. Think of it as the difference between asking a colleague a question versus delegating an entire task to them.
No. Most of the practical use cases covered at the summit — meeting summaries, email drafts, weekly project briefings — require no coding at all. They use tools already built into platforms many organisations already pay for. The key skill is knowing how to give a clear, specific instruction, not how to write code.
An AI assistant responds to a prompt — you ask, it answers, and then it waits. An AI agent acts autonomously toward a goal — it can take multiple steps, use multiple tools, and complete work without you orchestrating each move. The assistant is reactive; the agent is proactive.
Start by identifying the task in your week that you repeat most — a status update, a report compilation, a set of emails. Write down exactly what steps you take to complete it. That written process is the foundation of an agent. You don't need to build anything yet — just naming the task clearly is the first and most important step.
AI that actually works at work
Practical frameworks, real examples, and zero jargon — for professionals who want to use AI without becoming technical.
No spam. Unsubscribe any time.