Studying PL-900 is changing how I think about AI

The biggest lesson so far hasn't been learning Microsoft's tools. It's been learning to think differently about the problems they solve.

When I first enrolled on the Microsoft PL-900 course, I expected it to be a fairly traditional certification. I imagined spending my time learning product names, remembering where each application sat within the Power Platform and working towards passing an exam.

To some extent, that's exactly what it is.

There are services to understand, terminology (soooo much) to learn and capabilities to remember. Power Apps, Power Automate, Copilot Studio, Dataverse and AI Builder all have their own role within Microsoft's ecosystem, and understanding how they fit together is an important part of the learning journey.

What I didn't expect was that the biggest change wouldn't be technical at all.

Instead, PL-900 has gradually started changing the way I look at organisational problems.

From solutions to questions

Throughout my career in education, I've naturally been drawn towards solving problems.

How can we reduce teacher workload?

How can we improve communication with parents?

How can we make intervention for students more timely?

Those questions have shaped many of the projects I've worked on over the years.

Before starting this learning journey, I probably would have jumped fairly quickly to possible solutions. If a process felt repetitive, my instinct was to ask how technology might help improve it.

Now I find myself slowing down.

Rather than asking, ‘What could we build?’, I'm beginning to ask different questions.

• Why does this process exist?
• Who is responsible for it?
• Where does the information come from?
• Which parts require professional judgement?
• Are there any parts that are repetitive enough to automate safely?
• What happens when one of the steps is wrong?

Technology should support processes that are already good, not compensate for poor ones.

The more I study, the more that idea resonates.

It's easy to become fascinated by what AI can do.

Generate emails.

Summarise meetings.

Answer questions.

Build intelligent agents.

Automate repetitive work.

Those capabilities are genuinely exciting.

But introducing AI into an inefficient or poorly understood process doesn't automatically improve it. In many cases, it simply allows the same problems to happen more quickly.

Good technology cannot replace good organisational design.

If anything, AI places even greater importance on understanding how work is currently carried out before deciding where automation genuinely belongs. Process Mining is a particularly useful tool for this analysis.

Building while I'm learning

One of the best decisions I've made is not waiting until I feel ‘ready’ before building things.

Alongside studying PL-900, I've been experimenting with Copilot Studio and Power Automate to design practical proof-of-concept solutions for schools. Every time I build something, the theory suddenly becomes much more meaningful. The Microsoft Labs are difficult to follow if the example use case is for something more commercial than I’m used to, like a Sales Forecaster, so I’ve been deliberately trying to adapt the resources to my own educationally-focused ones before I build in the developer environment.

When Microsoft talks about human approval, I immediately link it to safeguarding, data protection and professional accountability within schools.

When learning about connectors and automation, I naturally begin considering how different departments share information and where this process can become disconnected.

The certification provides the concepts. Building real solutions provides the context. It's the combination of the two that is changing how I think about AI adoption.

Together, they reinforce one another in a way that neither could achieve alone.

Governance isn't the obstacle

Perhaps the biggest surprise has been my changing perspective on governance.

Like many people, I initially viewed governance as something that slows innovation.

(ALL. THE. RED. TAPE.)

Policies.

Approvals.

Security.

Permissions.

Rules.

The more I explore practical AI adoption, however, the more I see governance differently.

Good governance doesn't exist to stop people using AI. It exists to build AI confidence and secure AI-literacy within an organisation.

If teachers know when AI is appropriate, if leaders understand where risks exist and if organisations establish clear expectations from the outset, innovation becomes much easier rather than more difficult.

Trust doesn't emerge by accident.

It develops because people understand the boundaries within which new technology operates.

Looking ahead

I'm still very early in this journey.

There is a huge amount left to learn, both technically and strategically.

I'm sure some of the opinions I've written here will change over the coming months as I complete PL-900, move on to PL-200 and continue building more sophisticated solutions.

In many ways, I hope they do.

If this Staffroom is going to be an honest record of learning, then changing my mind should be seen as progress rather than inconsistency.

For now, though, one lesson already feels clear.

PL-900 isn't simply teaching me how Microsoft's products work.

It's teaching me to ask better questions before I start looking for answers.