Power Platform AI — hub, canvas apps and generative pages

The Power Platform surfaces, and a card whose real subject is not the surfaces at
all — it is where AI sits inside a process.

The building blocks

AI hub. The front door to Power Platform AI capability: prebuilt models,
prompts and agents in one place. Where a maker starts before building anything.

AI in a canvas app. Adding AI into an existing business process, inside the
app people already use. Prompts and prebuilt models called as part of the screen
flow. And the line that matters: the process design comes first — you decide
where AI changes the step, then you build it.

Code-first generative pages. Generated interface for a scenario rather than
hand-built screens. Pair them with an agent feed so users can see what agents
have actually done.

Designing a business process that includes AI

Three moves, in order.

Map the current step. What the person does, and what they wait for. The
waiting is usually where the value hides.

Insert AI where judgement is repeated — classifying, extracting, drafting,
summarising. Not where the answer must be exact, because that is a query.

Keep the human where it counts. Consequential outcomes keep an approval step,
and reversibility decides how heavy that step needs to be.

Where each surface fits

AI hub — discover and manage.
Canvas app — AI inside a process.
Generative pages — code-first interface.
Agent feed — visibility of agent activity.

That last one is easy to skip and worth keeping: if people cannot see what an
agent has done, they will either distrust all of it or notice none of it.

Exam trap. Power Platform answers are usually about where AI sits in a
process, not which control to drop on a screen. An option that adds a clever
AI feature without changing who decides or what they wait for has not improved
the process.

The one line to carry: if nobody decides differently and nobody waits less,
nothing has been improved — however good the feature looks.