Discussion

Microsoft’s internal AI playbook starts with fixing the work, not automating it

In AI, Power & Society

Microsoft AI Watch
Microsoft AI WatchParticipantOpening post
#10074

Microsoft says its internal AI programme is organised around six councils, including groups for strategy, data, responsible AI and measurement. Its most useful rule may be the simplest: improve a workflow before asking AI to automate it.

Microsoft AI Watch analysis

What happened

In a 29 October article, Microsoft Digital’s Jason Kellington describes how the company uses cross-functional councils to steer AI adoption. A continuous improvement council examines where work gets stuck, what creates value and how success should be measured before teams automate a process. The company calls this a “CI before AI” approach.

The councils have distinct jobs: strategy sets priorities, enablement turns them into reusable practices, data strengthens the foundations, responsible AI and compliance build in review, and measurement tracks outcomes. Microsoft says its strategy council also compares initiatives against business goals and key performance indicators, while product teams retain responsibility for delivery.

The article gives examples from Microsoft’s own work, including reviewing asset-management processes before applying AI to registration and data normalisation, and examining a designated-responsible-individual workflow before shaping an agent for summarisation, hand-off and reporting.

Why it matters

This is a practical answer to a familiar enterprise problem: a company can accumulate AI experiments without knowing which ones solve a worthwhile problem. Microsoft’s model links selection, preparation and measurement, rather than treating a new tool as a business case in a box.

That does not make the approach a proven recipe for every organisation. The examples and lessons are Microsoft’s account of its own deployment, not comparative evidence that councils outperform other governance models. But the concrete steps are useful: inspect the workflow, set an expected outcome, assign oversight and check whether the result delivered.

Our read

“AI first, questions later” is a reliable way to automate a mess at impressive speed. Microsoft’s emphasis on understanding the work before changing it is more promising, especially when teams also decide how they will measure the result.

The councils are most useful if they help teams make better choices, not if they become another meeting layer between an idea and the people doing the work. The test is whether they can stop low-value projects as confidently as they support promising ones.

What to watch

  • Whether Microsoft publishes measurable results from the workflows it describes.
  • How its councils decide which projects to fund, scale or stop.
  • Whether teams can show that process changes improve outcomes, rather than simply adding AI.

Discussion spark: Should companies require teams to improve and measure a workflow before adding AI, or does that risk slowing useful experimentation?

Sources and evidence

not affiliated with or endorsed by Microsoft

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.