What's new at Microsoft from a generative AI point of view?

Below my takeaways from yesterday's Microsoft AI #Breaking Boundaries event at Gooiland Events :

1) Licencing model: Copilot Chat Professional: Free, But at What Cost?

Offering Microsoft Copilot Chat Professional for free to Enterprise clients is a classic market penetration strategy. The motivation to do so at this point of time might have been motivated by the launch of their Chinese competitor DeepSeek AI as well.

While it might drive generative AI adoption, it also means that paying attention to AI training and (data) governance becomes more important to avoid e.g. that employees unintentionally feed sensitive company data into the public cloud or incorrect information being treated as the truth.

Microsoft did acknowledge that to a certain extent though. See below.

2) AI Adoption Stories: A Glimmer of Realism, But Still Selective

Siemens put a lot of effort in engaging the works council as well as information security professionals at an early stage. KPN indicated that success of their generative AI adoption was also by running POCs/POVs while setting up training and governance alongside. Additionally they piloted only with domain specific use cases and demanding training completion prior to granting licenses.

While interesting the case studies still felt a bit curated. E.g. the only hurdle KPN seemed to have faced was lack of a sponsor to scale their licenses.

Questions remaining: Did they address the challenges fully? What were the unexpected hurdles? What about the failures? Presenting only success stories might paint an incomplete picture and might not offer a realistic perspective on the complexities of AI implementation. I'd have preferred a more balanced and nuanced discussion of the real-world obstacles and lessons learned, even if they touched upon piloting with small use cases, training & governance.

3) Microsoft's AI Agents: Promising Data Access, But Limited Workflow Integration

From a personal productivity point of view Microsoft's AI agents show promise in their ability to access data from various systems via APIs. This is a positive step towards making information readily available to users

Although Microsoft demoed collaboration on retrieved information by the AI agents on Microsoft Pages and further team collaboration as a next step it remained unclear how Microsoft aims to achieve extended collaboration across departmental and partner siloes.

Integration with Microsoft's Power Platform orchestrator or 3rd party process orchestration tooling wasn't demoed (yet). Parties like Salesforce or ServiceNow seem to have made greater strides in this area at first glance (although at the same time Salesforce does this at the cost of some vendor lockin).

So, while API access is a good start, it doesn't address (yet) the core issue I tried to get answered at the event: how can Microsoft's AI agents truly enhance complex end to end value streams and drive meaningful business value within existing workflows? This remains a key question to me.

4. AI Governance by design: Still Somewhat Opaque

The brief demo of Microsoft Copilot Insights in combination with Purview triggered my interest to explore further how AI observability and governance is solutioned by Microsoft and whether claiming "responsible AI" or "governance by design" are buzzwords or not.

I've explored Purview previously, offering promising features to autodetect and autolabel sensitive (PII) data and assign data owners. The Copilot Insights demo seemed to focus mostly on adoption KPIs.

However what I'd like to explore further on this topic is:

• How is Microsoft addressing bias in algorithms?
• How are they ensuring data privacy?
• What mechanisms are in place for accountability?
• What is the impact of Microsoft's genai on the environment (are they e.g. able to compete with DeepSeek AI announcing less CPU power, energy and infrastruture investment is required when using their algorithms).

These are critical topics Microsoft touched upon as what they consider responsible AI but more of a deep dive on the how would have been appreciated.

5. Europe's AI Adoption: A Convenient Narrative, Even with Nuance

Lastly, Microsoft tried to boost adoption of their generative AI features by means of their Keynote at the start of the day where they indicated that Europe and within Europe especially the Netherlands is lagging behind compared to e.g. USA, China and India. Netherlands e.g. dropped from place 2 to place 8 on the global Innovation index apparently... but still time to catch up.

Microsoft's analysis of the hesitation to adopt generative AI in Europe/Netherlands is that we are more cautious about ethics and governance maybe compared to the current frontrunners.

6. Conclusion

Overall, the Microsoft AI #BreakingBoundaries event felt like a marketing exercise mostly, especially because Microsoft didn't offer opportunity to raise questions during the presentations to truly engage the audience. Also during the closing keynote the speaker indicated to look forward to talk further but once we left the keynote area all (mostly genai implementation partner) boots were closed and we were directed to the exit as the event was over.

So do I regret going? Certainly not as I was able to effectively network with other visitors of the event and saved myself quite some reading time to catch up on where Microsoft stands on their generative AI journey.

As always looking forward to your feedback.

#governancebydesign #responsibleAI #processautomationinnovation #futureofwork