Op-Ed: Inevitably, Microsoft merges Copilot with 365 in new ‘superapp’ on August 18

In March, Microsoft announced Wave 3 of Microsoft 365 Copilot including “Agent 365” in the First Frontier Suite. The new development looks very like an evolution of this idea, merging Copilot with the standard Microsoft Office 365 stalwarts.

The Microsoft merger announcement goes into quite a lot of detail as to how Microsoft accounts are affected. Existing accounts don’t need to do anything.

As of August 18, Group Chat, Podcasts, and Deep Research are going. Group Chat “will be migrated as individual chats”, Podcasts and Deep Research are being retired.

This looks very like a necessary housekeeping exercise to tidy up the diverse range and huge scale of operations Copilot inevitably generates. Expanded volumes must inevitably require streamlining of apps.

Not everyone’s happy about this situation. Comments on The Verge article regarding the superapp  indicate the sort of “traditional” dissatisfaction with Microsoft’s UX overall you’d expect.

AI vs digital culture

The overall distrust seems more cultural, with AI in general seen as an intrusive and unwelcome factor. This may seem purely reactive and unrealistic, but there is a point to the dissatisfaction.

During the glory days of the internet 30 years ago, operations were much more basic, but it was fun, interesting, and actually pretty efficient. Getting rid of mountains of paper didn’t bother anyone at all. Digital systems created their own audit trails. Everyone adapted in their own niches and simply enjoyed the smoother ride.  

AI has created imponderables. That’s the real problem. Your friendly, folksy, fun-loving chatbot can also be clunky. The perception of “efficiency” doesn’t include scripted and sometimes lengthy responses to anything and everything. Particularly when a two-minute phone call is so much more productive.

This impasse will eventually have to clear itself up because it has no choice. The efficiency of AI is geared to comprehension, and users aren’t necessarily good at defining what they want.

The other cultural question is privacy. Microsoft lays out its rules for Copilot, but there’s a gigantic cultural issue here. Implied surveillance is as bad as actual surveillance in that you may or may not know your rights or where your information is going. Nobody believes that data brokerage doesn’t include unwanted snooping.

Microsoft has also unwittingly or otherwise exposed a reliable knee-jerk response to change. Change for the sake of change, making people do things differently or adding extra layers to basic jobs is simply annoying.

Change because the entire framework of AI data management is totally different is still annoying. Necessary it may be, but it doesn’t have to be alarming or stressful. A bit of time like six months to assess and to adapt would have helped.

Does the superapp make sense commercially?

Far less straightforward is the loss of established features. In purely commercial terms, what are the benefits?

The loss of Podcast is a case in point. Podcasts are often business operations, and Microsoft Podcast runs quite a few well-known podcasts. The sudden demise of Podcast will definitely affect some users negatively.

It’s also more than a bit odd that Microsoft seems to have temporarily written itself out of servicing this very popular global medium.

Losing Deep Research is also a particularly opaque option. Deep Research is/was a useful tool for a very wide range of professional users. As part of the Microsoft family of tools, it seems natural that people using Microsoft products would prefer an in-suite option.

Stranger still is the fact that the core functions of Deep Research are very much AI territory. Copilot is a natural built-in Microsoft resource, true, but did they simply not say “You can do all that stuff with Copilot anyway”?

AI inevitably requires systemic environmental reorganization

The much deeper part of the Microsoft move is far more existential than purely contemporary and local. AI is now at the toddler phase, roughly the two-year-old, precocious klutz stage.

Ignore the breathless raptures about the tech. The historical default evolution of technologies comes in predictable phases:

Disruption to current practices: Superseded practices have to go anyway. Some method in the madness would be much appreciated. Operational values for AI must be better and able to prove it in dollar terms.

Adaptation to functional needs: AI’s biggest marketing weakness so far is the almost total failure to recognize that users have specific needs. Agentic AI, in its clumsy clown suit of convenient customization, has barely addressed this issue. Adaptation happens as users create better fits for their tech. There’s no One Size to fit. This will take time and experience.

Expansion of technologies and capabilities: AI may well turn out to be a blessing in shoddy packaging and droning terminology. It’s obvious that AI can deliver value and deliver it fast. You could set up an entire business with shopfronts, distribution, affiliate programs, etc., tailored to your market in an hour or so. This requires practice, training, and above all support.

Evolution of AI roles and reach: This is barely the Cambrian Era for AI. The future ecologies of the AI universe will have to evolve in direct relation to its environmental realities. The future will include unknown environments and technologies. Maybe AI agents will need to evolve their own generational self-defence mechanisms for their roles. Maybe they’ll find new markets in which to graze. Their reach will be their ever-expanding frames of reference.

 Microsoft’s new move might just be the very first iteration of how AI ecologies will change.