[ethbr1]

Your experience essentially mirrors the state of the product category.

Incapable of serving as a general purpose tool, outside of specialized use cases (NLP + action, summarization).

There are more generalized things in the works that would improve that (from Microsoft and others), but afaik nothing has been released to consumers.

The issue with Microsoft's Power nee Copilot strategy has always been unification -- it may share one branding, but you're really talking about different teams, writing different integrations, with different amounts of functionality.

It's hard to meet customer expectations when "magic thing that works with PowerPoint Copilot doesn't exist in Excel Copilot."

And right now, GenAI is dependent on access to integration scaffolding with the specific service+activity -- ergo, much more IFTTT and less general purpose magic box.

[gowings97]

Correct.

People say they want an AI but really they want something to automate and complete their daily workflow/tasks.

How that is accomplished is through the integrations, as you've described. But I am at a loss as to how these integrations have anything to do with NLP.

[ethbr1]

What most (non-developer) users want is an ability to describe what they do (in non-code, natural language terms) and have a moderately intelligent system iteratively build/modify a solution for them.

From that perspective, the product is essentially constructing a mapping between NLP and available pre-built code integration points.

[gowings97]

So in some ways it tries to shoehorn what's available via APIs and attempts to shoehorn it based on the NLP parsed prompt...I guess?