Microsoft Copilot – Smooth AI Operator?

by Heidi Elmore

Microsoft and Google productivity platforms both “entered the chat” this week, and just like that, we are all trying to understand how this new wave of AI will impact some of our most critical applications. While I am admittedly still trying to untangle all the AI news dropped in our lap, one word stood out to me more than any other: Copilot.

I have been struck by the elegance of the Microsoft Copilot naming since it first dropped with GitHub. They laid out the product purpose and aspiration in one fell swoop without esoteric terms, buzzwords, or jargon. In a very un-Microsoft move, Copilot appears to be a stroke of product naming genius. 

But Copilot has now been launched across GitHub, Dynamics, Power Apps, and Microsoft 365. So the real question is – will product execution fulfill this inspired and unified naming promise?

The naming analysis is worth diving into first. Every company is different in its naming principles. For companies with large multi-product portfolios, the best product names tell you what it does or who it’s for. Amazon Supply Chain. Microsoft Teams. Twilio Customer Data Platform. If you are a small startup whose product is your company brand, obscure translations of “work” in Swahili are probably ok. Memorable even. (It’s kazi, in case you didn’t know. Just ask ChatGPT.)

The Copilot term is self-explanatory. You, as the pilot, are in control. AI is merely there to take the wheel when you need some help. Technology companies are also abandoning the personification of AI. Instead of naming tools Alexa, Cortana, or Lex (all female names, by the way), Microsoft is highlighting the function. 

What I also like about the Copilot term is that it is aspirational. Perhaps it’s because I come from a family of pilots, but it’s a different concept than “digital assistants.” In real life, assistants work for you, but they also work independently of you. Digital assistants never really fulfilled that promise. But you can (potentially?) train a Copilot.

Copilot also addresses some of our fears of AI. We want AI to help us, but we don’t want Skynet. AI gone wild is deeply ingrained in our shared consciousness as a doomsday for humanity. Copilot assures us that we are the boss. Copilot is our friend. Our Riker. Our #1.

Now, I’ve repeatedly praised Copilot as a product name. But in reality, a Copilot is a varied set of AI features embedded across the Microsoft portfolio. And this is where things could get a bit hairy for Microsoft. Having (proudly) worked there twice, I’m unfortunately very aware of how siloed product teams can be in an organization of over 200,000. 

As described by their public documentation and various blogs, Copilot can mean many different capabilities depending on the specific area of Microsoft technology. But by all accounts, it generally seems to offer the automation of asset creation through a conversational prompt. 

Copilot could be a chat call to the Microsoft Graph scrubbed by the Large Language Model (LLM) GPT in Microsoft 365. The organizational context here is undoubtedly the most compelling. Rather than just a text-based response, you could receive images, documents, or formulas into a new asset based on the content in your existing tenant. It seems like search, Delve, and a creative tool combined and on steroids. 

Copilot with Power Apps promises to automate the creation of no-code applications and virtual agents, which are generally associated with heavy lifting, even for citizen developers. Within virtual agents, Copilot can help automate the creation of  “topics,” typically associated with a specific use case through the conversational prompt. (For example – an insurance quote.) Those same virtual agents can be plugged right into Teams as well.

Copilot for Github supports the creation of code in your editor of choice based on conversational prompts. It can also automate testing. I’m no software engineer, but starting with pseudocode to develop code is not new. This just seems to supercharge the process, and potentially narrow the chasm between engineers and citizen developers. 

The conversational experience for Dynamics with Copilot is the most familiar, as many tools in the market already have similar capabilities. Salespeople can ask queries of their accounts in conversational prompts, informed by both account data and assets like meeting recordings. We all collectively shudder at the coached email authoring, as most of us have undoubtedly received a GPT-driven email. 

While the consistency of Copilot across products is the automation of asset creation, all the instances of Copilot appear to be implemented inconsistently. You can call the Copilot plugin from text editors in Github. Still, all the other mockups across products, menus, iconography, or even the chat prompt boxes seem unalike. That’s a bit concerning. A beautifully branded concept like Copilot needs some visual congruency.

In addition to the inconsistency of Copilot features in the UI are the monetization strategies across products. GitHub is a per-person per-month strategy with both Individual and Business tiers. We don’t yet know the pricing for Power Apps, Microsoft 365, or Dynamics. One would assume it will be included to elevate enterprise value in the suites. But then all we have to do is look at Teams Premium to second-guess ourselves. 

Microsoft Copilot is an inspiring concept and the best cohesive story of any major SaaS provider yet. There are threads of uniformity throughout all these features and capabilities across the Microsoft portfolio. But there is a real threat of capabilities tossed under the Copilot naming without much thought. Unless there is a centralized, cohesive set of principles around Copilot (dare I say – product principles), we may start to see the value of this shiny new brand degrade.