Microsoft has spent years building toward an enterprise AI moment it believed was coming.

The infrastructure it bet on was not a model. It was Office: a user base built up over decades, their documents, their email histories, and the workflows running through all of it.

Copilot is the product now sitting on top of that base.

Bank of America analyst Tal Liani hosted an investor session with Jacob Andreou, executive vice president of Copilot, on Nov. 20. The setup was different from a standard earnings call. Liani brought Andreou directly in front of investors to walk through where the enterprise AI strategy stands, what Microsoft is still building, and what it still needs to prove.

Liani reiterated a Buy rating on Microsoft (MSFT) with a price target of $630, against a Nov. 20 closing price of $498.30. In a note shared with TheStreet, he laid out the case Andreou made and what it means for investors watching the stock.

Why enterprise AI still depends on context more than capability

Andreou’s starting point was a problem Microsoft has been pitching for a while. Frontier models score well on benchmarks. Put them inside a company with no access to its files, client histories, or the conversations employees have been having for years, and the results start to disappoint.

Also read: Microsoft CEO Nadella faces uncomfortable reality after outage

Andreou called Work IQ Copilot’s intelligence layer. He described how it tracks the way work actually moves inside an organization: an email thread about a client deal, the Teams call that followed it, the files that came out of that call, all linked together. When a question touches any part of that chain, the system is built to pull it up.

The outside providers have capable models. Getting those models to understand a specific company’s internal work, its email threads, project history, and files going back years, requires access Microsoft already has.

Andreou made that case at the session. The note put it at the center of Copilot’s competitive position.

3 tiers and what they tell investors about how revenue scales

Andreou described three types of users, each with a different revenue profile. Starters already have basic Copilot features inside their Microsoft 365 plans. They’re using the product, just not paying for it separately. Getting them to eventually buy a seat is the first conversion Microsoft is after.

Subscribers are paying for a Copilot seat. The more engaged users go further, into modules like Autopilot for background task execution, Code for building custom apps and dashboards, and Cowork for working across multiple files at once. Those modules bill on consumption, meaning Microsoft charges based on what the system actually processes.

Both moves matter for where the revenue goes. Getting Starters to pay for a Copilot seat is one path. Getting paying subscribers to actually use the consumption modules is a different bet entirely.

The note put both as necessary for the paid seat growth Liani’s Buy rating assumes.

Liani’s $630 price target is built on 28 times his calendar year 2027 earnings estimate.

Massimo Giachetti / Getty Images

What keeps enterprise customers inside Microsoft’s ecosystem

Andreou’s third line of argument was harder to put a number on. Organizations that bring in Copilot don’t have to reorganize how they work. Word stays Word. Teams stays Teams. Someone who hasn’t adopted Copilot at all still opens the same files from the same folders as everyone else around them.

For companies trying to displace Microsoft in enterprise software, that’s key. Pulling an organization off Office means moving document storage, file sharing, security controls, and compliance workflows at the same time, through legal, IT, and security teams simultaneously.

The note put that barrier among the three main structural advantages Copilot brings to competition with outside AI providers.

More Microsoft:

Copilot Workspaces, launched at Ignite, runs on the same premise. A team gets a shared space to run work through Copilot. The person in the organization who hasn’t adopted the tool still accesses the same documents under the same security controls.

Andreou said that was intentional. The product is built to function without a full company rollout behind it.

Microsoft’s pricing structure and what Bank of America is betting on

Pricing came up near the end of the session. Andreou laid out a flat monthly rate for everyday Copilot use and a separate consumption meter for heavier tasks.

The note said Microsoft structured it that way so companies can roll the product out broadly without committing upfront to a budget that depends on predicting how much people will use it.

Auto mode runs inside the consumption tier and handles model selection automatically. It routes each step in a task to whatever fits: lower-cost models for simpler work, more capable ones when the task gets heavier.

Andreou also raised a broader issue during the session: AI products still require users to know what to ask. The Today feature, rolled out at Ignite, pushes relevant work to users before they ask for it.

Liani’s $630 price target is built on 28 times his calendar year 2027 earnings estimate, a premium to the peer group range of 18 to 25 times. He wrote in the note that Microsoft’s revenue growth and margin profile justify paying above that range.

At $498.30, the target implies roughly 26% upside. He also listed three downside risks: gross margin pressure as Azure scales, faster AI innovation from outside Microsoft, and a slowdown in enterprise application spending.

The last of those is the one most enterprise AI bulls are not modeling.

Related: Microsoft’s Oracle move raises new question for investors