Amid the AI frenzy of investment and hype of across Big Tech these past two years there has been one overarching yet unanswered question on the lips of most AI industry executives:
“How do we monetise this?”
That’s because the reality of the AI roll-out is that the “innovate and they will come” model that worked so well to drive cloud adoption hasn’t worked in quite the same way for AI. Not least because the market has multiple competing options that make enough headline news to make customers aware of their ability to choose.
Ultimately, we will probably see consolidation and simplification of the market, but right now public awareness of AI is high, customer choice is wide, and the ability to convert tokens in the datacentre to cash in the bank isn’t playing out quite as investors may have hoped.
There are some serious players in the market, including the big three hyperscalers, with all offering GPU-as-a-service type consumption models. Beyond that their services and approaches are quite differentiated.
AWS principally offers AI enablement: AI tooling for developers that fits their largely IaaS/PaaS model. This is powered by generally available LLM’s and their own model – Nova – built on the voice-activated Alexa services most of us already have in our homes.
Google has a broader range of directly-embedded web services with integration to their search engine and workspace services. They also have three distinct variants of in-house models: Gemini, Vertex and Gemini Enterprise.
Microsoft goes all in
As usual, however, it is Microsoft that has gone all-in. They have AI everywhere; not just in their cloud services but also desktop operating systems and productivity software.
Copilot is on every Windows PC in the world, every office suite, all M365 cloud services and is peppered through the Azure landscape.
It is rapidly becoming pervasive. You literally cannot escape it, and even when not actively using Copilot you may be relying on its outputs in ways you hadn’t realised, as the former Chief Constable of West Midlands Police recently found out to his cost.
Copilot may be software designed for entertainment purposes and not to be relied upon, as some reports of Microsoft caveats suggest, but it’s also a multi-billion dollar revenue stream for Microsoft – or at least that appears to be the plan.
Per-agent licensing
The recent suggestion by Microsoft VP for Copilot and M365, Rajesh Jha, might prove to be the most lucrative change to licensing practices since vendors adopted subscription-based models. Namely, to force users to buy licences for software sub-routines running inside the software they’ve already paid for.
Announcing plans to require user-type licences for AI agents to offset potential losses of revenue from staff downscaling due to AI, Jha has gifted Microsoft an idea to exploit the untapped licence value of AI. It’s one that should ring alarm bells for every enterprise locked into their product ecosystem.
This is not just software licensing based on compute or storage resources – which already leads to unexpected bill inflation for many customers – but charging Microsoft customers by the sub-routine they run inside the software.
As a company grows into AI and develops new streamlined ways of working, Microsoft’s revenues will actually increase. Every developer-created agent, every user AI request, will be captured, tracked and billed – not just on a consumption model, but also on a repeatable annual licence subscription, billed per agent.
Genius move?
For Microsoft it’s potentially a genius move. And the new Microsoft Agent 365, and Microsoft 365 E7: The Frontier Suite products – which go live on 1 May – are the means for them to do this.
This shouldn’t surprise us. The only thing secondary to Microsoft’s ability to create new products is their drive to gain revenue from them, which of course for a commercial entity like Microsoft is entirely to be expected.
What won’t be expected by their customers is the likely impact on their purse. The very organisations rushing to adopt Agentic AI in Microsoft Cloud and M365 are the same ones who will find their bills rising rapidly.
I’ve no doubt the new Frontier Suite products bring amazingly detailed insights into how your organisation uses AI, and controls over its use. The security and management reporting value of that capability is self-evident.
It’s also going to be a hugely necessary insight into how to manage your bills, and that’s required because Microsoft are changing the billing rules. They’ve found a way to tax innovation at source, and every single user of their AI services is going to pay for it, and will do so in ways we hadn’t even realised were possible.
Most of Microsoft’s users might still be waiting to enter the AI age, but software licensing is waiting for them to arrive.
