Have you tried using AI for that?

David Bowman:

You propose an investment in employee communications infrastructure. An intranet, a knowledge management function, a content strategy. The response comes back from above: have you tried using AI for that?

It’s not arriving in a vacuum. Every budget conversation right now is a three-way collision between cost reduction, operational efficiency, and AI investment.

Those three things have fused into a single programme in most organisations, and everything that’s not one of them is, by implication, the old way of doing things.

Employee communications infrastructure has the wrong vocabulary for this moment. Intranet sounds like 2005. Knowledge management sounds like a consulting engagement that never delivered. Content strategy sounds like something marketing should own. The AI question is not really a question about capability. It’s a signal that the category has already lost the argument before you made the case.

The problem is that the category losing the argument is precisely the category that determines whether the AI investment returns anything above meeting summaries and marginally better slide decks.

Microsoft’s 2026 Work Trend Index surveyed 20,000 workers across 10 markets and found that only 19% of organisations sit in what they call the Frontier zone, where individual AI capability and organisational readiness are both high and reinforcing each other. The constraint holding the other 81% back isn’t the technology. Organisational factors, culture, documented processes, how knowledge is captured and structured, account for more than twice the reported AI impact of individual effort alone. People are ready. The systems around them aren’t. 

The question “have you tried using AI for that” assumes the bottleneck is effort. In most of the organisations I’m talking to, the bottleneck is substrate.

Copilot doesn’t editorially improve the content it reasons over.

It doesn’t fill gaps where no one captured anything. It doesn’t distinguish between the policy document your comms team spent three weeks getting right and the draft someone abandoned in a shared folder in 2021. It reasons over whatever is there and returns an answer with confidence proportional to the signal it finds. Where the signal is thin, the answer is thin. Where no one has been responsible for what goes in, the answer reflects that.

Fresh Intranet’s own research, published in May 2026, found that 28% of employees now use AI-generated summaries as their primary discovery channel for company news, ahead of the intranet, ahead of email from leadership, ahead of Teams notifications.

Employees aren’t just using AI to compress content they have already decided to read. A significant number are using it to decide what is worth reading in the first place. That means the quality of what AI returns is no longer a search problem. It’s an editorial problem, and most organisations have no one accountable for it.

The same research found that 98.6% of internal communications professionals are aware that employees are using AI to interpret their content. Only 33% have any formal governance in place for monitoring it. The gap between awareness and action isn’t laziness. It’s that the function responsible for closing It’s the one being defunded or bypassed by the organisations asking whether AI could do the job instead.

This is the compounding problem. The organisations most likely to ask “have you tried using AI for that” are also the organisations least likely to have built the conditions that make AI return something worth having. No intranet means no defined boundary between intentional organisational content and everything else in the tenant. No comms function means no one whose job it’s to turn business objectives into something an employee, or an AI, can find and understand. No content strategy means Copilot is reasoning over an undifferentiated mass of files, emails, and abandoned drafts, returning answers that reflect the state of what was never governed.

The organisations getting genuine value from Copilot, in the synthesis and reasoning use cases that go beyond meeting summaries, are the ones where someone has been making decisions about what gets captured, in what form, for whom, and where. Those decisions look like communications strategy. They look like intranet governance. They look like a knowledge manager who is listening for the questions employees are actually asking and building content that answers them. They are, in aggregate, the content infrastructure layer that determines whether an AI investment compounds or stalls.

A well-governed intranet inside your existing Microsoft 365 environment isn’t the whole answer to this, but it’s a meaningful part of it. A defined scope, content rules, and clear ownership does two things simultaneously. It gives employees a legible, well-structured place to find what they need. And it gives tools like Copilot a coherent, bounded, intentional signal to reason over, distinct from the noise of everything else in the tenant. The same investment, two payoffs, and most organisations are only pricing in one of them.

The internal communicator or knowledge manager doing this work isn’t managing a publishing channel. They’re curating the organisation’s AI surface, deciding what the organisation wants Copilot to know, in what shape, with what structure. Microsoft’s Work Trend Index calls the output of this work Owned Intelligence: institutional know-how that compounds over time, is unique to the firm, and hard to replicate. The organisations building it are treating content governance and employee communications not as a cost line but as AI readiness infrastructure. The organisations not building it are replacing that process with a question.

The answer in most cases is that you try. The readout will tell you exactly what you have built. What you have not built will not show up at all.

+++

David Bowman is Product Director for Fresh Intranet, which conducted the Employee Attention Recession research.