Microsoft's Work IQ API reached general availability on June 16. Two weeks in, most developers I've talked to haven't absorbed what it actually is.
Work IQ isn't a productivity dashboard or a reporting tool. It's an intelligence layer — a continuously updating semantic model of how your organization works, built from every email, calendar invite, meeting transcript, chat message, and file in Microsoft 365. That model is then served to enterprise AI agents so they can act with real organizational context: scheduling across a team's actual focus patterns, drafting communications that match someone's real style, surfacing the right files based on who's working with whom on what.
Here's the part worth sitting with: the model is built entirely from your work. But the entity that accesses it is your employer.
What Gets Processed
Work IQ continuously processes email, calendar, meetings, chats, files, "people signals," and collaboration patterns across the Microsoft 365 tenant. The output isn't raw files — it's an inferred organizational graph: topic clusters, collaboration networks, who-talks-to-whom patterns, and work rhythms extracted from the aggregate activity.
Microsoft frames this as aggregated, anonymized organizational intelligence rather than individual monitoring. That framing matters, and it's probably accurate at the architecture level. But the governance of what gets surfaced and to whom is decided at the enterprise tenant level — by the organization paying for M365 licenses, not by the people whose activity generated the underlying data.
A post from the "Intranet from the Trenches" newsletter, published days after GA, captured the governance gap cleanly: "Microsoft is moving deeper into a model where the platform doesn't just store and route work; it interprets it. That has huge productivity potential, but it also means governance must evolve from managing access to managing intelligence."
That's the right framing. Managing access — who can read which files — is a solved problem in enterprise software. Managing intelligence — what can be inferred from patterns across those files, and who can query those inferences — is a different problem, and most organizations haven't caught up to it.
The Data Asymmetry
You generate the raw material. The organization accesses the refined product.
The emails you write. The calendar you fill. The meetings you attend. Every artifact you produce inside M365 becomes an input to Work IQ's continuous processing. The semantic understanding derived from those inputs — the patterns, the inferences, the organizational knowledge graph — is what enterprise agents get to work with.
You can use those same agents. You don't get to see how the model thinks about your own patterns.
This isn't new. It's the default architecture of enterprise software: the platform accumulates context about how you work, that context makes the platform more useful, and the value flows to whoever controls the platform. What Work IQ changes is that the intelligence layer is now explicit. Named. Accessible via API. Priced on a consumption model with published documentation.
Making it explicit makes the asymmetry visible in a way that was easy to ignore when it was implicit.
Where This Isn't Surveillance
The employee monitoring framing is wrong here, and using it misses what's actually interesting.
Classic monitoring — ActivTrak screenshots, keystroke loggers, Teramind's "productivity scores" — is about real-time observation. The premise is continuous visibility into what you're doing at each moment. It's coarse, mostly measures compliance, and generates the kind of pressure that corrodes trust without improving output.
Work IQ is doing something more sophisticated: building an interpretive model from signals you've already emitted. The email was sent. The meeting happened. The calendar was filled. Work IQ reasons about patterns across those artifacts after the fact — more like a statistical model than a window into your activity.
That's harder to object to on the same grounds as a screenshot tool. It's also harder to reason about. You can enumerate the inputs. The inferences are opaque.
The researcher in that governance piece put it well: "Workers may like AI that remembers their projects; they may be less comfortable with AI that appears to remember patterns they never explicitly gave it. The line between helpful continuity and unsettling surveillance is not a technical boundary. It is a trust boundary."
That trust question isn't going to be resolved by better documentation of what Work IQ aggregates. It'll be resolved — or not — by whether employees actually understand what's been built from their activity and have any meaningful visibility into it.
What You Actually Get
Here's the part that's not discussed in Microsoft's launch coverage: what does the employee get from Work IQ?
Indirectly, they get smarter enterprise agents. A scheduling agent that understands focus blocks. A drafting agent that understands communication tone. An assistant that surfaces context without needing to be told what's relevant. These are real benefits, and they're built on the intelligence layer that Work IQ provides.
But the employee doesn't get the model itself. If you want to understand how you actually spend your time inside M365 — what your collaboration patterns look like, how your meeting load has changed over quarters, whether your deep work blocks are actually protecting your best output hours — you can't query Work IQ for that. The enterprise can query it about their organization. You can't query it about yourself.
That gap is older than Work IQ. But Work IQ makes it concrete, because now there's a named product that contains a semantic model of your work patterns that you have no access to.
The Individual Version of the Same Problem
The intelligence that Work IQ sells to organizations — "here's how your people actually work" — is exactly the intelligence that should be available to individuals about themselves. You generated the raw material for both use cases. The enterprise version is increasingly well-funded, productized, and technically sophisticated. The individual version is mostly missing.
This is the gap that personal analytics tools are trying to close. Not a corporate intelligence layer that your employer queries, but a model of your own work that you own and can interrogate. Your app usage, your focus patterns, your coding session data, your energy levels, what you actually spent your time on this week — aggregated into something that surfaces what the enterprise layer can't, because it isn't accountable to you.
At xeve, this is the exact gap. The data you emit every day as a developer — app switches, coding sessions, focus blocks, biometric signals from your wearable, activity across your tools — belongs to you and should generate insights that flow back to you. Not as a byproduct of someone else's enterprise intelligence system, but as a model of your own work that you control.
Work IQ GA is a useful prompt for asking who should have access to the intelligence derived from how you spend your days. Microsoft's answer is the enterprise that licenses M365. That seems like the wrong default — and it's not one you have to accept.