Microsoft IQ Explained: What It Changes for Work

Microsoft IQ is the context layer Microsoft launched at Build 2026 that grounds AI agents in both world knowledge and your company's own enterprise knowledge, so agents can act inside the Microsoft 365 tools your team already runs instead of being bolted on as a separate app.

That is the one-sentence version. If you build, buy, or operate back-office automation, the rest of this page exists to translate that sentence into something you can act on: what actually shipped, how it works in plain language, why it arrived now, who is behind it, the limits nobody should hand-wave past, and where we think it lands for small and mid-size businesses over the next few years.

The reason this page exists is simple. "Microsoft IQ" is a days-old term. It was announced this week, the search results are mostly press headlines and launch-keynote recaps, and almost nobody has written the calm, sourced, jargon-free explanation that an owner, an IT lead, or an operations manager actually wants. So that is what this is — the plain-English entity page, as of June 2026.

TL;DR

• Microsoft IQ went generally available across GitHub Copilot, Microsoft Foundry and Copilot Studio on announcement day. Microsoft's Build 2026 blog describes it as "a new context layer that grounds agents in both world knowledge and enterprise knowledge."

• It is the grounding layer underneath a wider Microsoft Agent Platform. Windows News describes that platform in three pieces — an Agent Runtime, an Orchestration Engine, and a Governance & Trust Layer — in its Build 2026 report.

• The Work IQ APIs that expose this layer to developers are generally available on June 16, according to Microsoft, whose Build 2026 blog calls them "programmatic access to this intelligence layer."

• The strategic point is reach, not novelty: Microsoft is aiming agent rollout at the 400 million Microsoft 365 seats it already sells, according to Windows News, whose coverage ties the bet to "billions of devices running Windows, and the Azure backbone."

• The honest limit: shipping a governance layer raises the ceiling of what an agent can be trusted to do inside your tenant; it does not by itself make any specific workflow safe to leave unattended.

What actually happened

At Build 2026 — Microsoft's two-day developer conference held June 2-3 at the Moscone Center in San Francisco — Microsoft moved AI agents from demo to default. Microsoft IQ went generally available on day one across GitHub Copilot, Microsoft Foundry and Copilot Studio, and Microsoft's keynote blog positions it as the context layer beneath the company's agent stack. The same conference "drew an estimated 15,000 in-person attendees and tens of thousands more online," a scale signal worth noting in the Windows News event recap.

Underneath the marketing, three things shipped together, and it helps to keep them straight:

• Microsoft IQ — the context/grounding layer. It is what lets an agent "know" both the public world and your company's documents, data, and rules.

• The Microsoft Agent Platform — the place you build, deploy, and run agents. According to Microsoft's blog, "you can build your agent in GitHub, deploy it to Microsoft Foundry and optimize it automatically with models best suited for the job."

• Agent 365 — the control plane for governing those agents once they exist. According to Microsoft, "Agent 365 for local agents extends Entra, Defender and Purview into a single control plane to observe, govern and secure agents across your estate," language drawn directly from the Build 2026 blog.

Build week was also a model launch. Microsoft unveiled seven homegrown AI models, including its first reasoning model, MAI-Thinking-1, according to Microsoft, whose Build 2026 blog describes it as "a mid-sized, 35 billion active parameter model with a 256K context window." The models matter because they are the engines the Agent Platform can route to — but for most businesses, the platform-and-governance story is the part that changes daily work.

What shipped at Build 2026

Sources for this table appear in the cited paragraphs above and below — the GA scope of Microsoft IQ, the June 16 Work IQ date, and the MAI-Thinking-1 description all come from Microsoft's Build 2026 blog.

How it works, in plain language

Think of an AI agent as a worker that can read instructions, look things up, and take a few steps on its own. For that worker to be useful inside a business, it needs three things: it needs to know the right context, it needs somewhere to run, and someone needs to be able to watch and stop it. Microsoft's announcement maps almost one-to-one onto those three needs.

Knowing. This is Microsoft IQ — "a new context layer that grounds agents in both world knowledge and enterprise knowledge," language taken from Microsoft's Build 2026 blog. In practice, "grounding" means the agent answers from your actual files, email threads, and records — not just whatever a general model happened to memorize. The developer hook for the work side of that context, the Work IQ APIs, becomes generally available on June 16, the date the same blog gives while describing the APIs as "programmatic access to this intelligence layer."

Running. This is the Microsoft Agent Platform. Windows News describes it in three pieces — an Agent Runtime, an Orchestration Engine that chains agents together, and a Governance & Trust Layer — per its Build 2026 report. The plain-English read: instead of standing up your own servers and glue code to host an agent, you point it at Microsoft's runtime and let the platform handle scaling and hand-offs.

Watching. This is Agent 365 plus the orchestration rules. Governance here is concrete: the orchestration layer can enforce business rules ranging from simple guardrails like "Never approve contracts above $500,000 without human sign-off" to more sophisticated meta-rules, an example drawn from the Windows News coverage. That dollar-threshold example is the whole story in miniature: an agent can do the work, but a human stays on the hook above a line you set.

The mechanism that ties the three together — and this is the part that actually changes back-office work — is that all of it is being delivered inside the Microsoft 365 stack most companies already run. You do not adopt a new vendor; you turn on configuration in tools you already pay for.

Why now — what constraint broke

New capabilities ship constantly. The interesting question for any "why now" is: what was blocking this, and did the block actually move?

The block was never raw model quality — it was trust and plumbing. Surveys from late 2025 showed that 78% of CIOs cited governance, security, and explainability as the top barriers to adoption, according to Windows News, a figure it uses in its Build 2026 analysis to explain why a governance-first launch landed now. Read plainly: leaders did not refuse agents because the agents were dumb. They refused because they could not see what an agent did, prove why it did it, or guarantee it would stop at the right line.

That is exactly the gap Agent 365 targets by extending Entra, Defender and Purview into a single control plane, according to Microsoft's Build 2026 blog — identity, threat protection, and data governance an IT team already operates, now pointed at agents. The constraint that moved is not "can a model do the task." It is "can IT govern the thing doing the task using controls it already trusts."

There is also a distribution reason it is now and not later. The agent capabilities are being aimed at the 400 million Microsoft 365 seats Microsoft already sells — the Windows News report frames the bet as making agentic AI "as mundane as email." When the distribution channel is the software a company already runs, rollout stops being a procurement project and becomes an IT setting.

Build 2026 by the numbers

Each figure is sourced in a cited paragraph elsewhere on this page; the attendance and seat figures come from Windows News's event report, and the model count from Microsoft's Build 2026 blog.

Availability at a glance

The general-availability dates and preview status above come from Microsoft's Build 2026 blog and Windows News's Build 2026 report, which dates the "Polaris" Windows runtime to later in 2026.

Who shipped it, and the supporting cast

Microsoft shipped Microsoft IQ, the Agent Platform, and Agent 365 as a single, integrated push — the through-line of its own Build 2026 blog, which ties context, build/deploy, and governance into one story. The platform is explicitly model-flexible: in Microsoft's words you "deploy it to Microsoft Foundry and optimize it automatically with models best suited for the job," meaning the agent layer is not welded to a single model.

That flexibility is why the same week's model news matters as supporting cast rather than headline. Microsoft's MAI-Thinking-1 is a mid-sized, 35 billion active parameter model with a 256K context window, according to Microsoft's Build 2026 blog. Separately, according to Microsoft AI, a MAI model tuned for Excel matches GPT-5.4 while being up to 10x more efficient, in its model launch post. For a business, the takeaway is not the parameter count — it is that the platform can route a cheap, fast model at a routine task and a stronger one at a hard task, without you rewiring anything.

The honest limits

A platform launch is not a finished deployment. Three limits deserve plain statement, kept strictly to what the sources support.

First, "generally available" is uneven. Microsoft IQ itself went GA, but the deep Windows-level integration is future-dated: the agent runtime support lands in the next major Windows update, codenamed "Polaris," slated for release later in 2026, per that Windows News report. The same coverage notes that early adopters can expect previews of those Windows capabilities "via the Windows Insider Program in the coming months" — and the in-house reasoning model MAI-Thinking-1 shipped the same week as the first of Microsoft's seven new in-house models, per Microsoft's Build 2026 blog — newly released, not yet broadly proven. Announced is not parity.

Second, pricing was not pinned down. Windows News describes only a hinted "consumption-based model with reserved capacity options" in its coverage — meaning total cost at your usage level is genuinely unknown until you measure it, and "it's in M365" does not mean "it's free."

Third — and this is the one that bites compliance-heavy teams — a governance layer is not the same as a governance outcome. The fact that an agent can be told to stop at a $500,000 contract threshold, per the example in Windows News's report, does not mean your specific rules are written, tested, and enforced. Somebody still has to author the policy, map the data the agent is allowed to touch, and verify the stop actually fires. The platform makes that work possible and centralized; it does not make it done.

Signal vs Speculation

Everything above this line is sourced fact. Everything in this section is our forecast, clearly labeled as such — read it as analyst opinion, not reporting.

What is demonstrated fact (sourced). Microsoft IQ is generally available across GitHub Copilot, Foundry, and Copilot Studio; Work IQ APIs reach GA on June 16; Agent 365 extends Entra, Defender, and Purview into one agent control plane; and the whole push is aimed at the 400-million-seat Microsoft 365 base. All four are from Microsoft's and Windows News's Build 2026 reporting cited above.

Our read — the next-few-years picture for small and mid-size businesses. If the "it's already in your M365 stack" framing holds, the most likely outcome is that agent deployment stops being a custom software project and becomes a configuration task owned by whoever already administers your tenant. That is a meaningful shift for a 20-to-200-person firm that could never staff an "AI integration" build but can absolutely toggle settings in tools it already runs.

Our read — the catch. We expect the binding constraint to migrate from capability to governance hygiene. The 78% governance-barrier figure is the tell: the platform answers the "can IT control it" objection, which means the next bottleneck is the unglamorous work of writing the rules, mapping which data each agent may touch, and verifying the human-sign-off thresholds actually trigger. Firms that already keep their documents and approvals in clean, well-routed workflows will adopt fastest; firms with messy data will find the agent only as trustworthy as the context it is grounded in.

Our read — the practical move for now. We would not rip anything out to chase a days-old platform. We would inventory which back-office processes already run through a documented, automated workflow, because those are the processes where a Microsoft IQ-grounded agent is a model swap rather than a rebuild. Teams already routing documents through US Tech Automations workflows are positioned to add an agent step at the point of extraction or routing without re-architecting the pipeline — the connector changes, the process does not.

What it means for your industry

This page is the hub. The implications differ sharply by the kind of business you run, and we have written dedicated explainers for the three verticals where the governance-and-context story changes daily work the most:

• For owner-operated and small teams, see what Microsoft IQ means for small businesses — where the "configuration, not a build" shift matters most.

• For practices handling reconciliations, filings, and client books, see what Microsoft IQ means for accounting firms — where Agent Mode in Excel and human-sign-off thresholds land directly.

• For document-and-deadline-driven legal work, see what Microsoft IQ means for law firms — where data-governance controls and matter confidentiality dominate the decision.

A useful first internal step before reading any of those: list the processes you would want an agent to touch, and mark which ones already have a clean intake and a defined approval step. Teams that map documents into US Tech Automations workflows can use that same inventory to decide where an agent plugs into an existing extraction or routing step rather than a new one.

Key Takeaways

• Microsoft IQ is a context/grounding layer, generally available across GitHub Copilot, Foundry, and Copilot Studio — the part that lets an agent answer from your company's own knowledge, per Microsoft's Build 2026 blog.

• It sits under a broader Microsoft Agent Platform (build, run, govern) and Agent 365 (the control plane), so the news is really about deployment and governance, not a single product.

• The strategic bet is distribution: aiming agents at the 400 million Microsoft 365 seats companies already run, per Windows News's report.

• The constraint that moved was trust, not capability — a 78% CIO governance barrier, answered by extending Entra, Defender, and Purview to agents.

• The honest limits: previews are not parity, pricing is consumption-based and unpinned, and a governance layer still needs you to write and test the rules.

Frequently asked questions

What is Microsoft IQ in one sentence?

Microsoft IQ is the context layer that grounds AI agents in both world knowledge and your company's own enterprise knowledge. According to Microsoft, it went generally available across GitHub Copilot, Microsoft Foundry and Copilot Studio at Build 2026, language from its keynote blog. In practice it is what lets an agent answer from your files and records rather than only a general model's memory.

How is Microsoft IQ different from Copilot?

Copilot is the assistant you talk to; Microsoft IQ is the grounding layer underneath that tells agents what your company actually knows. Microsoft describes IQ as available across GitHub Copilot, Foundry, and Copilot Studio in its Build 2026 blog — so think of IQ as the context plumbing and Copilot as one of the front doors that uses it.

When can developers actually use it?

Developer access arrived fast. The Work IQ APIs are generally available on June 16, the date Microsoft's Build 2026 blog gives while calling them "programmatic access to this intelligence layer." The deeper Windows-level agent runtime, by contrast, is tied to a later 2026 Windows update.

Does Microsoft IQ change how my back-office work gets done?

Potentially, yes — because it ships inside the Microsoft 365 stack most companies already run. The reach is large: Microsoft is aiming this at 400 million Microsoft 365 seats, a figure from the Windows News coverage. That turns agent rollout from a custom project into a configuration task for whoever administers your tenant.

Is it safe to let an agent run unattended?

Only within limits you set and verify. The platform supports rules like a $500,000 human-sign-off threshold on contracts, an example given in the Windows News Build 2026 report — but a governance layer is not a finished policy. Someone still has to author the rule, scope the data, and confirm the stop fires.

Why did Microsoft launch this now?

Because the blocker was governance, not model quality. 78% of CIOs cited governance, security, and explainability as the top barriers to adoption in late-2025 surveys, a figure in the Windows News analysis. Pointing existing controls — Entra, Defender, Purview — at agents is Microsoft's direct answer to that objection.

Do I need to switch models or vendors to use it?

No. The Agent Platform is model-flexible: Microsoft says you "optimize it automatically with models best suited for the job" in its Build 2026 blog, and the same week it unveiled seven in-house models of its own, per Microsoft AI's model launch post. For most teams it is a configuration change, not a rip-and-replace.

Where to go from here

If you operate document-heavy or approval-heavy processes, the practical question is not "should we adopt Microsoft IQ" — it is "which of our workflows are clean enough that an agent could plug into them as a step rather than a rebuild." That is an inventory you can do today, before any platform decision. To see how a documented, governed automation pipeline is structured so an agent can drop into an existing extraction or routing step, walk through our agentic workflows overview, then read the industry explainer that matches your business above.