By Errin O'Connor — Founder & Chief AI Architect, EPC Group · Microsoft Solutions Partner across all six designations · G2 Leader in BI consulting
What is Microsoft Ignite, and why does it matter to an executive?
Ignite is Microsoft's annual event for announcements, technical sessions and labs. For anyone running Microsoft 365, Azure, Power BI or Fabric, it sets the product direction for roughly the next six months.
The risk for a leadership team is not missing an announcement. It is mistaking the announcement for the plan. A new control plane does not create an inventory of agents. A new billing dashboard does not set a limit. A new data feature does not certify a number. Those are decisions people make, and most of them were already due before the event.
How should an executive read an announcement?
Four questions, in this order.
- Is it generally available, in preview, or on a roadmap? Only the first is something to plan a rollout around. Fabric IQ ontology, for example, is still a preview item, and Microsoft retires the old ontology experience on January 31, 2027. [3]
- Is it on by default? Microsoft has been shipping metered services with defaults. In Copilot Cowork, a spending policy both funds and grants access, and "Auto-apply new services" is on by default for spending policies. [4][5]
- What does it cost, and on which meter? A seat price, Copilot Credits, or Fabric capacity. Three different budgets.
- Does it reach your cloud? Features arrive in the government clouds on their own schedule. See the seven routes into GCC and GCC High.
The Seven-Decision Filter
The Seven-Decision Filter is EPC Group's way to sort Ignite news: every announcement either changes one of these decisions or it does not.
1. Who owns AI in the company?
One named executive, with authority over budget and a seat at the risk table. Not a committee. For companies that cannot justify a full-time role, EPC Group provides a virtual Chief AI Officer.
Listen for: anything that changes who administers Copilot and agents. Decide regardless: the name.
2. Which agents exist, and who owns each one?
In the same CIO survey, 84% said employees are creating AI agents and applications faster than IT can govern them. [2] An agent with no owner has no one to answer for it.
Listen for: agent registry, lifecycle and identity features. Decide regardless: one register, one owner per agent, and what happens to agents nobody claims. See AI agent fleet management and Microsoft Agent 365 consulting.
3. Who can spend metered Copilot, and how much?
Copilot now has a variable tail. In Copilot Cowork, any user in the scope of a spending policy that selects Cowork has access, however small the limit. [4]
Listen for: new metered services and changes to cost management. Decide regardless: the security group, the per-user limit and the monthly review. See Copilot Cowork: access, cost and controls and the usage-based billing playbook.
4. Which numbers is AI allowed to answer from?
Copilot answers business questions from Power BI semantic models. A model that has not been certified produces confident wrong answers in front of the people least likely to doubt them.
Listen for: Fabric IQ, ontology and data agent changes. Decide regardless: which semantic models are certified, by whom, and by what date. See Fabric IQ ontology vs. semantic model and semantic model certification.
5. Who may publish what an agent builds?
Agents now build reports and data apps. Fabric Apps (preview) give each app its own hosting, database and sign-in.
Listen for: app creation in Power BI Desktop and Fabric Apps availability. Decide regardless: the publishing lanes and who signs off. See apps in Power BI and Fabric Apps governance.
6. What evidence can the company produce in 60 days?
Texas's AI law gives a company 60 days to cure after notice from the Attorney General. Cyber insurers and boards ask for the same record: an inventory, a purpose for each system, and a log.
Listen for: audit, compliance and security-for-AI features. Decide regardless: who assembles the evidence pack and where it lives. See TRAIGA compliance for Copilot and AI agents and AI insurance readiness.
7. What will be measured by the end of the quarter?
76% of CIOs in the survey said their organization lacks mature, highly capable ROI measurement across all or most AI initiatives. [2] Budgets for 2027 are being set against whatever can be shown now.
Listen for: usage and value reporting. Decide regardless: the baseline, the three measures the CFO will accept, and the date they are reported. See the AI ROI assessment and the CFO AI governance conversation.
Which of the seven can a tenant score today?
All of them, against EPC Group's 8-surface Tenant AI Readiness Standard, TAR-8.
| Decision | TAR-8 surface |
|---|---|
| 1. Who owns AI | 8. Ownership and evidence |
| 2. Agents and owners | 1. Identity · 6. Business applications |
| 3. Metered Copilot | 8. Ownership and evidence |
| 4. Numbers AI answers from | 5. Reports and dashboards |
| 5. What agents build | 5. Reports and dashboards · 6. Business applications |
| 6. Evidence in 60 days | 8. Ownership and evidence |
| 7. What is measured | 8. Ownership and evidence |
TAR-8 is EPC Group's published standard. It is not a Microsoft program or a certification.
What is already decided for you, or about to be?
Several defaults changed this year without a keynote. Each one is a decision someone made on a company's behalf unless the company made its own.
- Metered Copilot is arriving with a default limit. Microsoft announced that usage-based billing will be the default for new Microsoft 365 Copilot Business licenses bought through Cloud Solution Providers from January to March 2027, with a default limit of 4,000 Copilot Credits per user per month that administrators can change. [6] Detail: the usage-based billing playbook.
- A spending policy became an access control. In Copilot Cowork, scope decides who has access. [4]
- New metered services can attach themselves. "Auto-apply new services" is on by default for spending policies. [5]
- Copilot started answering from semantic models. Whether or not those models were certified.
- Agents started building reports and apps. Under whatever publishing rules already existed, which in most tenants is none.
None of these is a criticism of Microsoft. A platform has to ship with defaults. The point is that a default is a decision, and the company can make a better one for itself.
Who should watch what?
| Role | Watch for | Bring back |
|---|---|---|
| CIO | Agent registry, identity and lifecycle | A yes or no on whether the agent register changes |
| CFO | Anything metered, any licensing change | The meter, the default and the limit |
| CISO | Security for AI, audit and data protection | Which control is available now and which is preview |
| Chief data officer | Fabric IQ, ontology, data agents, apps | What changes for semantic model certification |
| General counsel | Audit, retention, compliance templates | What new evidence the platform can now produce |
One page from each, in the format the filter uses: the announcement, its status, the decision it changes.
Five questions for your Microsoft account team this week
- Which of this week's announcements are generally available today in the cloud this tenant runs in?
- Which are on by default, and where is the setting?
- Which bill on Copilot Credits, which on Fabric capacity, and which are in the seat price?
- What changes for tenants that already have a spending policy?
- What is the retirement date for anything being replaced?
What to do in the 30 days after Ignite
- Week 1. Sort the announcements with the filter. Write one line per announcement: which decision it changes, and whether it is available, in preview or on a roadmap.
- Week 2. Close decisions 1, 2 and 3. They need no new product.
- Week 3. Score the tenant on the surfaces behind decisions 4 and 5.
- Week 4. Rehearse decision 6 and report decision 7 to the CFO.
Where EPC Group fits
EPC Group has delivered Microsoft consulting since 1997 and holds all six Microsoft Solutions Partner designations, with more than 60 Copilot rollouts and more than 240 readiness assessments since August 2025.
The data behind this (sources and verification)
- [1] Microsoft Ignite event page
- [2] Dataiku, Global AI Confessions Report: CIO Edition 2026, press release (September 24, 2026), Harris Poll survey of 685 CIOs, July 9–29, 2026
- [3] Microsoft Learn, What is ontology (preview)?
- [4] Microsoft Learn, Determine access controls for Copilot Cowork
- [5] Microsoft Learn, Usage-based billing and cost management for Copilot Credits
- [6] Microsoft Partner Center announcements, November 2026
- [7] GovInfoSecurity, CIOs are accountable for AI agents they can't fully control (secondary report of the Dataiku survey)
- [8] Microsoft Learn, What is Fabric Apps (preview)?
- [9] Microsoft Learn, Fabric IQ in Microsoft 365 Copilot
Third-party figures above are attributed to their named sources as of the Last verified date. EPC Group audit figures are directional findings from client engagements.
Frequently asked questions
When is Microsoft Ignite 2026?
November 17 to 20, 2026, in San Francisco, according to Microsoft's event page. EPC Group's brief lists the decisions to make, not the announcements.
What will Microsoft announce at Ignite 2026?
This brief does not predict or list announcements. Microsoft publishes them during the event, and EPC Group recommends judging each against the Seven-Decision Filter.
What should a CIO do before Ignite?
Make the seven decisions that no announcement will make: ownership, agents, metered spend, certified numbers, publishing, evidence and measurement. EPC Group lists them in the Seven-Decision Filter.
How do I tell whether an Ignite announcement matters to my company?
Ask whether it is generally available, whether it is on by default, which meter it bills on, and whether it reaches your cloud. EPC Group then maps it to one of the seven decisions.
Is Copilot Cowork generally available?
Yes. Microsoft's documentation describes Cowork as generally available, with access granted through a spending policy. EPC Group covers the controls on its Cowork page.
Is Fabric IQ ontology generally available?
No. Microsoft lists ontology as a preview item. EPC Group recommends certifying semantic models before generating one.
What is TAR-8?
It is EPC Group's 8-surface Tenant AI Readiness Standard, a 24-point rubric for whether a Microsoft environment is ready for Copilot, agents and other AI. EPC Group maps each of the seven decisions to it.
Does EPC Group offer a post-Ignite briefing?
Yes. EPC Group runs a one-hour call with a senior architect that applies the filter to a client's own tenant. Engagements are fixed-fee and scoped after that call.
Ready to act on this?
Start with the practice most relevant to your estate, or reach out directly for a senior-architect conversation.
📧 contact@epcgroup.net · 📞 (888) 381-9725
Multiple models. One truth.
