In this blog post Ignite 2026 Preview What Australian CIOs Should Do Before 17 November we will explain what the event is likely to mean for your AI plans, security controls, cloud costs and 2027 technology roadmap.
Microsoft Ignite runs from 17 to 20 November 2026 in San Francisco and online. The published agenda points to a clear theme: Microsoft wants AI agents to move beyond experiments and become a managed part of everyday business operations.
For CIOs, that creates an immediate problem. The product announcements will arrive quickly, but decisions about budgets, security and governance cannot be made from a keynote slide. The organisations that get value from Ignite will be those that know their gaps before the event begins.
The main technology to understand is AI agents
An AI agent is software that can interpret a goal, find information and take actions across approved business systems. Unlike a basic chatbot, an agent may update a customer record, prepare a report, monitor an operational process or escalate an issue without waiting for another prompt.
CPI Consulting helps organisations design, build and secure Azure AI solutions, from Azure OpenAI and Microsoft Foundry to agents, governance and production deployment.
Visit CPI ConsultingUnderneath the friendly interface sits a combination of an AI model, company data, software connections, an identity and a set of permissions. The agent also needs rules describing what it may do, when a person must approve its actions and how every decision is recorded.
This explains why the Ignite agenda extends well beyond Microsoft 365 Copilot. Microsoft is connecting Foundry, its platform for building and managing AI applications, with Agent 365, which provides central visibility and control over agents. Entra manages identities and access, Purview monitors and protects sensitive data, and Defender detects security threats.
The technology is becoming easier to deploy. Governing it safely is becoming the harder job.
1. Build an agent register before buying more AI
Most mid-market organisations already have more AI activity than leadership realises. Teams are testing Copilot, ChatGPT, Claude, automated workflows and small internal assistants, often without a shared inventory.
Before Ignite, list every AI tool or agent currently being used or tested. Record its owner, purpose, data access, model, monthly cost and whether it can take actions without human approval.
This gives you a baseline for assessing Agent 365 and other governance announcements. It also exposes duplicated subscriptions, abandoned trials and agents that still have access after their original project ended.
If you are already using Microsoft Foundry, compare your register with the actions in our Microsoft Foundry September 2026 CIO checklist. The priority now is moving from separate technical projects to one operating model.
2. Decide what an agent should never do alone
Ignite sessions are expected to show agents completing increasingly complex work. That is useful, but technical capability should not automatically become business permission.
An agent drafting a response to a customer is different from an agent issuing a refund. Summarising an incident is different from changing a security policy. Reading a contract is different from approving one.
Define three categories before 17 November:
- Low risk: The agent can complete the task and record what it did.
- Approval required: The agent prepares an action, but a person must approve it.
- Prohibited: The agent cannot perform the task, even if technically capable.
This is consistent with Australian Signals Directorate guidance encouraging incremental agent adoption, limited privileges, continuous monitoring and human oversight. It also gives your security and operations teams practical rules instead of asking them to “use AI responsibly”.
3. Treat identity as the new AI security boundary
Every agent needs an identity, just as an employee does. That identity determines which files, applications and business processes the agent can access.
The risk appears when an agent inherits excessive permissions. It may expose information to the wrong employee, follow a malicious instruction hidden in a document or take an action outside its intended role.
Expect identity security to be a major Ignite theme. Entra, which controls access to Microsoft systems, is being positioned to manage access for both people and agents. Defender, Microsoft’s security platform, and Purview, its data protection and compliance platform, add threat detection, audit records and controls over sensitive information.
Before the event, review privileged accounts, unused application connections and shared service identities. Essential Eight, the Australian Government’s cybersecurity framework, remains an important foundation because patching, restricted administration and strong authentication reduce the environment an attacker can exploit.
However, Essential Eight alone will not answer who approved an agent, what data it used or why it made a decision. Your AI governance model must address those additional questions.
4. Measure AI economics before expanding licences
One of the easiest ways to waste money in 2027 will be licensing AI broadly without measuring whether it changes business performance.
Consider a 200-person professional services company that gives 120 employees an AI assistant. If each employee saves ten minutes per working day, that represents roughly 4,400 hours of annual capacity. But it is only valuable if those hours improve billable work, customer service or turnaround times.
For each AI use case, establish a baseline before Ignite:
- How long does the task take today?
- How often is it completed?
- What is the cost of errors or delays?
- How will quality be checked?
- What are the model, licence and support costs?
Foundry is increasingly focused on monitoring agent quality, reliability, speed, safety and cost. These measurements will matter more than model benchmark scores.
Model choice should also remain open. OpenAI may suit one workload while Anthropic Claude or a smaller model may deliver better cost, speed or accuracy elsewhere. Our earlier guide on comparing Claude across Microsoft Foundry and Anthropic explains why the platform, governance and commercial terms can matter as much as the model itself.
5. Check whether devices and cloud systems are ready
AI does not remove the need for reliable devices and infrastructure. It increases the dependency.
Intune, which manages and secures company laptops, phones and applications, is a prominent part of the Ignite agenda. Microsoft is positioning managed devices as the control point for deciding which users and agents can access company data.
Windows 365, which provides secure Cloud PCs accessed over the internet, is also expanding beyond traditional remote desktops. Microsoft is exploring Cloud PCs as controlled environments where agents can perform computer-based tasks without being given unrestricted access to an employee’s physical device.
Review your device compliance, software update and recovery position now. An exciting AI feature will not fix unmanaged laptops, inconsistent patching or weak administrator controls.
The same applies to Azure. AI services depend on data, networks, applications and backups remaining available. Use our Azure resilience checklist for Australian mid-market CIOs to identify recovery gaps before adding another business-critical dependency.
6. Prepare for Australian privacy changes in December
Ignite takes place only weeks before an important Australian privacy deadline. From 10 December 2026, covered Australian Privacy Principle entities may need to update their privacy policies when computer programs make, or substantially support, decisions that significantly affect an individual’s rights or interests using personal information.
This may be relevant to AI used in recruitment, credit, insurance, employee management or customer eligibility. CIOs should work with privacy and legal leaders to identify affected systems rather than assuming this is solely a compliance team issue.
Ask vendors whether their new tools provide decision records, data lineage and exportable audit evidence. If you cannot explain what information influenced an automated decision, proving compliance later will be difficult.
Your checklist before 17 November
- Create an inventory of AI tools, agents, models and owners.
- Document which agent actions require human approval.
- Record current AI spending and business performance baselines.
- Review agent identities, permissions and access to sensitive data.
- Check device management, patching, backups and recovery testing.
- Identify automated decisions that may affect Australian privacy obligations.
- Assign team members to follow Foundry, Agent 365, security, Intune, Windows 365 and Azure announcements.
- Book a post-Ignite review before calendars fill in December.
Watch Ignite with decisions in mind
The goal is not to return from Ignite with a longer product wish list. It is to identify which announcements reduce cost, close a known risk or improve a measurable business process.
CloudProInc brings more than 20 years of enterprise IT experience to these decisions. As a Melbourne-based Microsoft Partner and Wiz Security Integrator, we help organisations connect Azure, Microsoft 365, Intune, Windows 365, Defender, Wiz, OpenAI and Claude into a practical plan rather than another collection of disconnected tools.
If you are not sure whether your environment is ready for what Microsoft announces on 17 November, we are happy to review your current AI, cloud and security position and help you separate useful changes from expensive distractions — no strings attached.