Watch Microsoft Ignite (17–20 November 2026) and AWS re:Invent (30 November–4 December 2026) for five things that change an AI agent roadmap: how agents are governed, how they get identity and permissions, how they are evaluated, which protocols they support, and what they cost. Web Summit (9–12 November) is better for seeing what startups are shipping.

Picture an operations director with two agent projects planned for 2027: one that drafts replies to supplier emails, and one that updates the CRM after sales calls. In late November, the news feed fills with launches from Microsoft and AWS. Some will make part of the plan cheaper or safer. Most will be noise. The useful move is to decide before the events what you are listening for.

When are Web Summit, Ignite and re:Invent 2026?

The three events run back to back over four weeks, from 9 November to 4 December 2026. Each one has a different audience, so they are worth following for different reasons.

• Web Summit: Dates: 9–12 November 2026. Place: Lisbon. Why follow it: a broad technology conference where AI startups, investors and policymakers meet; good for spotting new tools and how founders talk about agents. Source: Web Summit.

• Microsoft Ignite: Dates: 17–20 November 2026. Place: Moscone Center, San Francisco. Why follow it: Microsoft's main conference for IT and business leaders; announcements land in Microsoft 365, Copilot, Azure and Entra, which many companies already pay for. Source: Microsoft Ignite and Moscone Center.

• AWS re:Invent: Dates: 30 November–4 December 2026. Place: Las Vegas. Why follow it: Amazon Web Services' main conference, with launches across AI, security, migration and developer tools. Source: AWS re:Invent.

Check each event's own site for which sessions are streamed and when the keynotes are scheduled.

Why should a business follow these events at all?

Because the platforms you already use will change what an agent project costs and how hard it is to run safely. A feature announced at Ignite or re:Invent can replace something you were planning to build, or add a control you were planning to buy.

Last year's events show the pattern. At Ignite 2025, Microsoft announced Microsoft Agent 365, which it described as extending "the infrastructure for managing users to agents", with a registry giving a view of "all agents in your organization", including shadow agents. At re:Invent 2025, Amazon announced three "frontier agents" (Kiro autonomous agent, AWS Security Agent and AWS DevOps Agent), plus Policy (in preview) and Evaluations for Amazon Bedrock AgentCore.

Both vendors spent their biggest stage on managing and checking agents, not only on building them. That tells you where the platforms think the problems are.

What should you watch for at Ignite and re:Invent 2026?

Watch for changes in five areas: governance, identity, evaluation, protocols and pricing. These are the parts of an agent project that are hardest to build yourself and most likely to be absorbed by the platform.

Governance

Agent governance is the set of controls that decide which agents exist, who owns them and what they may do. Listen for registries, approval flows and audit logs, and whether they cover agents built outside the vendor's own tools.

Identity and permissions

Agent identity means giving each agent its own account with its own rights, instead of borrowing a person's login. Listen for how agent identities are created, scoped and revoked, and whether they work across the systems your agent touches.

Evaluation and monitoring

Evaluation is testing an agent's answers and actions against a set of known cases before and after launch. Listen for built-in evaluators, live sampling and alerts, and whether you can bring your own test cases.

Protocols

The Model Context Protocol (MCP) is an open standard for connecting AI applications to tools and data. Listen for which products add MCP support, as a client or a server, and what security controls come with it.

Pricing

Listen for how agent features are charged: per user, per agent, per action or per token. A pricing model that fits a pilot can be expensive at production volume.

How do you tell a useful announcement from a demo?

Test it against your own plan with a few plain questions. An announcement is useful when it is available in your region and licence, solves a problem already on your roadmap, and has a price you can estimate.

Before you change plans, check:

• Availability: Is it generally available, in preview, or "coming soon"? Previews can change or be withdrawn.

• Fit: Does it solve a problem on your roadmap, or a problem you don't have?

• Licence: Is it included in what you already pay for, or a new SKU?

• Data: Where does your data go, and does that match your contracts and regulations?

• Lock-in: If you use it, how hard is it to move the agent to another platform later?

• Proof: Has anyone outside the vendor run it on real work yet?

Wait for the documentation and pricing pages, not just the keynote. The details that decide whether a feature is usable rarely fit on a slide.

How should you update your AI agent roadmap after the events?

Hold a short review in the second week of December, once the documentation is out. Go through your planned agent projects one by one and ask whether anything announced changes what you build, buy or postpone.

1. List your planned agent projects. Include the systems each one touches and the controls it needs.

2. Collect the relevant announcements. Keep only those from vendors you already use or are seriously considering.

3. Mark each project. Build as planned, use the new platform feature, or wait for general availability.

4. Re-check costs. Update running-cost estimates with any new pricing.

5. Update your security plan. If new governance or identity features cover something you planned to build, test them before relying on them.

6. Set a date to look again. Preview features often change before general availability; review in the spring.

If you are setting next year's budget at the same time, Sizing an AI Budget for 2027 covers how to fund the foundations these projects need.

How we approach it at Vizio AI

We read the announcements against the client's roadmap, not the other way round. Before the events, we list the agent projects a team has planned and the controls each one needs: identity, permissions, approvals, evaluation and logging.

After the events, we go through the vendors' documentation, not the keynotes, and mark which of those controls the platform now covers, which it covers only in preview, and which still need building. We test any feature we plan to rely on with the client's own cases before it goes into the plan. When an agent needs to reach business systems, we set access control first, as described in Connecting an AI Agent to Your CRM? Secure the MCP Server First, and we test it the way we describe in Before You Put an AI Agent in Front of a Client.

This is part of our DX Studio work.

Sources

• Web Summit, Lisbon, 9–12 November 2026: https://websummit.com/

• Microsoft Ignite: https://ignite.microsoft.com/

• Moscone Center, Microsoft Ignite 2026: https://moscone.com/events/microsoft-ignite-2026

• AWS re:Invent, 30 November–4 December 2026: https://aws.amazon.com/events/reinvent

• Microsoft, Ignite 2025 Book of News: https://news.microsoft.com/ignite-2025-book-of-news/

• Amazon, AWS re:Invent 2025 AI news and updates: https://aboutamazon.com/news/aws/aws-re-invent-2025-ai-news-updates

• Model Context Protocol: https://modelcontextprotocol.io/