Agentic AI moves from roadmap to rollout

By Alistair Johnston, Principal Analyst, Pro AV, Futuresource Consulting

After two years of pilots and roadmap slides, 2026 is the year collaboration AI shipped. Alistair Johnston, Principal Consultant for ProAV and Enterprise Collaboration at Futuresource Consulting, considers where the meeting room sits in the new wave of agentic platforms, and what it means for the integrators and end-users walking the InfoComm floor.

For most of the last two years, the conversation about artificial intelligence in the meeting room has been a conversation about features that were almost ready: meeting summaries, action item capture, smarter framing, real-time translation. The story was credible, but consistently ran ahead of what enterprises could actually deploy. The opening months of 2026 have changed that. Across every major UCaaS vendor, agentic AI has moved from roadmap item to shipping product, and the positioning has shifted with it. The platforms are no longer selling AI as an assistant that summarises a meeting; they are selling AI as an executor that turns the meeting itself into a trigger for downstream work.

Zoom has set the tone. AI Companion 3.0 now has a dedicated tab in the Zoom Workplace desktop application, and the company reported that AI Companion monthly active users more than tripled year-on-year in its most recent quarter. The more consequential development is the Custom AI Companion agent builder, a no-code orchestration environment allowing organisations either to deploy prebuilt agents for sales, IT and marketing, or to construct their own agents capable of interacting with Salesforce, ServiceNow, Slack, Box, Google Drive and OneDrive. Alongside this sit AI Docs, AI Sheets and AI Slides, which generate documents, spreadsheets and presentations directly from meeting context, together with real-time voice translation, deepfake risk detection and photorealistic avatars. The framing from Eric Yuan is that work itself is moving from person-to-person communication into a mix of person-to-machine and machine-to-machine task completion, with the meeting as the orchestration point.

Microsoft has taken a different route but arrived in similar territory. Recent Teams updates introduce a Facilitator Agent that produces meeting notes which update live during the call rather than after it; an AI Video Recap delivering narrated highlight reels with timestamped links to decision points; automatic spoken-language detection; and custom dictionaries that extend recognition to industry-specific terminology. Copilot now draws across chat history, transcripts and calendar context as a unified surface. With more than fifteen million paid Microsoft 365 Copilot seats in circulation, enterprise buyers are evaluating Teams Rooms refresh cycles against AI capability rather than basic conferencing parity.

Google's contribution has been less about Meet-specific features and more about the ambient intelligence layer surrounding the meeting. The March Gemini Drop extended Personal Intelligence to all US Gemini users, allowing Gemini to connect to Gmail, Drive and Calendar in order to prepare context for meetings and tasks without manual setup. Combined with Model Context Protocol support in the Gemini API, and the move to Gemini 3.1 Pro as the flagship model, Google has quietly positioned itself as the integration layer for any organisation already standardised on Workspace.

The hardware story has kept pace. Logitech's Rally AI Camera Pro, certified at launch for both Teams Rooms and Zoom Rooms, pairs a twenty-megapixel one-inch sensor for AI-driven digital PTZ with a mechanical PTZ lens for speaker close-ups. Jabra's PanaCast 55 VBS, announced at ISE in February, combines a 180-degree panoramic camera with the Huddly Crew multi-camera system, and is the first multi-camera Android solution available for both Teams Rooms and Zoom Rooms. HP's HP IQ platform runs a twenty-billion-parameter model locally on device and captures meeting notes without app switching, while AVer's MT500 targets the larger-room operator-free segment through intelligent voice tracking and automatic camera switching. The pattern is consistent: intelligent framing, noise suppression and speaker tracking are no longer differentiators but baseline expectations of the category.

Three implications follow for those walking the InfoComm floor. The first is that software and hardware certification cycles are genuinely converging: the Logitech, Jabra and HP launches are timed to Teams Rooms and Zoom Rooms certification windows, with the consequence that refresh decisions taken in the second half of 2026 are in effect decisions about which AI ecosystem an organisation is committing to for the next three to five years. The second is that the integration surface is widening rapidly. Custom agents reaching into Salesforce and ServiceNow are not meeting room products in any traditional sense; they are workflow products which happen to be triggered by meetings, and integrators who continue to treat the room as the endpoint of the conversation will find their scope of work shrinking accordingly. The third is that the procurement question itself is changing shape. Eighteen months ago, buyers were asking whether a given platform had AI; today they are asking where that AI is actually saving labour, reducing downtime or improving decisions, and that is a considerably harder question for any single vendor to answer alone.

The collaboration platforms have spent two years promising that AI would change the meeting. What 2026 has made clear is that the more interesting shift is happening on either side of it, in the preparation, the follow-through, and the workflows the meeting feeds into. The room remains central, but the value increasingly lies beyond it.

For more information on our collaboration reports visit here or our other solutions please visit here.