Inside Neat's push to bring meeting rooms into the agentic AI era with Tormod Ree

• Hybrid work isn't done. Most collaboration tools were designed before remote work was the default; big rooms still break down, and remote participants still feel disconnected even when the meeting technically "works."

• Neat bets on distributing compute across every device in a room rather than packing it into one central codec, which lets customers swap, add, or reconfigure hardware without re-architecting the space.

• Meeting rooms are the blind spot in the agentic AI stack. Until rooms can feed agents real context (who said what, what got decided), the productivity story everyone is selling will keep underdelivering.

Let's dive in. No floaties needed…

Revenge of the Nerds

Tormod Ree, Chief Product & Engineering Officer of Neat

Tormod Ree is the Chief Product & Engineering Officer at Neat, the Oslo-based video technology company building the bars, boards, and companion devices that sit in meeting rooms running Zoom, Microsoft Teams, Google Meet, and Neat BYOD. Founded in 2019 by former Tandberg executives Ivar Johnsrud, Simen Teigre, and Haakon Sporsheim, Neat has raised over $73M (Zoom is among its investors), sold more than 500,000 devices to 20,000-plus enterprises, including Atlassian, HubSpot, and Rakuten, and grown to roughly 500employees.

After 18 months in the Norwegian Army and an MSc in Telecommunications from NTNU, he spent nearly eight years at Cisco, then co-founded Ava Unified Security, a cloud-native, AI-built-in video security platform that Motorola acquired in 2023. He stayed on as VP of Avigilon Alta Video before joining Neat at the end of 2024. If you've ever been the remote person on a 16-seat conference call, wondering why the camera is pointed at the empty chair, you already understand the problem he's working on.

First, tell us about Neat & who you're building for

Neat was started to create what we like to call simple and intelligent devices for video meetings. We provide devices that go into meeting rooms for Zoom, Microsoft Teams, Google Meet, or Neat BYOD. They range from video bars for meeting rooms to all-in-one devices with integrated displays, which we call Boards. It's a global company now, growing 30x YoY. In addition to that, we provide a platform for third-party applications, so you can run things like interactive whiteboarding or project management tools in the meeting room.

We came to market with something very simple. This is a world that typically had a lot of complexity: lots of different devices, cameras, computers, controllers, and speakers. It was pretty hard to set up, maintain, and use. Neat did something very simple, and we've also been pioneering the use of AI from the beginning to add a layer of intelligence on top: automatically detecting people, framing them, better audio quality, making sure the people you want to see are the ones you see. That's the core of what Neat does.

Hybrid work was supposed to be solved by now. What's still broken?

I don't think anybody believes hybrid work or a distributed workforce is perfect. People typically prefer to be in the same room with someone. Remote is good for focused work, but in general, there's quite a lot that can be done better. Most of the tools we use to collaborate today were designed before distributed work became the norm and before modern AI existed. They're built around workflows and assumptions from a different era, which is why many of today's AI features feel bolted on rather than foundational.

Meetings are okay if it's one-to-one, or if you don't have too many people in the room. It can break down once you're in a room with 10 or more people; that's not always the best experience.

And remote people still have a level of disconnectedness from what's going on. You can't just walk over and talk to somebody; you need to schedule a meeting and decide when to talk about something. It's not as interactive. People sometimes feel a bit disconnected. There are definitely still things left to do, but that makes it interesting from a product perspective.

Why distributed intelligence instead of one central box?

When Neat started, most of what you put in meeting rooms for meetings was pretty complex. Neat brought to market a very simple, all-in-one device for small and medium rooms. You put in this thing, you're good to go, hardly any cables, and it always works. In larger rooms, things are still complex. You have central computing (often called a codec), cameras, speakers, and microphones. It's complex to buy, install, and troubleshoot, and complex for users to use. That area requires a new approach.

That's where our distributed architecture comes in. One, it connects everything over the network, not directly or with custom cables. Two, it distributes compute, so every device has compute to add AI and intelligence to the mix. You don't have to decide that upfront with one box that does everything. Every device you put into the room adds compute and edge AI. It gives you more flexibility because you can add and remove devices to change or increase what that room does for you. And we think we can cover more use cases because our things are connected through the network and software, not direct cables. If you want to change how you use one meeting room from one day to the next, that's possible because we use software and architecture to reconfigure and adapt to whatever meeting people are having.

What are cloud-native AI teams missing because they don't ship hardware?

Building hardware forces you to think differently because you're operating at the intersection of software, AI, acoustics, cameras, and physical environments. One, it's pretty hard to correct a bug once you've shipped it if it's actual hardware, because that could involve replacing the device. That means it requires a different level of care across development, manufacturing, QA, and even packaging. It's about the full development and production lifecycle working together as one system. You can't try something, fail, and fix it the same way. It's a different discipline when developing the product. It's hard in different ways, but we like it. I've always enjoyed working with both software and hardware.

From a feature and experience point of view, we have many more people working on software in our company than on hardware. The hardware is important and is an enabler for our experiences, but a lot of the experiences are driven by software design. We have to design the hardware for the experience and be mindful of what the hardware can do when we build new experiences. Our devices also get replaced less often than other things in an office. Customers will use them for at least three years, sometimes five, sometimes longer. We have to be mindful of that when we ship new features.

What did Ava Unified Security teach you about building AI products?

Quite a few things. We built a cloud-native solution for video security with AI from the start. The most important thing I learned is that if you build AI in from the start in your product, in the core fabric, you can design the product differently. You can assume it's always there. It's not something you turn on or off in this part of the product, or for that camera; it's not something you opt into or need an additional license for. That gives you more flexibility because you can use it for everything the product does. In our case, we used it to decide what footage to store in higher or lower quality. We used it to present the right video to the operator and to search for videos faster, which was a big advantage.

You can draw a parallel to AI today. Some companies are approaching AI features by adding a button on top of their existing interface or adding a new feature here and there. That doesn't take full advantage of it, because you're slapping something on top. What AI actually allows you to do is rethink your product from the ground up: the interaction model, the UI, and how people use it. So when there are bigger shifts like that, you need to approach them with an open mindset and rethink properly how it would shape the experience. That was probably the most important learning.

The other thing relevant to AI: at Ava, we did devices at the edge and a cloud service. It was important to flexibly think about where we wanted to do which type of AI and processing, because we had compute at the edge and compute in the cloud. We were constrained at the edge, but it was effectively free, because it was always there. The cloud has more power, but more cost and more latency. Having an architecture that lets you be flexible matters. It's not either-or, it's both-and. You might want to do something at the edge because it's more cost-efficient, time-critical, or because the data is sensitive (which is often the case for meetings). Other things can go to the cloud. That's something I learned, and I was happy to find a similar approach when I started at Neat.

If Neat gets this right, what does the meeting room look like in five to 10 years?

Most importantly for the user, the technology is not top of mind for their experience. Today it's too present and in the way. You go into a meeting room and have to figure out how you're supposed to join. You join a meeting and immediately have to deal with issues around audio, muting, volume levels, or camera framing. There are even scenarios where the people in the meeting room decide what camera framing the remote participants see. So while you're having an important discussion, you have to decide what camera framing other people are going to see. Most meeting experiences are still remarkably static.

With AI, it will be possible to read the room, for the platform and the devices to actually understand what's going on. We can already identify who's speaking, but what kind of interaction is taking place? Is somebody lecturing, presenting, is it a workshop, who's more or less likely to be relevant to other participants? That should just work. There shouldn't be any changing of framing, modes, or cameras. In that way, the technology is transparent.

The second area is that AI, and especially how people are starting to use agents to support their workday, will have an impact on meeting rooms. For agents to work effectively, they need proper context. They need to know what was discussed, what was decided, and who said what. Meeting rooms are a blind spot for them today. You can sort of work around it with a transcript, but you don't know who said what, then you can pipe it into Claude or whatever and try to do something. They're not connected to any kind of agentic workflow. That has to change.

The third area is potentially having something in the room that supports you during or after the meeting. If you're discussing products, maybe you need to update the Jira tickets. If you're on the sales team, it should update the CRM. Those things should just happen. Transparent or invisible in a good way, supportive with AI in another. That's clearly where the industry is heading and what people need, and it's our intention to do that well going forward.

Additional Reads

• The Future of Meeting Rooms — with Tormod Ree (Microsoft Teams Insider Podcast, October 2025)

• AV Interview: Tormod Ree, Chief Product and Engineering Officer, Neat (AV Magazine, August 2025)

• rAVe [TV] Episode 242: The Future of Meeting Rooms (rAVe [PUBS], November 2025)

Quick Poll

Rate This Edition