Gamma makes slides. Gamma just bought a startup to figure out whether slides should exist at all. Both of those things are true as of 2026, and the gap between them tells you a lot about where AI tools are heading.
The deal itself is small in the way most acquisitions in this space are small. Gamma acquired Lica, a design startup backed by Accel, and the terms were not disclosed. Lica’s co-founders are not being folded into an existing team or handed a product roadmap. They are being handed a new AI design research lab to lead, with a focus on new presentation methods.
If you are not steeped in startup news, that last sentence might read as corporate filler. It isn’t. A research lab is a specific kind of bet, and it’s a strange one for a company whose core product already works.
What Lica actually did
Lica’s product turned screenshots and recordings into presentations and videos. You gave it raw material — a screen capture, a recording of you talking through something — and it produced a finished artifact on the other end.
That is a different starting point than most AI presentation tools. The usual pattern is: you type a prompt, the tool generates slides. Lica’s pattern was: you already made something messy, the tool reshapes it. The input is evidence of work rather than a description of work.
For anyone trying to understand AI agents, that distinction is useful. A tool that takes a prompt is doing translation. A tool that takes a recording and rebuilds it into something else is doing interpretation. It has to decide what mattered in the thing you made, what was noise, and what shape the result should take. Those are judgment calls, and judgment calls are where agent behavior gets interesting and also where it gets wrong.
Why a research lab and not a feature
Gamma could have bought Lica, extracted the technology, and shipped a “convert your recording to a deck” button. That would have been the tidy version. Instead the co-founders get a lab and a research mandate.
Reading between the reported lines, the reasoning that makes sense to me is this: the slide is a legacy format. It exists because someone in the 1980s needed a metaphor for a physical transparency on an overhead projector. Every rectangle, bullet point, and title placeholder in modern presentation software descends from that constraint. The constraint is gone. The format stayed.
Once a machine is generating the artifact instead of a human dragging text boxes around, the old format stops being necessary and starts being a limitation. Reporting around the deal points at exactly this question — whether AI-generated presentations could adapt to their audience and move past conventional slide-based communication.
You cannot get to that answer with a feature sprint. You get there by letting a small team spend a long time asking uncomfortable questions about the thing that pays the bills.
What “adapts to the audience” would mean for you
Here is where it gets concrete for non-technical readers. Think about what a presentation actually is: one fixed set of content, delivered to many different people, all of whom need different things from it.
- Your CFO wants the numbers and will skip your setup.
- Your new hire needs the setup and will drown in the numbers.
- Your customer wants to know what changes for them and nothing else.
Today you solve that by making three decks, or by making one deck and talking over it differently each time. An agent-driven format could in principle solve it by reassembling the same underlying material differently for each viewer. Same facts, different path through them.
That is a genuinely different product than a slide generator. It also raises questions worth being skeptical about. If the content reshapes itself per person, who is accountable for what a given viewer saw? How do you review something before you send it if it won’t look the same on the other end? Those aren’t reasons not to try. They’re the reasons a research lab is the right container for the attempt.
The pattern to watch
Small acqui-hires like this rarely make headlines outside industry press, and they’re easy to skim past. But they’re a decent signal of where product teams think the ceiling is. When a company buys a team specifically to question its own format, it’s admitting the current version is a local maximum.
I’d watch for two things from Gamma’s lab. First, whether anything ships at all — research labs inside product companies have a habit of producing interesting papers and no shipped changes. Second, whether the output is a better slide or something that isn’t a slide. Those are very different outcomes, and only one of them justifies buying a company to get there.
🕒 Published: