Ex-DeepMind Founders Raise $20M for Airspeed — Agents That Act on Signals, Not Just Surface Them
Airspeed, the London startup founded by two former Google DeepMind researchers, has raised a $20 million Series A led by DN Capital to build what it calls an execution layer for revenue teams — autonomous agents that act on sales signals instead of merely surfacing them. Vi Partners, Framework Venture Partners and Atlassian Ventures joined the round, which brings total funding past $25 million.
The company was founded as Glyphic in 2022 by CEO Adam Liska and CTO Devang Agrawal, who met as research engineers at DeepMind, and rebranded to Airspeed on 20 May 2026. It says revenue has grown four-fold year over year and that it now serves close to 200 customers across 20 countries, including Persona, Qdrant and Pricefx.
What the product actually does
Airspeed runs agents across calls, emails, support tickets and CRM records. According to DN Capital's investment note, a proprietary orchestration engine processes each conversation within 60 seconds of it ending: the CRM updates itself, follow-ups are drafted, and coaching is delivered to reps and managers. On top of that capture layer sits an agent layer that briefs reps before meetings, re-engages stalled deals and flags forecast risk.
Customers built thousands of custom agents on the platform in the first four months of 2026, DN Capital said, with monthly run volume nearly tripling between January and April. Airspeed states on its site that every agent acts on live call data and asks for approval before anything goes out.
Liska framed the thesis as a problem of execution rather than intelligence. "You can have all the data in the world and still lose because no one executed at the right moment, in the right way," he wrote when announcing the round.
Why it matters for outbound teams
The funding lands on the same argument that has driven most of 2026's revenue-tech M&A and financing: detection is commoditising, and the money is moving to the layer that decides what to do next. Buying a signal feed is easy. Getting a rep to act on the right signal, in the hour it is still warm, is the part that stays broken.
That is the gap most sales stacks still have. Intent and engagement data pile up in dashboards nobody opens between calls. As DN Capital put it, the aim is to stop data "rusting on shelves." If you have never mapped what your own inputs look like, our guide to buyer intent signals covers the categories worth capturing before you automate anything on top of them.
There is a caveat worth naming. Airspeed's agents work on conversations that already exist — calls, threads, tickets, records. That is powerful for expanding and defending pipeline, but it is a post-conversation signal. The earliest signals sit further upstream, before anyone is in your CRM at all: the prospect quietly reading and reacting to your content. LinkedIn engagement signals fire weeks before a first meeting is ever booked, and no conversation-capture layer can see them.
The takeaway
The market is converging on a single conclusion: the scarce asset in 2026 outbound is not data volume, it is acting at the right moment. Teams building toward that should instrument both ends — the upstream attention signals that tell you who is warming up, and the downstream execution layer that makes sure someone does something about it.
traxy handles the upstream half: it learns your ICP, watches who engages with your LinkedIn content, enriches the ICP matches and pushes them to Slack or your CRM in real time, while the intent is still warm. See how it works.
Sources: The Next Web, DN Capital, Airspeed