James Thewlis and Sasha Haco / Unitary
“I want insurers to move faster,” says Sasha Haco, co-founder and CEO of London-based Unitary.
Her startup is bringing AI-powered “virtual agents” to the manual, legacy-heavy sector, using software that can learn complex human tasks and integrate like a new member of staff.
Haco founded Unitary in 2019 alongside chief scientist James Thewlis, who she met at Entrepreneur First. The company started life far from the world of underwriting and claims, with the pair originally solely focused on automating human review in content moderation, helping platforms like Substack, Depop and Trustpilot identify and manage harmful user-generated content.
“We started out doing automation in the content moderation world where there are lots of manual processes around human review of harmful content,” Haco tells Pathfounders. And we were building better AI models originally to identify that kind of content, and then we shifted a little bit to trying to actually help the human work.”
From there they moved into automating more complex cases, from user generated content (UGC) to marketplaces, replacing a range of back office tasks.
Agentic AI is far from new, with companies like Counterpart, which offers agentic underwriting, risk mitigation and claims management among other services, and Lyzr, which Accenture just backed to bring agentic AI to both banking and insurance companies.
But unlike agentic AI, which consists of exploratory, autonomous models that make their own decisions with limited or no human input, the “virtual agents” follow more of a formula.
“[It’s] a mix of software and AI which gives them built-in guardrails and means they're completely predictable,” Haco says.
That focus on precision and predictability has roots in both founders’ scientific backgrounds. Haco holds a PhD in theoretical physics and previously worked with Stephen Hawking on black hole research before turning to startups. The company’s name, a nod to the “unitary transformation” in physics, reflects that scientific rigour.
And, unlike AI co-pilots that have become common in productivity tools, the agents are designed to do tasks, rather than just assist.
“It's not about just helping someone do their job faster,” Haco says. “It's taking on the full workload of a bit they don't want to do.”
Agnostic technology
So while it perhaps seems like a surprising new industry to dive into — moving from content moderation to insurance — once the Unitary team realised they had built a very general model that was able to learn any task, rather than one specifically tied to content moderation, that this technology could then be applied elsewhere.
There is no tech integration required, with the agent able to log into a complex mix of tools and legacy systems. There’s no API integrations or engineering setup, and once the company sets up a username, password and some instructions, they’re good to go.
“We integrate like a human,” she says, “as though they’re hiring another member of staff.”
That flexibility has helped Unitary secure early partnerships in insurance, including its first client, Attune.
The startup has raised $25 million so far: an initial $10 million across pre-seed and seed rounds, followed by a $15 million Series A in 2023 led by Creandum with participation from Paladin Capital Group and Plural.
The startup will “probably look to raise at some point late next year” as they look to grow and explore new areas to branch into. Haco is interested in exploring applications in healthcare and logistics, but says the focus for the next year or so will be firmly on insurance.