Voker raises $2.2M to track how AI agents perform in the real world
What's the deal? Voker, an agent analytics platform for AI product teams, has raised $2.2M in pre-seed funding from Y Combinator and FundersClub. The startup helps companies understand how their AI agents perform once deployed — shifting focus from building agents to measuring whether they actually deliver value.
"Product teams have this onus to deliver on the marketing claims," co-founder and chief executive Tyler Postle told SiliconANGLE. "They're going to start getting asked by executives how many new products are sold through this agent. They don't have a way to measure that."
Why now? AI agents are flooding into customer-facing and internal workflows, but companies are discovering a gap between launch and results. Postle said that getting agents into production isn't hard with pretrained large language models — the challenge is that once deployed, things rarely go as planned.
Existing observability and evaluation tools are built for engineers doing trace debugging. They break down when teams need to analyse thousands or millions of monthly conversations and understand agent performance at scale. Voker targets the wider product team — designers, product managers, executives — not just engineers.
What could go wrong? Voker is entering a crowded AI tooling market where observability platforms are already well-funded and expanding their capabilities. The startup currently targets customers with agents already in production handling at least 1,000 conversations a month — a narrow slice that could limit early growth if enterprise agent adoption stalls.
There's also what Postle calls the "ask me anything" problem: companies set agent expectations too high based on marketing overpromises. If the broader market struggles to deploy useful agents, the demand for analytics on those agents could lag.
The signal: The AI industry is maturing past the build phase and into the accountability phase. Companies need to prove ROI on their AI investments, and that requires tooling beyond what engineering dashboards provide. Voker's bet is that agent analytics will become as essential as product analytics is for traditional software — surfacing not just errors, but user intent, unmet needs, and opportunities to improve.
For example, a hotel booking agent might field questions about an attached restaurant. Voker could surface those conversations and show product teams where to expand the agent's capabilities — turning raw user behaviour into actionable product decisions.
Read more: siliconangle.com
