London-based Ethos has raised a $22.75 million Series A led by Andreessen Horowitz, with participation from General Catalyst, XTX Markets, and others, for an expert network that uses voice agents to onboard professionals and build a knowledge graph of who actually knows what, rather than relying on job titles and credentials to match companies with relevant expertise.
The company was founded by James Lo, a former McKinsey consultant and SoftBank Vision Fund operator, and Daniel Mankowitz, an AI researcher who came from Google DeepMind. Their starting premise is that the economy is a knowledge graph of people, companies, and products, and that existing expert networks have been building that graph with blunt tools. Traditional networks like GLG match clients with experts based on job history, seniority, and company names. Those signals are useful but shallow. A former partner at a consulting firm who spent five years working specifically on pharmaceutical regulatory submissions in Asia knows something very different from a colleague at the same level who worked on generic strategy projects. Job title matching misses that distinction entirely. Ethos is trying to capture the difference using voice interviews conducted by AI agents, which can ask more questions, explore nuance, and extract domain-specific knowledge that a LinkedIn profile never surfaces.
The voice onboarding mechanism is the core product bet. When an expert joins Ethos, they do not fill out a lengthy form or manually tag their areas of knowledge. An AI agent conducts a structured interview, asks follow-up questions based on what the expert says, and builds a richer knowledge profile from the spoken responses. The company then combines that interview data with public signals, including academic papers, blog posts, GitHub repositories, and other indexed sources, to construct a multi-dimensional picture of what each person actually knows. Ethos says it is now onboarding approximately 35,000 experts per week, which at that volume suggests the AI-driven intake is genuinely scalable in a way that manual profiling could never be. The product then allows clients to query that knowledge graph in natural language, asking for experts who match complex criteria that traditional filters cannot handle, such as someone who worked at a funded startup solving for finance automation or a doctor who both specialises in oncology and has published on drug development processes.
The commercial traction is not theoretical. Ethos said it is on track for eight-figure annualised revenue without disclosing a precise number, and its client base includes top hedge funds, private equity firms, leading AI labs, and enterprise consulting firms. The company takes 30 percent or more as a per-project fee depending on the nature of the engagement. That fee structure is more aggressive than GLG's typical model but defensible if the matching quality justifies it. Clients who find the right expert on the first query rather than reviewing 20 candidates save time that is worth more than the fee difference. The bet is that better matching commands a premium, and that AI-powered knowledge profiling is what enables that matching to improve.
For startups trying to understand where GLG is vulnerable, the answer is in the acquisition and matching economics. GLG has built its network through years of manual expert recruitment, and its matching involves human analysts who understand client needs and guide conversations. That process is high-touch and expensive. Ethos automates both sides at scale, which means it can grow the expert supply faster and serve clients at lower operational cost. If the knowledge profiles are accurate enough, the automated matching can reach GLG's quality at a fraction of the overhead. The biggest risk is that voice onboarding produces lower-fidelity profiles than a human interview, or that the AI cannot ask follow-up questions at the level of nuance that truly distinguishes one expert from another in highly technical domains. Ethos will need to demonstrate over time that its knowledge graph is deep enough to justify displacing incumbents rather than simply undercutting them on price.
The voice onboarding angle also has implications beyond expert networks specifically. It points to a broader question about where voice becomes the lowest-friction data collection layer for enterprise AI. Forms require deliberate attention and consistent format. Text profiles require users to know what to write. Voice interviews meet the user in a conversational register they already understand, and AI can extract structured data from unstructured speech at a level of detail that forms cannot. If that extraction is good enough, voice becomes the obvious intake mechanism for any enterprise AI application that needs to build rich profiles of people, whether they are experts, job candidates, customers, or subject matter experts inside a company.
For SF readers, the Ethos round is a clean signal about how AI is reshaping professional services distribution. The company is not just building a better expert network. It is building the infrastructure argument that voice is the natural onboarding layer for professional knowledge, and that AI can run the intake at a scale that makes the knowledge graph continuously better as the expert base grows. A16z backing that thesis with a Series A lead is a vote that the technical pieces are far enough along and the market timing is right. If Ethos is right about the knowledge graph and wrong about nothing else, GLG has a problem.
Also read: Alphabet and Amazon's AI startup gains are now driving more than half their headline earnings • Match Group is slowing hiring to fund AI and consumer internet has its new margin script • Amazon's e-book market is getting flooded by AI-generated titles and the discoverability crisis is just beginning
