Twin1 AI Secures $20 Million to Develop Digital Replicas of Corporate Employees

AI startup raises $20m to build ‘digital twins’ of office workers

The San Mateo- and London-based startup is building AI replicas of individual employees, using workplace data to answer questions, perform some tasks and connect knowledge across a business. The pitch matters because investors are putting serious early-stage capital behind a product that sits directly on a company’s communications infrastructure.

The bet is on employee data as software

Twin1’s system can draw information from emails, meetings, documents and workplace tools including Slack, Microsoft Teams, Outlook, Gmail, Google Drive and SharePoint. The intended result is an AI “twin” that can respond on an employee’s behalf while making that person’s accumulated knowledge available to colleagues and other AI systems.

That is a narrower proposition than modelling an entire company. Salesforce has outlined a concept for an “enterprise digital twin” designed to model business areas and test decisions. Twin1 is taking the opposite route: start with the individual, then connect those individual systems across the organisation.

The company says its customers include law firms Linklaters, Orrick and Dechert, US lender Customers Bank and energy group Aegis Energy. It also claims early customers have automated between 30 and 50 per cent of communications work. That figure has not been independently verified, so it should be treated as a company-reported operating claim rather than an established benchmark.

For buyers, the distinction is material. “Automating communications” can mean anything from drafting routine responses to handling more consequential internal requests. The funding announcement does not establish how much work is completed without human review, how performance varies by customer, or whether the claimed range translates into measurable labour-cost savings.

Follow the money, then inspect the access controls

Twin1 plans to use the cash to expand its London and California teams, develop the technology and increase sales. The round also included EJF Ventures, F-Prime, Lakestar, Notion Capital and Orrick, alongside angel investors. Several backers had previously invested in Eigen Technologies, the AI document software company founded by members of the Twin1 team. Eigen was acquired by Sirion in 2024.

There is no valuation disclosed in the available material. That leaves the $20 million as the clearest signal of investor conviction, but not a basis for judging the startup’s eventual multiple or exit economics. Seed capital can fund product development and distribution. It does not prove durable demand.

The practical diligence questions are therefore straightforward:

• Which systems can an employee’s twin access by default?
• Can the employee restrict what is collected and what is shared?
• What can the twin do without approval?
• Are the reported automation rates measured consistently across customers?
• Does the product reduce paid work, speed up existing staff, or simply move review work elsewhere?

Twin1 says employees control what information their twin can access and what can be shared with colleagues or other AI systems. That is an important product feature, but it remains a company statement. Customers assessing the platform will need to test those controls in real workflows rather than infer protection from the interface.

A crowded knowledge-automation trade

Twin1 is entering a market where startups are trying to turn institutional knowledge into software. Lithuanian AI company Guideless, for example, raised €1 million in pre-seed funding for a platform that converts software workflows into structured, AI-narrated training guides.

The commercial logic is clear. Companies lose time when expertise remains trapped in inboxes, meetings and scattered documents. The harder question is whether an AI replica preserves that expertise accurately enough to be trusted, while keeping the employee in control of its use.

That is the part investors have not yet solved with a cheque. Twin1 now has capital, named enterprise customers and a credible distribution story. Its next test is less theatrical: prove that the “twin” creates repeatable savings, survives workplace complexity and does not turn every employee’s digital history into another expensive system to monitor.