millionways Builds Behavioral Intelligence for Enterprise AI

millionways is building behavioral-intelligence infrastructure for decisions that ordinary enterprise software cannot explain. The New York company uses its Thorsten-4 model to analyze text and audio for patterns related to trust, pressure, communication fit, and decision-making across workforce, customer, and human-agent systems.

Co-founders Martin Cordsmeier and Max Weidemann lead a company moving from a long research program into enterprise distribution. A strategic investment from Hunt Scanlon Ventures gives millionways a direct route into executive search and leadership assessment, while partnerships in AI governance place the same technology inside the emerging market for agent observability.

About millionways

millionways describes Thorsten-4 as a Large Psychology Model. Unlike a general-purpose language model optimized to predict and generate language, Thorsten-4 is intended to extract behavioral signals from communication and connect them to enterprise outcomes. The company says the model can analyze humans, AI agents, and the interaction between them.

The origin story begins before the current enterprise product. Cordsmeier has described interviewing people across ages and economic backgrounds through a predecessor nonprofit, then turning that listening process into a system that could scale. The company's current science overview frames that history as a decade-long behavioral research program.

Public sources do not support one clean legal founding year, so the more useful marker is the research continuity. millionways says Thorsten-4 draws on more than 35 million behavioral observations, more than 100,000 research hours, work by over 50 behavioral scientists and psychologists, and more than 7,000 participants. Those are company-reported development figures, not independently verified results.

What Thorsten-4 Does

The Thorsten-4 platform analyzes communication from meetings, interviews, customer conversations, emails, documents, support interactions, and AI conversations. Its outputs include behavioral measurements, derived behavioral factors, purpose-built Enterprise Behavioral Models, and recommended actions.

millionways groups those models into three operating areas. Workforce and leadership models support candidate intelligence, leadership effectiveness, team dynamics, succession, and burnout risk. Customer and revenue models focus on trust formation, buyer readiness, relationship strength, and customer experience. AI and human-agent models measure communication fit, friction, escalation risk, trust, and interaction outcomes.

Deployment is designed for enterprise constraints. Customers can use the millionways Platform, an API, MCP-based integrations, or self-hosted infrastructure. That flexibility matters because the source material may include interview transcripts, customer conversations, leadership discussions, and other sensitive data that cannot move casually between systems.

Why Behavioral Intelligence Matters Now

AI adoption has created an awkward gap between task completion and human acceptance. A support agent can produce a correct answer while increasing customer resistance. A hiring system can summarize an interview while missing how a candidate makes decisions under uncertainty. A sales workflow can automate follow-up while failing to recognize that trust is falling.

millionways is building for that gap. Its human-agent systems product is meant to show whether an interaction built confidence, created friction, clarified a decision, or should have triggered a human handoff. The strategic question is whether those signals improve outcomes consistently enough to become an operating layer rather than another analytics panel.

The company is already testing the idea across adjacent markets. millionways announced a partnership with Waxell to combine behavioral intelligence with AI governance and agent observability. Hunt Scanlon's investment creates another proving ground in executive search, leadership assessment, and private-equity diligence.

Leadership, Capital, and Market Access

Martin Cordsmeier is the founder and CEO voice behind the company's human-understanding thesis. Max Weidemann is the co-founder and CTO responsible for the technical direction of Thorsten-4. Their roles are current in company and professional sources, although public materials disclose little about the rest of the leadership structure.

The company's capital history also requires careful reading. In 2023, millionways said it had raised $10 million cumulatively through equity, convertible notes, grants, and angel backing, according to AlleyWatch. That was not one clean priced round. The amount and terms of Hunt Scanlon Ventures' 2026 investment were not disclosed.

Hunt Scanlon adds something the earlier capital stack could not guarantee: market access. Executive-search firms and talent leaders can test whether behavioral signals add decision quality beyond interview polish and conventional assessment. Scott A. Scanlon, co-CEO of Hunt Scanlon Ventures, is also advising the co-founders.

The Governance Test

Behavioral inference becomes more valuable as the decision becomes more consequential, and more dangerous for the same reason. A system used in hiring, leadership, care, or finance needs clear boundaries around consent, bias, explainability, and human review.

millionways' current privacy policy says customer-uploaded content is not used to train its models unless a customer separately opts in. Its terms place responsibility for notice, consent, audits, legal compliance, and meaningful human review on customers using the service in employment or other regulated settings. The terms also prohibit protected-class inference and certain automated final decisions.

Those controls are not a side note. They are part of whether behavioral intelligence can become trusted infrastructure. Buyers will need evidence that a model remains useful across roles and populations, that users can challenge its output, and that inference does not become a shortcut around accountable judgment.

What to Watch Next

millionways has product breadth, a differentiated research story, and new routes into executive search and AI governance. It does not yet publish the revenue, named-customer roster, independent validation record, or exact employee count that would make commercial scale easy to judge. LinkedIn lists an 11-50 employee band, but no official careers page or current open-role list was verified for this Spotlight.

The next signal is therefore not a recruiting slogan. It is whether millionways can convert behavioral analysis into repeatable outcomes that enterprises can audit and defend. If it can, Thorsten-4 may become part of the control layer between AI systems and the people expected to trust them. If it cannot, the market will have another sophisticated way to describe behavior without improving the decision that follows.