millionways and Waxell Form Alliance to Bring Behavioral Intelligence to AI Governance and Agent Observability

October 7, 2026 – millionways, the behavioral intelligence layer for the human-agent workforce, recently formed a strategic partnership with Waxell, an AI governance and agent observability company. The alliance was formed to help enterprises govern, observe, and improve AI agents operating in high-stakes human environments.

As enterprises move from copilots to autonomous and semi-autonomous agents, AI systems are being given adult responsibilities before they have adult behavioral judgment. They can speak, reason, recommend, escalate, persuade, and take action, but they do not naturally understand trust, hesitation, confusion, pressure, motivation, emotional response, or when their own behavior is creating risk. In practice, that is where the risk shows up: an agent that keeps pushing when a customer is hesitating, escalates a situation it should have calmed, quietly expands its own access, or grows more confident even as it drifts off course, often with no one watching closely enough to catch it.

Waxell and millionways are partnering to close that gap. The combined solution brings together Waxell’s AI governance and agent observability platform with millionways’ Thorsten-4 Large Psychology Model, adding a behavioral signal layer that helps enterprises understand both sides of the interaction: how the agent behaves and how the human responds.

Waxell provides the governance infrastructure to monitor, manage, and improve AI systems across complex workflows. millionways adds Thorsten-4, a large psychology model that analyzes human and agent communication to detect behavioral signals such as trust formation, stress response, decision friction, emotional regulation, communication fit, motive alignment, and escalation risk.

Together, Waxell and millionways give enterprises a deeper model of AI governance: one that does not stop at whether an agent completed a task, followed a policy, or generated a compliant response. It also evaluates whether the interaction created trust, confusion, hesitation, pressure, alignment, resistance, or escalation risk.

“AI agents are powerful, but they are still behaviorally immature,” said Max Weidemann, founder and CTO of millionways. “They can generate language and take action, but they do not naturally know how they are landing with humans. Thorsten-4 helps humanize agents by giving enterprises visibility into the behavioral dynamics of human-agent interaction: whether an agent is building trust, creating pressure, causing confusion, or helping a person move forward.”

“Waxell gives teams real control over the agents they run in production, from the guardrails agents operate within to the visibility needed to catch issues early and continuously improve performance over time,” said Logan Kelly, Founder and CEO of Waxell. “With millionways, those controls can now be informed by behavioral signals, enabling organizations to guide not only what an agent is allowed to do, but how it engages with people. Together, we’re helping enterprises create agent interactions that are more effective, predictable, and aligned with both business objectives and human expectations.”

AI Trust at Scale

The partnership is designed for enterprises deploying AI agents across customer service, financial services, healthcare, travel, talent, sales, internal operations, and other human-facing workflows where trust, judgment, and communication quality directly affect outcomes.

The combined Waxell and millionways solution is designed to be simple to implement. With only a few lines of code, enterprises can enrich agent interactions with behavioral intelligence, turning transcripts, chats, emails, meetings, and other text-based communication into structured signals that governance, observability, and agent improvement systems can use.

Through the partnership, Waxell and millionways plan to support joint proof-of-concepts with enterprise customers, starting with use cases where AI agents interact with customers, employees, advisors, operators, prospects, or other stakeholders in complex decision environments. These pilots may evaluate agent behavior, human response, communication risk, escalation patterns, and the fit between agent style and human need.

millionways Brings a New Behavioral Intelligence Lever to Executive Recruiters

Hunt Scanlon Ventures has completed its investment in millionways, a New York-headquartered company that builds a behavioral intelligence layer of AI for human agent workforces, utilizing a proprietary Large Psychology Model called Thorsten-4. The platform focuses on understanding human signals, motives, and behavior patterns, with applications used across AI governance, executive hiring, talent management, and enterprise environments.

“millionways is a trailblazer in this newly emerging space,” said Scott A. Scanlon, CEO and co-founder of Hunt Scanlon Ventures and a senior advisor to company co-founders Martin Cordsmeier and Max Weidemann. “As investors, we see behavioral intelligence as a new layer of AI infrastructure, giving artificial intelligence a missing layer – the ability to interpret the human dynamics behind communication and decision-making. In our view, organizations that can better understand the convergence of leadership and human behavior patterns will create meaningful competitive advantage in the years ahead. Behavioral intelligence will become central to everyday enterprise decision-making,” he said.

Mr. Scanlon noted that artificial intelligence is already making big inroads and reshaping the human capital sector. “Organizations have embraced AI to automate workflows, improve performance, and generate insights on people. As those capabilities mature, the next phase is poised to move beyond productivity and efficiency toward improving the quality of human judgement,” he said. “Reducing this friction is an essential next step for businesses in general, but a critical one for the human capital ecosystem.”

Thorsten-4 analyzes unstructured communication, including transcripts, chat, email, meetings, and other text-based interactions, and converts those signals into structured behavioral intelligence. Unlike traditional sentiment analysis or static personality assessments, Thorsten-4 is designed to evaluate the deeper behavioral structure beneath communication, including motives, cognition, affect, regulation, stress response, and operating style.

For enterprise AI deployments, this creates a powerful new signal layer. Instead of only measuring whether an agent completed a task, companies can evaluate whether the interaction produced confidence, resistance, confusion, trust, compliance, or escalation risk. And because that signal lives in the governance layer, it becomes something teams can act on, not just observe. They can adjust how an agent communicates, step in when an interaction starts to go wrong, and tune agents over time so they work better with the people they serve.

These pilots are designed to turn behavioral signal into action: better-calibrated agent communication, earlier intervention when an interaction goes off course, and more trusted outcomes for the people on the other side. Initial joint use cases may include:

Monitoring customer-agent conversations for trust, hesitation, pressure, motive alignment, and escalation signals.
Evaluating whether AI agents are communicating in ways that match user needs, motivation, and emotional state.
Detecting when agent-generated responses create confusion, defensiveness, overconfidence, or misalignment.
Creating behavioral risk scores across AI-assisted workflows.
Improving agent design, prompt strategy, and policy enforcement using human response data.
Supporting enterprise governance teams with a deeper understanding of human-agent interaction quality.
Helping enterprises humanize agents by teaching them how to adapt communication to the people they serve.

Contributed by Scott A. Scanlon, Editor-in-Chief and Dale M. Zupsansky, Executive Editor – Hunt Scanlon Media

Share This Article

0 Comments

RECOMMENDED ARTICLES