HAMMERSMITH, U.K. - s4story -- Not a Black Box, an early-stage company founded by HR-technology veteran Bill Fischer, believes it has identified the architectural reason and filed a UK patent application aimed squarely at the fix.

The bottleneck, the company argues, is not the AI. It is the substrate the AI is being asked to operate on: enterprise software is built for humans clicking buttons rather than for machines reading structure. Today's announcement covers a filing with the UK Intellectual Property Office on what NABB calls the Dual-Representation Workflow Engine: a way of holding every enterprise record in two synchronized forms at the same time, one written for human operators and one written for AI agents.

The decision to express the system of record as code, rather than as a conventional database with workflow screens on top, is itself a strategic bet on where the AI laboratories are spending their research budget. Anthropic, OpenAI and Google have collectively poured extraordinary resources into making their models exceptionally good at one specific thing: reading and writing code. That sustained investment is what produced the productivity asymmetry described above, and there is no indication it will slow.

More on S For Story

• Sylvester Anthony III Introduces His Artist Journey with Debut Single "Cherish"
• Rock 'n' Roll Wiener Dog Goes Global: "Longfellow Finds A Home" in a Japanese–English Edition
• Boston Industrial Solutions Introduces High-Performance Primer for Bonding Liquid Silicone to Epoxy
• Healthcare Leaders Publish New Integrated Behavioral Healthcare Guide, Led by Doctors of Behavioral Health
• Verbica Challenges Panetta to a Televised Debate on the Issues

"The frontier laboratories have already decided what they are going to make their models excellent at," Fischer said. "Rather than coax those models into doing something they were never optimised for, we built a substrate they intuitively know how to navigate. Every pound Anthropic or OpenAI spends improving how models reason about code is a pound of free improvement to the way agents reason about the data inside our system."

A category going headless

The filing also lands at a moment of structural transition for enterprise software more broadly. In recent months, several established applicant tracking systems and HR platforms have begun publishing what are known as Model Context Protocol layers  (open standards that allow third-party AI agents to plug into a platform, request data, and execute actions on a user's behalf).

What is at stake

The first commercial implementation of the dual-representation pattern is TalentClaw OS, an agentic operating system for the modern HR function. NABB's product covers the full breadth of human-capital management: core employee records and organisational structure, and the talent lifecycle from hiring through offboarding. The remainder of the HR stack runs on the dual-representation substrate.

More on S For Story

• Salt Lake City Families Turn to Private Autopsy Services for Faster Answers After Unexpected Loss
• Veteran Sportscaster Takes Readers behind the Mic in New Book
• K2 Integrity's U.S. and EMEA Teams Again Recognized in Chambers and Partners 2026 Guides
• Connecta Satellite Solutions Ready to Support Emergency Communications Following the Venezuela Earthquake
• Rare 1933 Harold Weston painting will be auctioned to benefit the Keene Valley Library

Not a Black Box, Inc. is building an Agentic OS for Talent Operations. Rewiring the $650B global HR ecosystem - with a human and machine legible substrate.  TalentClawOS facilitates claw to claw interactions that power the workflows.  It is a Delaware corporation.

For more information, visit http://www.notablackbox.ai