Could your company produce a complete inventory of every AI tool, model, agent and embedded feature operating across the business today?

In this episode of Business Tech Perspectives, I speak with Russ Fradin, founder and CEO of Larridin, and Michael Levine, partner at Hunton Andrews Kurth, about the enterprise AI visibility gap and the growing legal, financial and insurance consequences for companies that cannot identify their AI systems.

Russ has experienced several major periods of technology adoption across digital advertising, mobile and workforce communication. He launched Larridin after recognizing that standard measurement systems could not account for the speed and variety of AI tools entering businesses. The company helps enterprises understand which AI systems employees use, what they cost, which models they rely on and where adoption is taking place.

The scale of the problem can be surprising. Russ describes an early customer that expected to find approximately 70 AI tools operating across the company but discovered 220. Most were benign, including five separate AI meeting assistants, but the discovery revealed unnecessary spending and fragmented adoption.

Another customer found an AI agent built by a former employee that continued running at a cost of $1,800 each month. Nobody remaining at the company understood what the agent did. The example shows why an enterprise AI inventory concerns business spending and accountability alongside security and compliance.

Mike explains how embedded AI can expose a company to legal action without leaders realizing that AI was involved. He discusses a class action concerning an HR screening tool that allegedly used AI-assisted lie detection without the required disclosure.

We also examine the growing patchwork of US AI regulation. Mike discusses the Texas Responsible Artificial Intelligence Governance Act and the Colorado AI Act, along with the difficulties faced by businesses operating across several states. A web-based company may need to comply with a state law because it serves residents there, even when the company is based elsewhere.

Board accountability is another focus. Russ says boards increasingly want to know which tools and models are being used, what employees use them for, what data agreements apply and how much the company is spending. Mike adds that inaccurate statements about AI capabilities can create directors and officers exposure through shareholder litigation.

Insurance adds further complexity. Mike explains why traditional cyber coverage may not apply to many AI claims because they do not involve a data breach or loss of protected information. Insurers are beginning to introduce AI-related exclusions across several forms of commercial coverage, making the wording and definition of AI increasingly important during renewal.

Russ and Mike close with practical guidance for companies building a verified AI inventory. IT, finance, legal, HR, operations, risk teams and board members must contribute because no single department has a complete view of AI use across the business.

Could your leadership team identify every AI system, its owner, its cost, the information it touches and the insurance protection available if something goes wrong? Please share your thoughts with me.