An operator's path to the knowledge layer
Most of the noise in artificial intelligence is about the model. Chris Mahl spends his time one layer down, on the messy, unglamorous problem of the data that feeds it.
As President and CEO of Pryon, a Raleigh, North Carolina company with roughly 130 employees, Mahl has staked his reputation on an idea that runs against the grain of the AI hype cycle. The winning question, he argues, is not whose model is smartest. It is whose answers you can actually trust. Pryon builds a retrieval-augmented generation platform - a system that pulls answers directly from an organization's own documents and data, then ties each answer back to the exact source it came from.
That focus did not come from nowhere. Mahl has spent more than two decades inside enterprise software, and the pattern of his career is consistent: show up early, help a company find its footing, and build the go-to-market and operational machinery that turns a promising product into a business. He earned a degree in finance and marketing from Boston College's Carroll School of Management, then went to work in the parts of the industry that were quietly reshaping how companies buy and use software.
He was part of the leadership teams during the formative years of Salesforce and Informatica. Both are now considered market-defining names, but they were not always. Mahl credits those years with teaching him a lesson he still repeats: technology alone does not win. Customer focus does. Helping transform young companies into profitable, customer-centric organizations became the through-line of everything that followed.
"A startup is a ridiculously, wonderfully complicated, exciting experience."
Before Pryon, Mahl held an executive role at Right Media Exchange, a digital advertising marketplace that was eventually acquired by Yahoo. He served as President and Chief Revenue Officer at Botkeeper, an AI-based automated bookkeeping company, where he helped lead growth through multiple funding rounds. He was also Executive Vice President and Chief Revenue Officer at Opentron Labworks. Alongside the operating jobs, he has spent more than 15 years advising and investing in startups and venture capital firms, which gave him a view of the market from both sides of the table.
The CRO-to-CEO move is a meaningful one. A chief revenue officer is measured on growth. A chief executive owns everything - product direction, culture, capital, and the harder question of what the company is actually for. At Pryon, Mahl's answer to that question is unusually specific.
Why data readiness is the real bottleneck
Ask Mahl why so many corporate AI projects stall, and he does not point at the algorithms. He points at the data. In his telling, the hard part of enterprise AI is getting a company's scattered, unstructured information into a state where a model can use it safely.
"Data preparation ranked as the number one most expensive, time consuming and technically challenging part of the build."
The number he returns to is stark. By his estimate, only about 1% of an enterprise's important data is reflected in today's models. Everything else - the contracts, manuals, reports, and institutional memory locked in filing systems and databases - sits outside the reach of the AI systems companies are trying to deploy. Close that gap, and the model gets dramatically more useful. Ignore it, and even the best model produces confident answers built on almost nothing.
This is where Pryon's design choices come from. Mahl describes the company as working "at a more fundamental level of the stack - the knowledge layer," rather than competing to be another broad chatbot interface. The platform ingests multimodal content, prepares it, and makes it retrievable. Crucially, it uses granular data attribution, so every answer can be traced back to its specific source. For regulated industries - government, defense, finance, life sciences, energy - that traceability is not a nice-to-have. It is the difference between a system a compliance team can approve and one it cannot.
Relative emphasis in Mahl's public comments on enterprise AI. Illustrative, based on interviews.
The strategy has attracted capital. Pryon closed a $100M Series B in September 2023, part of a total funding history near $139.5M. That money buys time to solve a problem that does not reward shortcuts. Enterprise buyers do not want a demo that dazzles and then falls apart on their own documents. They want a system that survives contact with real data, real regulators, and real stakes.
Two coasts, one thesis
Pryon's headquarters sit in Raleigh, North Carolina, on Crabtree Boulevard, but Mahl himself is based in New York. That split is a small window into how he operates - comfortable moving between the engineering-heavy world of a technical AI company and the commercial world where deals get done and belief gets built.
Colleagues and interviewers tend to describe him less as a visionary founder-type and more as an operator: someone drawn to the hardest, least glamorous part of a problem because that is usually the part that matters. He talks about passion as a business input, not a slogan - the thing that keeps teams pushing through the complicated, exciting mess he clearly enjoys. It is not an accident that the word he reaches for to describe a startup is "wonderfully complicated."
Early to the giants. He was on the leadership teams of Salesforce and Informatica before either was a household name in enterprise software.
An exit under his belt. Right Media Exchange, where he was an executive, was acquired by Yahoo.
Investor and operator. More than 15 years advising and backing startups and VC firms alongside his day jobs.
Repeat revenue leader. He served as Chief Revenue Officer at both Botkeeper and Opentron before taking the CEO seat.
For all the talk of models and stacks, Mahl's pitch comes back to something plain. AI is only as good as what you feed it, and most enterprises have not yet fed it much. Pryon is his bet that the company which fixes that - quietly, accurately, and with receipts - ends up owning the layer everything else depends on.