Using AI is not the same as being an AI bet
Last month AIQ Global’s principal software engineer, Chuck Spencer, published a candid account of how the company built “Ziggy”, an internal AI system that gives its engineers the benefits of modern language models without handing patient data, regulatory material or trade secrets to a third party. The architecture is unglamorous and, we think, exactly right: open-source models running on AIQ’s own hardware for anything sensitive, with no per-token cost; a gatekeeper layer that lets a commercial model (Anthropic’s Claude, in their case) see only pre-filtered, metadata-tagged queries; a vector index built nightly from Jira, Notion and GitHub, with tiered trust rules so that a recent code commit outweighs an old meeting note; and dozens of validated test queries to keep the whole thing grounded.
Spencer’s conclusion is the line worth repeating: “Real solutions are well within the grasp of any decent engineering team, now.” The open-source tooling, he notes, is “robust and hugely useful”, and the hardware demands were far less than feared.
Why does this matter to an early-stage investor? Because it illustrates a distinction that the current market tends to blur. AIQ is a company that uses AI. Its product, TRAQinform IQ, gives oncologists quantitative intelligence on complex, costly patients, and earlier this year it secured a new AMA Category III CPT code, the start of a reimbursement pathway. The business is built on clinical evidence, regulatory clearance and reimbursement. AI is a tool inside it, and Ziggy shows the team treats it as a tool: something to be run cheaply, privately and under their own control.
Contrast that with a company whose value is a bet on AI itself: on model capability continuing to compound, on compute prices falling, on the willingness of customers to pay for intelligence as a line item. Those companies may do very well. But their value rises and falls with the AI cycle, and when 87% of venture dollars are chasing that cycle, as Bloomberg reported last month, they are also where the crowd is.
The test we apply is simple. If the price of frontier models rose tenfold tomorrow, or fell by 90%, what would happen to this company? For AIQ the answer, on Ziggy’s evidence, is “not much”: the sensitive work already runs on open-source models it owns, and the commercial model sits behind a gate that can be swapped. The product’s value to an oncologist comes from what it tells them about a patient, not from which model drew the inference.
That is the kind of AI exposure we like: companies that get the productivity gain from the technology while their value rests on something the AI cycle cannot take away, whether a clinical outcome, a regulatory clearance, a reimbursement code or a physical product. It is a less exciting story than being an AI company. It is, we think, a more durable one.
Read AIQ’s piece: https://www.aiq-solutions.com/building-ziggy-how-we-got-real-value-out-of-ai-without-handing-over-our-data/
Albion Trinity Capital is an investor in AIQ Global. This article is for information only and is not an offer or solicitation of any investment.