Flux adds AI engineering metrics to show real returns

Flux has expanded its engineering intelligence platform with five features aimed at measuring the return on AI-assisted software development. The additions are intended to address gaps in how engineering teams track work and spending.

The features - verified velocity, trusted review, auditable work, continuous quality and defensible spend - are designed to help engineering leaders assess whether greater use of AI tools is producing measurable outcomes in codebases, rather than simply increasing activity in tickets, story points or adoption figures.

The Boston-based company argues that widespread AI use in software development has exposed weaknesses in traditional measurement systems. Engineering teams, it says, often rely on project management data that reflects intent or workflow activity while missing what actually changed in the code.

The expansion targets five areas Flux describes as blind spots: velocity theatre, review debt, hidden work, quality drift and unproven spend. Each, according to the company, can be tracked directly through code analysis without requiring new instrumentation or changes to engineering processes.

Verified velocity classifies merged code changes as feature work, maintenance, bug fixes or refactoring, then compares deployment frequency and lead time with an organisation's historical baseline. Trusted review measures time to first review, reviewer availability and review load distribution, while also assessing review depth against the size and risk of code changes.

Auditable work uses commits and pull request histories to show code activity, regardless of whether ticketing systems fully capture it. Continuous quality tracks new and resolved security findings, newly introduced dependencies, and failure and recovery trends against a team's norms.

Defensible spend is aimed at finance and leadership teams seeking evidence for capitalised development costs, operational expenditure and research and development tax claims. It classifies work by whether it is capitalisable or operational, and whether it is tied to features or maintenance, rather than measuring AI spending through token use or licence adoption alone.

Flux tied the launch to growing scrutiny from senior executives and boards over AI spending in engineering. As AI coding tools become more widely used, technology leaders face pressure to show whether those investments are improving delivery, code quality and financial outcomes.

Research cited by Flux from DORA found that 90% of software professionals use AI daily, up from 76% a year earlier. The company said the trend has widened the gap between organisations with strong development practices and those struggling to maintain visibility over software work.

"AI is the biggest bet most engineering organizations have ever made, and the blind spots undermining adoption hide inside tools built for a different era of software development," said Ted Julian, Chief Executive Officer, Flux.

"Tickets and adoption rates describe intent. The code itself shows what the team actually built and delivered," Julian said.

Aaron Beals, Chief Technology Officer at Flux, said conventional tracking methods have become less reliable as AI makes it easier for developers to complete larger pieces of work in less time.

"AI made these problems move faster than the old measurement systems can keep up with. A senior engineer with an AI assistant can turn a two-week refactor into an afternoon project, and that's just one example from one team. Across the org, teams are shipping far more real code updates than they used to. Ticket-based tracking was already a rough approximation before AI, and now what a ticket says happened and what the codebase shows have drifted far enough apart to create serious blind spots.

"Unproven spend and quality drift both trace back to the same gap. One is Finance, in the quarterly review, asking what the AI spend actually bought; the other is the on-call engineer paged for a dependency change nobody reviewed. Analyzing the codebase itself closes that gap for both questions without slowing anything down," Beals said.

Julian said financial accountability had become a central issue in customer discussions.

"Every engineering leader we talk to has been doubling down on AI: more tooling and more code moving through the pipeline. In nearly every enterprise conversation, the first questions from senior leadership are about spend. Can you show CapEx versus OpEx? Can you help substantiate an R&D credit?

"That pressure is why unproven spend has become a bigger story now that adoption is a given. But the same organizations that can't produce that proof usually can't see quality drift in their code either, because both gaps come from the same root cause: measuring activity instead of analyzing what the code actually shows," Julian said.

One customer, Cobalt, said the system had provided material to support financial reviews.

"I've spent the last year trying to prove where our engineering budget actually goes," said Gunter Ollmann, Chief Technology Officer, Cobalt, the leader of human-led, AI-powered offensive security.

"Now I have evidence I can hand to finance for R&D credit substantiation, and to the board when they ask if these bets are paying off," Ollmann said.