At a senior‑engineer dinner hosted by Shift CT O Craft in Toronto, leaders confronted the reality that AI adoption in engineering still lacks a clear financial model. Two years ago companies spent on AI without question; today CFOs demand measurable ROI within a year. The group agreed that token consumption, not headcount, now drives cost, leaving finance blind to actual spend still.

Participants described a “free‑for‑all” phase where contracts were signed but no internal controls existed to track tokens usage. They likened the situation to early cloud adoption before FinOps emerged, noting that metrics to evaluate AI value are still being invented. One engineer suggested establishing a baseline before chasing ROI, hoping future data will map tokens to feature value.

The dinner also surfaced hiring shifts: teams now separate AI‑product engineers from system‑design engineers, emphasizing product leadership alongside architectural rigor. Interview processes are being retooled to prioritize code‑review skills, reflecting that human judgment, not raw output, is the bottleneck. A consensus emerged that AI speeds code creation but also creates “cognitive debt,” demanding new technical‑health roles to manage long‑term maintenance.