From headcount to outcomes
The traditional IT services pitch was straightforward: the same work, done for less, by a larger pool of engineers in a lower-cost location. That model is under real pressure — not because the cost advantage disappeared, but because AI-assisted tooling has made raw implementation hours cheaper for everyone, everywhere.
What's left as the differentiator is judgment: knowing which problem is worth solving, which architecture will hold up in three years, and which shortcut will quietly become next year's incident. That's harder to commoditize, and it's what clients are actually paying for now.
What AI changed, and what it didn't
AI tooling has made a real dent in the time it takes to write boilerplate, scaffold a service, or migrate a codebase. It hasn't changed the time it takes to understand a client's actual business process, negotiate scope, or decide what not to build.
Services firms that only sold implementation hours are the ones most exposed to that shift. Firms that sold — and delivered — engineering judgment are less exposed, because that part of the job hasn't gotten any easier to automate.
What we think clients should actually evaluate
Fewer clients are asking 'how many engineers can you put on this' and more are asking 'will this still work in two years, and can someone else maintain it if you're not around.' Those are better questions. We'd rather be evaluated on documentation quality, handover process and long-term maintainability than on headcount or hourly rate.
Where we sit in this
We're a small engineering team by design, not by constraint — we take on work we can do well and say no to work we can't. That's a different model from traditional outsourcing, and it's the one we think holds up as the industry keeps changing.