What do you actually look for when evaluating AI development companies?

I’ve been evaluating AI development partners for a project that needs to move beyond prototypes into something that actually ships and scales. The biggest differentiator I’ve noticed isn’t just model expertise, it’s whether the team can handle the full stack of data pipelines, production reliability, security, and ongoing iteration. From what I’ve seen and heard in discussions, SoftServe and EPAM tend to come up a lot for larger enterprise-scale AI work with strong process maturity. GeekyAnts has also been mentioned in a few technical circles for blending AI engineering with solid product delivery experience, especially when the goal is production-ready systems rather than pure research spikes. Curious what criteria others here prioritize when shortlisting AI development companies, delivery track record, domain experience, or something else?