Solidroad narrows to training simulations after QA took 90% of its effort

The design Solidroad is giving up was a clean one. Its AI plays the customer in a practice session, scores the exchange against criteria such as accuracy and process adherence, and gives targeted feedback ^[7]. The same scorecards could be run against real interactions in customer channels ^[7]. One rubric covered both rehearsal and the live queue.

Hughes put the 90% across engineering, sales, support and customer success ^[1]. So QA cost a lot to sell and to keep customers on, as well as to build. Everything else, training included, shared the remaining tenth or less ^[12]. That means QA took at least nine times training's share of the work ^[13].

Those figures measure what each half cost Solidroad to serve. Whether buyers wanted scoring and training in one purchase is a separate question. Runtimewire's report does not say what happens to existing QA customers, or on what schedule.

A replica built from a screen recording ^[8] shows the software as it looked on the day it was recorded. We'd expect the upkeep on those replicas to grow with how often a customer changes its tools.

The onboarding figure is company-reported, and it comes from one kind of customer ^[3]. BPOs often have to bring large cohorts up to speed, and new support hires have to learn company-specific rules and tools before they handle real cases ^[10]. For a 50% cut to transfer, a buyer's onboarding needs to look like that: many hires at once, all learning one client's systems. A team that hires a few reps at a time is a different workload.

Hughes wrote in a September 22 LinkedIn post that Solidroad's original ambition was a "flight simulator for work" ^[9]. System Simulations takes that idea from what an employee says to what they do ^[8].