PP&A completed ten one-hour interviews in November and December 2024 with technology development leads and IT executives at enterprises of more than ten thousand employees in the United States and Europe, spanning financial services, telecommunications, media, technology, industrial equipment and defense-related engineering. The guide followed the adoption journey: which assistants were in use and for what, how adoption started and scaled, how decisions were made and by whom, what metrics justified the investment, the challenges met at each stage and the tactics that overcame them, the role vendors played, and how leaders saw AI-driven development evolving. The team synthesized the interviews into a report covering utilization, adoption dynamics, evaluation criteria, decision-making, strategic and cultural shifts, challenges and mitigations, vendor roles and future trends.

The interviews confirmed a staged journey. Adoption typically began with developers experimenting on their own, moved to a proof of concept with a small willing team, then to limited and enterprise-wide rollouts, with tools re-evaluated yearly because the market moves so fast. One organization grew from about one hundred licenses in mid-2023 to fifteen hundred developers by the end of 2024; another went from an R&D pilot to three hundred and then two thousand engineers. Active use ranged from roughly a quarter of developers where training was thin to well over half where it was not, highest in modern languages and lowest in legacy and backend systems. Hybrid decision-making, where grassroots experimentation informs an executive mandate, was the most common path, often overseen by a technology governance board spanning engineering, IT, legal and procurement.

Leaders measured value in task completion time, pull requests per day, code coverage, vulnerabilities per commit, defect counts and developer satisfaction. Reported productivity gains ranged from 10 to 15 percent in controlled comparisons to 30 to 55 percent in self-reported estimates, and one telecommunications group moved from quarterly to biweekly releases. Documentation, unit tests, comments, code completion, debugging, language conversion and mentoring junior engineers were the use cases that stuck.

The obstacles were consistent: early skepticism about job displacement and code ownership, resistance from senior engineers, limited context windows and no knowledge of internal frameworks, security and intellectual property exposure, and underuse where training was missing. Successful teams answered with phased rollouts, prompt-engineering training and lunch-and-learns, review processes for AI-generated code, private tenants for sensitive work and communities that share practices. Vendors that offered private instances, structured training, direct feedback channels and partnership pricing earned the deepest adoption. Leaders expected assistants to be ubiquitous by 2030 and were already standing up AI governance teams and reskilling programs. The report gave the client the evidence base for its adoption playbook.

• --