Tenex on AI-native SDLC practices at Fortune 2000 companies

Tenex is an AI application services firm serving large enterprises. It has 75 people, about 90% of them engineers or AI strategists, and its clients are split evenly between Fortune 2000 companies and mid-sized firms with $100M–$500M in revenue. The company keeps paying per story point: pay among engineers varies by roughly 4x, not the 10x originally envisioned. CJ Hess says that as models got stronger, he writes less code directly and spends more time shaping client context into a form models can use; rewards also need to cover work like architecture design and requirements judgment that is hard to break into tickets. The technical core is MetaHarness: a shared set of skills, project docs, and command-line tools, with plans and decisions committed into Git alongside code. When an agent returns to a task, it can read why the team built things that way, not just what was built. Tenex's AI-native SDLC article frames this as a project record that humans and agents reference together, wired into Linear, GitHub, and client update views. CJ also notes the cost: for a small team doing just one project, this level of structure may be overkill. The podcast also discusses the real cost of building software in-house: when replacing a $43K-a-year speaker management tool with agents, the option picked by human judges ranked only 19th on the AI judges' leaderboard. Alex warns that the further you go into large enterprises, the higher the cost and risk of replacing existing systems, and that anything built in-house must be maintained by someone long term. For teams driving agentic coding inside enterprises, the practical value is turning plans, acceptance criteria, and decisions into repo assets agents can read — while calculating maintenance costs up front.