Operations | Monitoring | ITSM | DevOps | Cloud

For the past two years, the conversation around AI in software engineering has focused on one thing: productivity. Engineers are shipping faster, writing more code, and completing work in hours that once took days. Every new model promises another leap forward. What gets far less attention is what all that speed demands from the people using it. Guillaume Moigneu, Field CTO at Upsun, has spent the past year watching engineering teams adapt to AI-assisted development.

Most organizations have responded to the rapid growth of AI by creating written policies that define acceptable use. A clear AI policy can establish expectations, assign responsibilities, and help employees understand how AI should and shouldn’t be used. But policy alone can’t provide oversight and is almost impossible to enforce at scale without the right tools.

Most of our runtime security habits were built for deterministic workloads. A service does what its code says: review the code, sign the image, and its behavior is bounded. Agents are different. An agent’s behavior emerges from a model reasoning over whatever lands in its context window, and some of that context comes from places we don’t fully control — a retrieved document, a tool’s output, a user’s prompt.

Your employees aren't trying to create compliance incidents by using AI tools. They're trying to finish their work faster. According to a 2026 BCG survey, 74% of frontline employees now use generative AI every day or a few times a week.

It's Monday afternoon and that feature you've been working on is mostly done. There's just one item still sitting untouched at the bottom of the ticket: "Add monitoring." You know you should. You also know the sprint ends tomorrow, nobody on the team is an observability expert, and figuring out what to measure—let alone how to write the PromQL for it—feels like a project all on its own. So it gets the same treatment it always does: "We'll add it when it breaks.".

Codex pricing runs six tiers, from free to $200 a month, but the sticker price is not your real bill. OpenAI Codex pricing 2026 charges by the token, not the plan, a change that took effect in April. Plus is $20, Pro starts at $100, and everything past that depends on how many files you let the agent read. Most Codex pricing guides hand you a price list and call it done. That is like pricing a taxi ride by the door handle. The meter is what matters, and OpenAI put a real one on Codex this year.

Claude Opus 5 launched July 24, 2026 at $5 per million input tokens and $25 per million output tokens, identical to Opus 4.8. It delivers near Claude Fable 5 performance at half Fable's price and is now the default model on Claude Max. New effort settings let teams trade capability for token savings, which means two teams on identical pricing can now run up very different bills. Finance teams, that last part is your problem. Anthropic has shipped a model that costs exactly what the old one cost.