Operations | Monitoring | ITSM | DevOps | Cloud

Engineering teams are shipping more AI-generated code than ever, but at what cost? Learn how to build a telemetry pipeline to monitor Claude Code usage and costs directly in SquaredUp. It is estimated that 85-90% of engineering teams are now using AI coding assistants such as Claude, Codex and Cursor. This is not just for small-scale pilot projects— around 40% of all code now being shipped is AI-generated, and in start-ups the figure is around 95%. This can result in incredible productivity gains.

Picture this. It's the end of the financial year. Your audit team is clipboard in hand, walking floor to floor, cross-referencing serial numbers against a spreadsheet that was last updated six months ago. Three days in, two people are still checking warehouses, and someone has already found a printer that the system says was disposed of in 2022. This is how most enterprises still run their physical asset audits in 2026.

AI models can reason over language, summarize findings, and explain patterns. What they cannot do on their own is see the real-time operational state of your environment. Ask a model about a critical incident and it will answer from whatever context it is given, which means the answer is only as trustworthy as the input. In operations and compliance workflows, an answer is only useful if it is grounded in current service context and governed access to the systems that define reality.

A majority of office professionals (72%) believe they understand how to use AI for their job better than the team responsible for managing AI at their company. While it’s encouraging to see employees embrace AI with such confidence, organizations will want to ensure they are providing the tools, guidance, and safeguards needed to help employees use AI safely.

Legacy infrastructure was not designed for the requirements of the AI era. While large-scale model training remains centralized in data centers, test-time inference is rapidly shifting to the edge to reduce latency and bandwidth consumption. This shift creates a new frontier for enterprise AI, but deploying at the edge introduces significant manual complexity, interoperability issues, and security vulnerabilities.

Infrastructure and Operations (I&O) teams have long operated under a familiar paradox: the faster the business scales, the more pressure I&O absorbs. Every new application deployment, every endpoint added, and every cloud workload spun up generates more complexity, more risk and more tickets. The traditional responses to this pressure — more headcount, more tooling, more scripts, more APIs — have delivered incremental relief at best.