Want a quick email when a new article is published? Subscribe
For boards and investors
Whether agent-built software replaces a SaaS product depends on what that product actually sells. Three categories, and what each one means if you're writing cheques.
Latest articles
The companies that designed for mobile beat the ones that bolted it on. AI agents are the same shift, and five questions tell you which of your tools are ready.
Jira, Confluence and Atlas each held a slice of the picture at a different level of detail, so every week someone rebuilt the whole thing in PowerPoint.
Your strongest engineers lose Friday afternoons to reports that people skim at best. Transparency should come out of the work, not sit on top of it as a second job.
Building the product was the easy part. Why giving most of it away is a commercial decision, not the absence of one.
The complete operating model I'd bring to a product and engineering organisation: squad structure, every role defined, sprint and release process, DevOps principles, product tiers, AI use cases, and how the whole thing flexes down for smaller headcounts. Twelve documents, written to be used rather than admired.
The cost of an AI feature is set by its design, not the price of the model. A pilot can't show you the difference, and the pilot is when the design gets fixed.
The engineers most resistant to AI coding tools aren't who you'd expect. And the gap between those using them properly and those refusing to look is widening at a pace that should worry anyone planning a career in this profession.
Stop using AI as autocomplete. Start using it to build systems. Here's the practical, no-shortcuts approach.
Your team's internal tools backlog is the perfect low-stakes proving ground for AI development. Here's why, and how to make it happen.