How my development workflow has evolved with AI over 2026.

As we are solidly in the back quarter of the year I want to talk about the radical way that my work has altered over the course of this year. At the start of the year my focus was deep on modelling, using industry knowledge and producing Kimball methodology-based models to be implemented by other parts of the team. AI was introduced as an autocomplete, something that could help me work faster, and it did, allowing me to edit Mermaid diagrams far faster than if I had been creating them by hand and reducing the time needed to express ideas.

Moving on from that we get to the point where advancements in models and reduction in hallucination, along with agents, allow me to leverage AI for research – helping me to understand the limitations and available data of an API and use that as a starting point for discussions around product development that can then better inform the model design.

Next, we get to faster implementation. Taking the diagram as an artifact and target AI can build the SQL to from the landed data to the target data model. It still needs human review, good guide rails and tests, but it’s done far faster than it could be by hand.

Next we get to a point where taking the input from the research and review, combined with the product insight and direction I can use models and rules to generate multiple designs with differing trade-offs and then review and tweak those designs before taking them to the wider team for review and then implementing them.

Every step still has humans in the chain, reviewing and making decisions. Those parts of the work that are now concentrated in human hands are the parts that Dave Farley has stressed as the real work of software engineering. What’s challenging is the implementation is more hands off, team collaboration has changed with more possible to be delivered by fewer people and the decision points are coming faster.

The comfortable work of known problems, churning out that SQL, editing that diagram have been eroded and almost completely removed. Pressure and the level at which everyone operates has changed. Productivity has grown massively, if the level of communication and collaboration can be maintained. Almost ironically it seems that AI is making development more of a human problem.