Skip to main content

Insights

Answers to the questions that come up before a project starts

Written from work we have actually done rather than from general practice. Each piece answers one question properly, including the cases where the honest answer is that the work is not worth doing.

For material you can work directly from, see our reference resources.

  • Artificial Intelligence & Automation · 9 min read

    Why AI prototypes stall before production, and what production actually requires

    The model is rarely the problem. What separates a prototype from a production system is the work nobody demonstrates: explicit interface contracts, defined behavior when a dependency misbehaves, observability, an evaluation set you can rerun, and a rollout that can be reversed. That work is unglamorous, it is usually scoped last, and it is the entire difference.

    Read the full piece

  • Artificial Intelligence & Automation · 8 min read

    When workflow automation is worth engineering, and when it is not

    Automation has a permanent maintenance cost, so it has to earn its place. The test is volume, repeatability and what an error costs. The harder question underneath it is which decisions in a process are genuine judgement and which have simply never been written down, because only the second kind can be automated safely.

    Read the full piece

  • Technology & Engineering · 8 min read

    Do you need data engineering, or do you need one spreadsheet to stop being retyped?

    Written for the person who has to sign it off rather than the person who would build it. Most teams who ask about data engineering are describing a symptom: the monthly report takes three days and the numbers disagree. Sometimes that is a pipeline problem. More often it is two systems that were never connected, or one report nobody has questioned in two years. The three are not the same purchase, and buying the wrong one is expensive in a way that is hard to unwind.

    Read the full piece

  • Artificial Intelligence & Automation · 9 min read

    What AI integration actually costs, and the part that is missing from every quote

    Almost every AI integration quote prices the build and stops there. The costs that decide whether the thing was worth doing arrive later: inference that scales with use rather than with headcount, integration work against systems that change without warning, and retraining nobody scheduled. Published figures put the 24 month total at 1.6 to 2.2 times the build cost. The gap is not padding, it is the part of the work that was never quoted.

    Read the full piece

Talk to us about it

Book a call if you would rather talk it through, or send a few lines and we will reply. Either is fine, and neither commits you to anything.

Book a call or send a message