Skip to content

Writing

Essays on leading technology teams and introducing AI into their work: what I’ve learned, and questions I’m still working through.

  1. Agent permissions after the OpenAI and Hugging Face incident

    How OpenAI’s agents got into Hugging Face’s systems through a shared service, and what the case says about tool permissions, scoring and accountability.
  2. Delegating to managers at MED-EL

    What I learned about giving managers responsibility at MED-EL, and why I regret waiting too long when one appointment wasn’t working.
  3. Popular AI tools: What about data protection?

    I compared the terms behind 38 AI tool tiers to help teams assess their use with personal, confidential and secrecy-protected data. Every assessment links to its source.
  4. Supporting engineering teams through an agentic AI rollout

    Notes from an agentic AI rollout across hundreds of engineers: the teams that adapted quickly, the one that fell behind, and what I would change.
  5. Salesforce Headless 360 and the value of a SaaS interface

    Salesforce’s Headless 360 makes more of the platform available to agents. That raises a question for SaaS: what value remains when customers do less of their work through your screens?
  6. GenAI in the workplace: What 51 successful deployments show

    What research on enterprise AI suggests about oversight, data and adoption, and how it connects to the systems I once had to take over in Corporate IT.
  7. The AI production gap: Why enterprises stall between strategy and systems

    I keep meeting teams with an AI strategy and working prototypes, but no clear route into production. Specifications, evaluation and operational ownership need attention earlier.
  8. A constitution you can run: Building governable agents for real work

    Funding matching shows why an agent needs explicit rules, evidence and a way to handle missing information. Model behavior is only part of that design.
  9. Agentic AI and the future of SaaS business models

    When agents do more of the work, user counts tell us less about software value. Pricing still has to cover costs, define a useful outcome and give buyers a budget they can manage.
  10. Show me the code: Keeping our judgment when we use GenAI

    A sticker on my laptop says: nice story, now show me the code. It is a useful reminder to check AI-assisted work and stay able to explain it.
  11. Building Flex.Insight, beating enshittification

    A dinner with Cory Doctorow and Roger Dingledine made me think about consulting, workplace bureaucracy and the practice we were building.
  12. Leading IT into the Frontier Firm era

    Microsoft’s Frontier Firm idea, considered alongside my own experience of AI adoption, hesitant users and the work needed after deployment.
  13. Designing an IT stack for political uncertainty

    Data location, supplier access and exit options belong in an architecture decision. Political uncertainty makes it worth checking how those options would work in practice.
  14. AI teammates in ITSM: What changes for the service desk

    I started on a helpdesk. AI can take on some of that routine work, but service desks still need to develop their people and take care over handovers.
  15. Why we founded Flex.Insight

    What Stefan and I wanted to build at Flex.Insight, and the tension between wanting to try new technology and being responsible for its use.