My thoughts and advises about AI deployment, are generally built around these six positions

Published 22 October 2026

I wanted to write down my stronger (controversial?) positions around AI deployment.

Generally, these six positions make the foundations of my work and advice:

  1. There is only one metric that matters in an AI deployment and it has nothing to do with AI: Sustainable Long Term Profit. All AI usage, token usage metrics are useless. Focus only on improving profitability and AI is one of many tools
  2. Process understanding is more important than good data quality. AI can turn messy email, Slack or WhatsApp into structured data. Clear processes are what make agents easy to deploy.
  3. Automation is more important than AI agents. If a repetitive process can be written as if-statements, a cron job or Airflow schedule is cheaper and easier to control than an agent.
  4. Buy is better than make. Unless a company has a culture/team that can build, maintain, monitor and update agents, it should buy from a specialist provider.
  5. Forward Deployment Engineers do not have the best interest of your company. The vendor's engineer is there to implement the vendor's product so don’t involve them in strategic decisions.
  6. The independent forward deployment engineer does not have the best interest of your business. He will still try to install tech or AI when the best option may be to remove or maintain the current process.