Intelligence
AI prepares. People decide.
AI becomes a capability of the company, with clear rules on who decides. And the official path is the easiest one, otherwise shadow AI appears.
Inside the letter I
Intelligence treats AI like electricity: as infrastructure the whole company runs on, and not as a feature somebody installs on their laptop. It works in two circles. The first circle makes the organisation itself more able to work: information is findable, data reliable, processes automated where it pays off. The second circle accelerates value creation: AI in products, services and delivery. The same limit applies to both. AI prepares, the human decides, and every dependency has a documented exit.
The trap
When everyone uses their own tool, you get fragmented knowledge, workflows nobody knows about, dependencies nobody sees, and a slow erosion of judgement, because people trust the output instead of checking it. The reflex of many companies makes it worse: a central body that wants to approve every tool. Then people build around it all the more.
Mechanics
- 01The Intelligence Steward sets the AI guardrails: AI prepares, people decide; autonomy levels per use case; an inventory in which every use case is listed.
- 02Three stages, chosen explicitly per task: assist, augment, autonomous. The higher the stage, the closer the review.
- 03Bias watch as a guardrail: automation bias, confirmation bias and status quo bias are actively checked for every use case. Whoever trusts the model blindly makes its mistakes their own.
- 04Anchor Core Jobs for Digital Workplace, data, AI in the organisation and AI in value creation. Outcome accountability for the product stays in F; I delivers the capability.
Rooted in research
Each source with a short sentence on what it actually says.
- Shneiderman (2022), Human-Centered AI
High automation and high human control are not a contradiction. Good systems give both: strong tools and clear control.
What this does not settle: design principles, not a finished operating model.
- Brynjolfsson & McAfee (2014), The Second Machine Age
The biggest productivity gains arise when human and machine do what each does best in the same workflow.
What this does not settle: a macro argument; it does not say what the roles inside the company look like.
- Parasuraman & Manzey (2010), Automation Bias
People trust automated results more than is justified, especially under time pressure. Without a built-in check, mistakes of the model become mistakes of the team.
What this does not settle: awareness alone does not help; it takes structural review steps.
- Note: FLAIMS-specific constructs such as Steward, Anchor Core Job, Core Job, Joblet, FlowCoach, Gravity Decision Model and the learning chain are developments of the model out of practice, not cited research.
In practice
AI stops being a private hack and becomes a capability of the whole company. For every use case you know which stage applies, who decides and how to get out again.