Across this series we have brought together three management points: purpose, values and boundaries; AI as a socio-technical system that redistributes power and roles; and the governance, risk and learning architecture needed to carry all of that.

Time to capture this in something you can actually use: an AI compass on one page.

Not a technical blueprint, because that belongs to execution, not direction. Not a legally complete document, because that is the work of compliance, not leadership. Not a marketing story, because that convinces nobody who genuinely has to work with it. A clear set of choices, short enough to remember and concrete enough to test against.

Four core questions form the backbone of that compass. What may AI work on for us? Which values and boundaries are non-negotiable? How do we distribute roles, power and accountability, including the authority to override? And how do we learn from what AI sets in motion, in rhythm, in incidents, in ongoing adjustment?

Each question deserves half a page, no more.

The first half page concerns your purpose for AI. Three to five sentences on how it contributes to your purpose and strategy. Two to three sentences on what it never replaces. That sounds short, and that is precisely the point. A compass that is too long does not get read at the moment it matters.

The second half page concerns values and red lines. Three to five values, translated into concrete principles: always a final human judgement, the right to explanation for whoever is assessed. Plus three to five no-go’s, clear enough to recognise without debate when a boundary is being crossed.

The third half page concerns roles and relationships: who carries ultimate accountability, how dissent is protected, how you handle the difference between what the model says and what the professional sees.

The fourth half page concerns governance and learning: who the board-level owner is, how use cases are classified as learning zone or critical zone, what reflection rhythm exists, and what escalation paths are in place when something goes wrong.

The process of building this compass is itself the intervention.

That may be the most important observation of this entire series, and I do not want to tuck it away in a subordinate clause. Assembling this compass is not an administrative exercise you tick off before the real work begins. It is the real work. While making it, you talk about identity: who do we want to be in a world rich with AI? You make implicit assumptions explicit, about trust, about control, about professionalism, about how you handle mistakes. And you show, as management, that you lead in reflection, not only in ambition.

That is how AI becomes an amplifier of maturity within an organisation, rather than an accelerator of alienation. That distinction matters. Alienation arises when technology is introduced without anyone taking the trouble to explain why, and without room to ask questions before it was already a done deal. Maturity arises when an organisation deliberately takes the time to become clear in advance about what it does and does not want, and has that conversation seriously, even when it is uncomfortable.

How do you approach this concretely? Start with a small core group and a short timeframe. Two to three members of the management team, someone from HR, someone from IT, and someone from daily practice who knows what the work actually looks like. Two two-hour sessions are enough for a first version. More people or more time rarely produces a better result, more often a diluted one, because a larger group falls back more quickly on vague wording nobody feels uncomfortable about.

Then present the draft to a mirror from within the organisation. Professionals, the works council, and if possible an external critical friend with no stake in a good story. Ask them directly: what does this make easier, and what does it make more tense? Both answers are valuable. If nothing becomes more tense, you have probably not recorded anything substantial.

And finally, make it formal, but keep questioning it. Adoption by the management team or board gives it the weight it deserves. But plan an evaluation immediately, in six to twelve months: what actually happened, what did the compass set in motion, and what do we sharpen? A compass that is never revised is no longer a living instrument. It is a document that was once well intended.

A reflective close, because that is what this whole series deserves.

If you look back in five years on the choices you are making now about AI, what do you want to be able to say? That you were early and fast? That is a legitimate answer, and for some organisations the right one. Or would you rather be able to say that you stayed true to who you wanted to be, even when the technology moved faster than your own capacity to carry it?

Both answers are honest. The difference is not which answer is best in the abstract. The difference is whether you consciously choose the answer now, or whether others apply it to you, retroactively, later.

Which single step will you take this month, to move closer to the answer you genuinely want to be able to give?

This was the final instalment of a five-part series on implementing AI without losing yourself. The series began with the question most organisations skip, namely what AI does to who you are, before any tool was ever chosen. It ends here, with a one-page instrument that offers no guarantee, but is an honest starting point for the conversation that matters.

Notes for those who wish to read further:

1. Virginia Dignum, Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way (2019, Springer). On translating abstract values into concrete, testable design choices.
2. Amy C. Edmondson, The Fearless Organization (2018, Wiley). On how the process of collective reflection itself contributes to psychological safety and organisational maturity.
3. Donella Meadows, Thinking in Systems (2008, Chelsea Green). On how a compact, clear instrument can have more influence than an extensive but unusable policy document.
4. Chris Argyris & Donald Schön, Organizational Learning: A Theory of Action Perspective (1978, Addison-Wesley). On the importance of repeated evaluation in keeping an instrument alive rather than letting it become a formality.
5. Peter Senge, The Fifth Discipline: The Art and Practice of the Learning Organization (1990, Doubleday). On how organisations build learning capacity by deliberately reflecting on their own assumptions, rather than merely reacting to outcomes.