GRC Leadership in an age of algorithmic decisions

The leaders I work with are noticing something they cannot always name. The work of leadership feels different from the work they were promoted for, and the gap seems to be widening. A recent Harvard Business Review piece helped me put words to part of it. It revisits the seven transitions a leader makes in moving from running a function to running an enterprise, a framework Michael Watkins first set out in 2012, and argues that while the transitions endure, what each one now demands has changed, largely because of generative AI.

I have a soft spot for Watkins. I leaned on The First 90 Days through two of my own larger moves, and have recommended it to more coachees than I can count. So a revisit of his thinking, with the ground now shifting under it, was always going to have my attention.

And one shift he notes stands out most for those of us in governance, risk and compliance. The move Watson calls analyst to integrator has changed the most. For years, the value of a senior leader lay in synthesis: pulling together work and views from across the silos and business areas and making sense of it. AI now generates more analysis than any of us can read, let alone absorb. So the task is no longer to produce the synthesis. It is to design the decision architecture around it: deciding which questions get an algorithmic answer and which still require human judgment, and, harder still, keeping a named person accountable for outcomes that emerge from systems no one fully understands, and that are non-deterministic by nature.

GRC leaders are the people the organisation turns to when it asks who is accountable. So an automation choice is an accountability decision, and a model deployment is an accountability decision. The analysis may be automated. The judgment, and the responsibility, cannot be.

So how to approach it. The value is moving from synthesising the work yourself to deciding which machine-produced options deserve your trust, which to combine, and which to overrule. That is judgment, and it is leadership rather than management: direction and discernment rather than control of the work. It also asks for less certainty than we are used to offering, because the ambiguity is real and will not fully resolve. Feeling less sure is not a failure of competence. It is the honest response to a genuinely harder job.

If the analysis increasingly writes itself, and your value now sits in building and governing the human and AI decision systems around it, are you spending your time there?