Article:

The AI governance triad: balancing management, accountability and innovation

Written by Dr Ibrahim Peerzada CMgr FCMI Wednesday 08 July 2026
As AI transforms the way organisations operate, leaders must move beyond speed and efficiency to build governance systems that protect trust, accountability and long-term value
Dr Ibrahim Peerzada CMgr FCMI

Let’s be honest with ourselves: as managers, our default setting is optimisation. We look at AI and our eyes light up at the prospect of automated workflows, predictive pipelines and autonomous agents executing tasks while we sleep.

But if you try to manage AI the same way you manage a traditional software rollout, you are stepping into an enterprise minefield.

Management is about driving speed, value and execution. Governance is about guardrails, ethics and risk mitigation. If you drive a sports car at 150mph without brakes, you aren’t an efficient driver; you are a liability.

 

Artificial Intelligence; Real Leadership

Our new report reveals that a striking 70% of managers seek advice from generative AI, rather than going to their managers for guidance. It reveals that the biggest barriers to successful AI adoption are not technical, but human, with a significant capability gap preventing many organisations from translating ambition into measurable business impact.

Find out more

 

Industry data from McKinsey shows that public trust in AI technologies has slipped significantly in recent years, dropping from 61% down to 53%. Consumers, regulators and boards are demanding proof that our systems are safe. With modern regulations like the Colorado AI Act and the EU AI Act – which hands down penalties up to €35m or 7% of global turnover for violations – the old tech mantra of “move fast and break things” is officially dead.

1. The core tension: management v governance

The breakdown usually happens because organisations treat management and governance as enemies. Management wants to democratise AI, giving every department access to foundation models to boost throughput. Governance, often isolated in legal or compliance silos, wants to restrict access until every imaginable edge case is documented.

When these two functions operate adversarially, you get one of two disasters:

  1. Bureaucratic gridlock: innovation stalls entirely because approval pipelines take six months.
  2. Shadow AI: frustrated teams bypass the rules completely, uploading proprietary company data into unvetted third-party platforms, opening the door to massive regulatory exposure.

To bridge this gap, managers must understand that traditional IT governance cannot handle the unique nature of AI.

 

Attribute Traditional IT governance Modern AI governance
System behaviour Predictable and deterministic. Input X always equals Output Y. Probabilistic and dynamic. Outputs drift based on data training and prompts.
Audit cadence Static, point-in-time reviews (e.g. annual security audits). Continuous control loops operating in real time as models adapt.
Risk metrics System uptime, data access logs, encryption status. Algorithmic bias, explainability, hallucination rates and model drift.
Ownership Confined strictly to IT and security teams. Cross-functional councils spanning legal, data and business owners.

Keep reading – more from Ibrahim

Login

If you are already registered as a CMI Friend, Subscriber or Member, just login to view this article.

Confirm your registration

Login below to confirm your details and access this article.

Log in

Register for Free Access

Not yet a Member, Subscriber or Friend? Register as a CMI Friend for free, and get access to this and many other exclusive resources, as well as weekly updates straight to your inbox.