Generative AI is everywhere and it’s tempting to reach for it whenever something feels messy, slow, or frustrating.
But when a tool is this powerful – and this non-deterministic – the real question isn’t “Can we use GenAI?” It’s “Should we?”
Used well, GenAI boosts productivity. Used indiscriminately, it quietly introduces risk.
This is where GenAI stops being just a productivity tool and starts becoming a governance challenge.
What is it?
GenAI is exceptionally good at language, synthesis, and pattern-matching. What it is not good at is replacing deterministic systems, disciplined data management, or clear accountability.
I see GenAI proposed for problems that are actually different types of issues:
- System integration problems: When organizations have multiple sources of truth, broken interfaces, or inconsistent identifiers, GenAI is sometimes dropped “in the middle” to smooth things over. GenAI can sit on top of integrated systems very effectively — but is it a substitute for integration itself? Not really.
- Data quality issues disguised as AI opportunities: Using GenAI to clean up reports or “fix” inconsistencies often hides upstream data problems. You may get nicer output, but it’s now built on non-deterministic coherence rather than reliable data.
In all cases, a non-deterministic model is being applied to a problem that really needs determinism, structure, or better foundations and should be pushed back downstream to fix.
What does it mean from a business perspective?
This isn’t just a technical concern, it has real business implications:
- Hidden risk: GenAI outputs are probabilistic, not guaranteed to be correct – even when they sound confident. This can introduce subtle errors into decisions, reports, or customer-facing material that may go unnoticed until something goes wrong.
- Loss of traceability: When GenAI influences outcomes, it can be difficult to explain how or why a specific result was produced. This creates challenges for auditability, compliance, and executive accountability.
- Operational fragility: Model behaviour can change over time due to updates or data changes, leading to inconsistent outcomes. Businesses may experience variability without clear signals that anything has changed.
- False efficiency: While GenAI can create quick wins, it can also paper over deeper structural problems. Over time, unresolved data quality or integration issues compound and become harder and more expensive to fix (technical debt anyone).
- Governance gaps: Traditional control frameworks weren’t designed for non-deterministic systems. Without clear rules, ownership, and oversight, GenAI can influence decisions in ways the organization hasn’t explicitly agreed to.
In short: applying GenAI everywhere doesn’t reduce complexity it just moves it somewhere harder to see.
What do I do with it?
Practical steps matter, as well as enthusiasm:
- Classify problems before choosing tools: Take a moment to identify the true nature of the problem you’re trying to solve. Many “AI opportunities” turn out to be data, process, or integration issues that need a different solution.
- Use GenAI where uncertainty is acceptable: GenAI works best when outputs are advisory rather than authoritative (or require human oversight to complete).
- Put guardrails around usage: Be explicit about where GenAI can be used and where it cannot. This includes defining approval points, human review requirements, and boundaries for automated decision-making.
- Treat GenAI as a governance topic, not just a productivity win: Assign ownership, define accountability, and integrate GenAI into existing risk and governance frameworks. This shifts AI adoption from experimentation to sustainable, enterprise-ready use.
GenAI is a big hammer – and it’s an impressive one. But mature organisations don’t swing their biggest tools everywhere just because they can. They build judgement, structure, and governance around how those tools are used.
If you’re serious about scaling GenAI responsibly, the real work isn’t adopting the technology – it’s deciding where not to apply it.
I’d love to hear where you’ve seen GenAI used well – and where it clearly wasn’t the right tool.
#GenerativeAI #AIGovernance #ResponsibleAI #DigitalTransformation #EnterpriseAI #DataStrategy #AIAdoption #TechLeadership