With over 20 years' experience in the industry, I have witnessed AI transform from an unrealistic concept to a business non-negotiable. Time after time, businesses are influenced to buy into several ‘autonomous’ AI tools and when it comes to implementation, ‘speed’ and ‘automation’ are always the first words that we hear. I call this trend ‘AI mushrooming’: implementing multiple AI systems with no clear objective other than gathering as many tools as possible so the ‘toolbox’ looks impressive!
However, this sudden wave of urgency to implement can be incredibly expensive, especially from a risk and compliance point of view. A tool (or multiple) meant for transformation can become parasitic, draining your budget and leaving your business vulnerable from a risk and security perspective. Hence, experts in the field are treading with extreme caution.
At an executive level, it is imperative to understand that we must approach the AI conversation with control and clarity at the forefront. We must ask probing questions about these tools BEFORE implementing them on a company-wide level. For example:
- Who owns these agents?
- What decisions are they making?
- What data are they using and how do we evidence that?
- And how do we remain confident that outcomes are accurate, explainable and defensible?
It is important to remember that while these tools can be transformative, we are not yet at a place where they can be reliably autonomous. This is not a flaw of AI, but rather a limitation of complexity. If we automate already complex processes without addressing the underlying design, we risk making them harder to understand and harder to govern. It’s like asking a machine without any information on general relativity or mathematics to build a rocket fit for space travel!
The way I advise our clients is the same way that I would advise other businesses: keep it simple and specific. Once an AI agent is given guidance and parameters on the task at hand, these tools can be exceptionally powerful. Before automation, take a step back and think about what can be simplified, standardised or removed. A huge amount of risk actually stems from ambiguity. For example: handoffs, exceptions and uncertainty around who is supposed to make what decisions.
The very best AI agents are not general purpose or ‘fully autonomous’. They will be narrowly scoped, have a clear business owner, operate within defined controls and produce outputs that can be audited and explained to regulators. In this way, AI becomes a way to strengthen governance rather than weaken it. AI can improve consistency, reduce manual error and increase transparency, provided it’s applied deliberately.
Deploying multiple AI tools for the sake of it isn’t a step towards transformation, it’s a step off a cliff. Using AI in a clear, controlled manner, not only protects your business, it will pave the way for sustainable growth, delivering genuine results and tangible benefits.
