Milk and AI have nothing in common, right? That’s what you would think. However, in a recent interview with the BBC, Brad Smith, president of Microsoft, one of the world’s biggest AI powerhouses, warned that AI currently faces fewer safety requirements than milk, writes Chris Hawley, managing director for Great Britain at the Noledge Group
It’s a surprising comparison, but it makes a complicated debate more relatable. And it is not a niche concern: more than one-third of UK businesses are now using AI. If we take care to keep milk safe and prevent it from spoiling, we should be just as thoughtful about how we use technology that can affect businesses and society in general.
There is a lot of talk about AI, but the technology itself shouldn’t be the focus. I see many businesses trying to scale too quickly with a heavy reliance on AI, only to then have to retrofit and work backwards. In fact, according to Gartner, at least 50% of generative AI projects were abandoned in 2025 due to poor data quality, inadequate risk controls, escalating costs or unclear business value.
When I talk to businesses about AI, I start with the outcome: Could it simplify a process, reduce costs or support growth? Where are people spending time on manual work and what would help them serve their customers better? In my opinion, that is the right place to start. Finance/ERP systems, for example, store a lot of information relating to orders, stock, finance, production and customer activity. But how hard is that data working for your business? AI can ask why a number has fallen, spot an anomaly, track a change or identify ways to improve performance.
None of that happens simply because a business ‘has AI’. The system needs access to the right information. Used without clear boundaries, AI can produce misleading conclusions or expose information that should not be shared. Businesses need to define what data AI can access and ensure someone is overseeing and sense-checking everything. The human element has not become obsolete; it’s more important than ever. And the prompts need to be carefully constructed to make sure you get what you ask for. The old adage of ‘be careful what you ask for’ has never been more apt.
In his BBC interview, Brad Smith called for global safety rules, independent evaluation, incident reports and an emergency brake. These are important questions for governments and Big Tech, but they matter to ordinary businesses too. Organisations need to know what their AI tools can do, how to use them effectively, what happens when they fail and who is responsible. That picture can blur when multiple applications are involved.
There are a lot of start-ups and software suppliers offering AI tools to work in conjunction with your Finance/ERP system, and these products promise to solve particular problems. This, in essence, is a good thing, but it’s important to remember that each new product can bring another supplier, integration and support responsibility. A business that accumulates bespoke tools for every process can end up with a complicated system that is harder to maintain than the problem it was meant to solve. A knowledgeable partner can help you see the full picture, steer you away from quick fixes, and guide you toward a solution that grows with your business and works with your existing systems.
Before even looking at AI tools, paving the way to success requires mapping the process and defining the desired outcome, be that less manual data entry, fewer bottlenecks, faster access to useful information or better visibility of performance. Only then should specific tools be looked at, with considerable thought put into whether a carefully chosen add-on or a tailored solution is the right fit. The aim should be to improve the process while keeping the core system dependable and supportable.
Finance/ERP projects have a significant impact on the people who keep a business running and the customers they look after. Any changes mean teams have to manage their day-to-day work while also learning new processes and systems. If the plan is unrealistic, or the wrong system or partner is chosen, even good software can become painful, leading to budget overruns or delays. An overly ambitious or wrongly scoped project risks grinding your team down as they attempt to juggle their workloads with a new system that doesn’t make sense.
That makes due diligence vital. Be clear about what your chosen AI is promising, check that the system fits your processes, agree realistic timelines and test the results before you go live with any changes (nobody likes AI hallucinations). Choose a partner who takes time to understand your business, clarifies the value the project should deliver and supports your team through the change. That work upfront can help reduce rework and keep the project focused on practical results.
AI will keep developing, and businesses should look for sensible ways to innovate. Adoption for its own sake is no measure of success. The goal is to run the business better, make sound decisions, serve customers well and compete with confidence. That means matching tools to real needs, introducing them carefully and ensuring people understand the work they are meant to improve.
There’s a reason we don’t leave milk out and hope for the best. Without the right care, even something valuable can quickly turn sour, and AI is no different. Rushing in without safeguards is the fastest way to pour an investment down the drain. With sound judgement and a strong partner, though, AI can help businesses work more efficiently and serve their customers better.


