In the corporate world, firing a newly hired assistant after one or two small mistakes would seem unnecessarily harsh. That’s because there’s a broad understanding that humans need time and experience to learn and improve on the job. Yet, when it comes to AI, there is a much lower tolerance for error, with research indicating the average tolerance rate for AI sits at 6.8%, much lower than the 11.3% average for human employees.
This double standard raises the question, are we subconsciously rooting against AI?
This might seem incorrect on the surface – the UK AI sector is attracting an average investment of around £200 million daily and jobs are being created as a result. However, beneath the high-level enthusiasm, there is scepticism, with data indicating less than half of the UK population trusts AI. It’s clear then, there needs to be a mindset shift if we’re to truly reap the benefits of AI.
AI hype distracts from ‘boring’ but impactful use cases
AI headlines are often sensationalist, ranging from bold promises of fully self-driving cars to alarming reports of the presence of hallucinations in court documents. For different reasons, these types of headlines feed negative perceptions of AI — either that it consistently overpromises and underdelivers, or that its risks far outweigh its benefits.
Meanwhile, quieter but more regular and meaningful applications of AI are being overlooked. Across sectors including healthcare, logistics and retail, AI is already delivering impressive, measurable results. The Central and North-West London NHS Foundation Trust is a great example of this. By deploying AI agents, the Trust has successfully automated a range of time-consuming manual processes, such as patient data entry, scheduling and administrative tasks. As a result, it has saved around £370,000 in manual effort a year and freed up 56 hours a day for healthcare professionals.
Focusing on a small percentage of non-representative use cases creates a damaging narrative around the measurable impact of AI. To move past this, the focus needs to shift from the spectacular to the practical examples, recognising the true power of AI often lies in incremental, day-to-day operational improvements.
Why short pilot phases are crucial for unlocking AI’s potential
One of the biggest mistakes businesses can make with AI is immediately rushing into large-scale deployments, whether it be with LLMs or agentic systems. AI is a costly investment and needs to be integrated carefully with existing workflows, data systems and business processes to deliver long-term results.
This is why short pilot phases with a clear roadmap to production are essential for successful, long-term implementation. Running a focused pilot project creates a controlled environment for businesses to trial a new AI tool. This is a crucial step to de-risk integration, providing an opportunity to troubleshoot issues and pivot strategy ahead of wider deployment, scaled deployment of the technology.
Importantly, pilots, alongside investment in thorough training initiatives, can help to improve employee confidence and buy-in. Many employees are wary of AI, concerned it could replace them or devalue their expertise, a fear compounded by a lack of knowledge and support in using the technology. This isn’t theoretical – a staggering 73% of the UK public recently stated they haven’t received any AI education or training. By involving employees in pilot phases, businesses can measure their feedback on the new tool and provide training in the early stages of deployment so worked can confidently harness the technology to its full advantage.
Pilot phases are not just a key practical step; they’re a cultural one. They’re fixed term experiments that serve to build the confidence of teams, stakeholders and customers, clearing the path of uncertainty ahead of widespread deployment.
Patience and governance are key for building trust
Strong governance is also a non-negotiable for successful AI deployment; measures must be embedded from the outset, not bolted on as an afterthought. AI tools are frequently granted access to sensitive business and customer information so a lack of oversight doesn’t just pose operational risks, but it can lead to reputational harm and even legal consequences. Although AI tools, specifically agents, are increasingly autonomous, there should always be the presence of both strong governance and human touch to ensure safe usage and to mitigate issues in real-time.
Transparency and accountability also lay the groundwork for trust. Employees and customers need to understand exactly how AI is being used. This means being open about what decisions are influenced by AI, how systems are monitored and what checks and balances are in place to ensure fairness and accountability. Transparency is what will convert fear into confidence, creating an environment where innovation can thrive.
Over time, the combination of pragmatic pilot phases and robust governance strategies will enable businesses to strike the right balance of controlled experimentation. It’s this approach of orchestrating agents, robots and systems with governance and security built in from day one that will allow AI to mature steadily and successfully.
Setting up AI for long-term success
AI will inevitably make mistakes, especially in its early stages of deployment. By giving it the same grace period we’d offer a new employee, issues can be resolved in real-time within a well-governed environment. This approach will lay the right foundations for long-term success and ROI.
A necessary step will be addressing the misconception that AI technology is a direct replacement for human intelligence. AI should be viewed as a supportive tool, taking on time-consuming, administrative tasks and freeing people to focus on strategic tasks such as orchestrating workflows and determining outcome based KPIs rather than chasing model accuracy. Ultimately, the success of AI depends not just on the technology itself, but on how thoughtfully it’s introduced and the mindset of those using it.
