How far should your rural business trust AI?

Artificial intelligence (AI) is widely discussed and increasingly used in many businesses, with some suggesting that it is the answer to almost every problem. It comes in many shapes and forms – some free though often only its entry level tier (such as ChatGPT, Google Gemini and Microsoft Copilot), some bespoke and paid-for. Drawing on our own experience, this article offers practical guidance on adopting AI with caution, and provides tips for avoiding pitfalls.

The CLA’s legal team has been trialling a variety of AI tools to help with our research, drafting and analysis. The results have been interesting and dangerously wrong in equal measure. For some relatively narrow tasks, it augments our work efficiently – for instance, summarising some known case law or generating suggestions for where to look for the answer to a given question in our legal database.

However, analysis results are nowhere near consistent enough to be relied on, and, occasionally, answers can prove deeply problematic.

Responses invariably sound highly believable yet frequently turn out to be wrong, incomplete, focusing on the wrong question or subtly misleading. For instance, when asked to simply review primary sources, AI recommendations have produced errors that, if they came from an adviser, could be considered professional negligence.

For example, I asked an AI tool to review a question on firearms licensing, primarily looking at the Home Office’s Firearms Licensing Law. Its answer, if followed without verification and amendment, could have led to the committing of a criminal offence. Likewise, AI’s interpretation of the requirements within the Torts (Interference with Goods) Act 1977 could have resulted in unlawful interference with property and potential criminal liability.

When questioning the AI tool, after I had reviewed primary sources and found its answers to be wrong, it cheerily acknowledged that, yes, that answer had been not just wrong but entirely made up, and then gave the reason why.

While the pitfalls can feel daunting, AI remains incredibly useful. Consider an imperfect analogy, also relating to potential improvements to rural communities’ productivity.

In the early 1700s, towards the beginning of the agrarian revolution, English farmer and barrister Jethro Tull arguably developed and promoted the horse-drawn seed drill. It doesn’t sound exciting now, but this helped to replace the practice of broadcasting seeds by hand – not at all efficient – by enabling precise, even placement at a consistent depth and spacing.

Tull held several ideas that seem eccentric today, such as his opposition to fallow fields. He also believed that constant mechanical ‘pulverisation’ of the soil could eliminate the need for manure by making nutrients more accessible to plants. In any event, the agricultural community slowly adopted the seed drill. Its use led to substantially reduced seed waste, improved germination, better weed control and higher yields.

This pragmatic approach of cautious adoption of new technologies allowed the best elements to move forward. Similarly, AI is a powerful augmenting tool for CLA members who manage farms, estates and rural businesses – but its true value depends on how it is applied.

AI systems can integrate satellite imagery, drones, soil sensors, weather data and machine learning to deliver incredible insights that parallel the seed drill’s precision. They can enable variable-rate applications of seed, fertiliser and pesticides that lower input costs while maintaining or increasing output. For landowners and rural businesses, these tools can help to plan new income streams, make better use of land and inputs, improve livestock management and give earlier warning of weather and price risks. There are huge potential upsides; when applied thoughtfully, AI can strengthen productivity, sustainability and resilience.

However, when using generative AI for complex questions, the importance of relying on verified information rather than substituting AI for your own or professional judgement cannot be overstated. Current AI models are known to ‘hallucinate’, meaning they will confidently present inaccurate or fabricated details. Hallucinations are inherent to generative AI, which prioritises producing fluent, plausible answers over verifying accuracy. Rates of hallucination vary, but for complex queries, a rate of around 15-30% is not unusual – and many figures are higher.

AI can also flip or shift its position markedly, particularly if its initial response does not align with what the user hoped to hear and it is queried again. Systems often default to agreeable answers. In a world saturated with convincing but potentially misleading information, the role of independent verification or human judgement is crucial.

Relying on AI for decisions that should involve qualified experts can expose members to serious errors, liability and unnecessary costs. An AI-generated recommendation that contributes to financial loss, regulatory breach or operational failure will not be easily remedied by professional indemnity insurance.

Tull’s seed drill delivered benefits only when skilled operators adapted it to local conditions while exercising appropriate judgement on his other suggestions. Similarly, AI functions best as an augmenting technology that amplifies expertise without replacing human accountability. CLA members have a strong tradition of responsible stewardship and pragmatic innovation.

• Assume that AI can be, and frequently is, wrong – confidence and accuracy are not the same.
• Check any answer that you might rely on – particularly for complex legal, financial or compliance outputs. They must be verified either from a primary source or through professional advice.
• Treat answer changes as a warning – shifting responses may signal uncertainty, not improvements.
• Don’t accept ‘helpful optimism’ – AI tends to downplay risk and ambiguity, so be thorough in checking.
• Use AI to assist, not to decide – it can augment your own judgement, but accountability remains with you.
• Stick to trusted sources – primary materials and professional advisers still matter most, particularly CLA advisers.
• Slow down before acting – speed is AI’s strength, but caution is yours.

If you have questions about your land or business that can’t be answered or verified by AI services, don’t hesitate to reach out to us at the CLA.