IFA Conference 2024: Is AI moving too fast? How to capitalise on this fast-growing trend

For those who are reluctant to adopt AI technologies, the co-founder of online accountancy platform provider Capium assures they can enhance productivity rather than cannibalise accountancy work. Asked whether accountants should fear AI, he says “don’t get daunted”, but start using the technology “personally before thinking about business” to build their AI expertise over time.

Accountants must manage AI risks

Patel stresses the need for accountants to verify AI-generated information for accuracy and potential data breaches in the absence of AI governance regulations.

He says “a lot of regulation needs to be sitting behind this” to mitigate the risks from unethical AI use. Potential risks include data privacy breaches, copyright infringement, discrimination and lack of insurance cover. In the absence of any global regulatory standards for AI, it is challenging to establish responsible governance to manage AI-related risks.

AI models are trained to enforce ethical limits.

“It won't tell you how to rob a bank due to human training and is taught to avoid harmful behaviours,” Patel says.

Mastering prompt engineering produces better AI outcomes

Managing these risks means understanding the importance of “prompt engineering”. These are the instructions that describe a task or search for a generative AI model to achieve the most relevant response.

Generative AI models are trained on vast quantities of data to create text, images and audio. Patel says most of the foundational AI models are large language models (LLMs) trained on natural language that can be adapted to perform a vast array of tasks. LLM architecture mirrors the human brain with a complex neural network with “billions of parameters fed into it”.

The model converts “input” text into numbers and transforms “outputs” back into text, images, audio or video. Achieving the best output relied on the quality of prompts people use to search or develop content.

“Prompt engineering is about effective communications,” Patel says. “It’s as important as us being able to read and write. The more effective you are with prompt engineering, the better results you will get.”

Language models “learn from patterns and the environment” and require “human feedback” to learn through reinforcement. There are a variety of models that vary in speed and cost and developers have identified “emerging capabilities springing up from these models”. Patel says they can be used for role playing, poetry and ideas for strategic discussions and provide a structure or an agenda, but were not for plagiarising.

“Be a sponge and open your mind,” Patel says. “The capability of AI is really limited by us and our prompt engineering tools. If we open it up, it is capable of doing much more for us.”

“AI won't take your job, but if you are not using AI someone using AI will take your job,” Patel says.

He cautions against being over-reliant on the technology and to “make sure you make the final judgement” and decisions. Accountants could assess areas “causing pain” that AI could address, such as capturing data for potential clients or streamlining bookkeeping.

Study finds 83% of UK firms boosted wages for AI skills

A recent UK Future Workforce Index survey of more than 2,267 freelancers, business leaders and workers, revealed 83 per cent of businesses were willing to offer higher pay for AI ability to address a skills gap. This willingness to boost remuneration underscores the critical role of AI skills in innovation and maintaining a competitive advantage.

According to the new Fiverr Future Workforce Index, nearly a third (32 per cent) of companies report a lack of AI expertise in high-demand areas such as natural language processing, AI content creation, and chatbot development.

Catch up on previous news from the IFA Conference 2024: Driving change