What would you trust AI to do?

Answers from business leaders are growing increasingly broad, with AI entrusted with anything from data collection and processing to identifying suitable candidates in the hiring process.

Without a doubt, AI is a time saver – with some estimates sitting as high as 40 per cent over completing a task manually – but what else is it?

As its role in the business landscape grows from a fringe concept to something like a leadership position, it’s important for business owners to understand the potential risks of AI and how they’ll need to mitigate them if they’re going to use it successfully in the long term.

An artificial sense of security

Contrary to sensationalist opinion, the current risks to businesses of AI are unlikely to be world-ending or lead to mass unemployment.

In fact, AI creates a real opportunity for innovative businesses to introduce time-saving measures which make them more efficient and offer clients a cost-effective, top-tier service – particularly in areas such as business operations, accounting, payroll and client administration.

However, this leads us onto the first major risk to businesses which embrace AI without proper planning and training – the skills gap.

The skills gap

Although AI in some form has existed for a while now, its use in business is generally new and evolving constantly, meaning that many talented professionals lack the necessary skills to effectively use an AI tool.

For example, using generative AI such as ChatGPT requires the user to know exactly what they need and how to ask the system for it, as well as common errors made by the programme.

This is easily solved but will require investment in training from firms seeking to innovate. Higher education may also play a role, with AI being incorporated into degree courses sought by AI-reliant industries.

In the meantime, such missing skills are a contributing factor to another difficulty brought about by using AI – accuracy and accountability.

Project tracking and fact checking

Although the capabilities of commercial AI systems are constantly growing, the fact remains that it isn’t perfect.

Many generative systems lack up-to-the-minute information even with the capacity to search the web, while others can easily misunderstand an instruction and pull through the wrong information.

This is particularly critical when dealing with client projects or financial information, where an error could create major loss of assets or trigger legal proceedings.

The reality is that the timesaving and automation capabilities of AI is limited when crucial information is involved, as it will need to be manually checked and verified.

This is also true beyond the scope of information and generative AI. While we may like to consider machine learning to be free of ‘human’ error, it is far from it. This feeds into the third major stumbling block for AI adoption, particularly in hiring practices and personnel decisions – unintended bias.

Machine-driven prejudice

I asked ChatGPT to write a description of a CEO of an accounting technology firm, five times over. On each iteration, it returned a description of a white man in his mid-40s.

While this may seem trivial, it is indicative of a well-documented phenomenon – bias sneaking its way into AI programming.

The major causes of this include programmer bias, conscious or unconscious, and machine learning based on online information. This means that historic prejudices, such as women leaders in business, and current majorities, find their way into AI representations of datasets.

This is most risky when used to HR and onboarding processes. Business owners open themselves up to accusations of discrimination and allowing bias to affect the hiring process.

Teams may also miss out on qualified candidates without certain keywords on their CV, for example.

Mitigating the risk

These are just some of the risks posed by AI to businesses and their owners.

However, there are many uses for it that can make your business run more smoothly and take on more work for sustainable growth.

Learning how to mitigate the risks is essential to successful AI integration.

You can do this through:

• Identifying the risks facing your specific sector
• Identifying whether each task or operation really requires AI
• Assigning staff to monitor work produced by AI systems
• Reviewing human and machine bias within the team
• Identifying the necessarily skills and investing in training

As with human workers and existing technologies, error cannot be fully eliminated from AI programmes, but with the right support, AI can make accountancy and other business operations more efficient and client-focused.

Our experts can guide you through how to identify the issues with AI facing your business and how to mitigate them within your business operations.

For more information on how we can support you, please get in touch with a member of our team.