Which business tasks can AI take on? And which can it not?

Receive free Artificial intelligence updates

We’ll send you a myFT Daily Digest email rounding up the latest Artificial intelligence news every morning.

How many businesses currently use AI?

Taking the UK as an example, about 15 per cent of all businesses — roughly 432,000 — had adopted at least one form of artificial intelligence technology by January 2022, according to a government survey. It found that the businesses most open to using AI were those in IT and telecoms, finance and accounting, media, advertising and sales, and law. 

As businesses grow, they become more likely to adopt AI, with 69 per cent of large companies adopting at least one form of AI, according to the research.

However, these figures predate the huge leaps in generative AI, which became more widely known about at the end of 2022, with the launch of OpenAI’s ChatGPT: a chatbot that can create large swaths of text and answer questions convincingly. It triggered an explosion in similar chatbots, as well as customised generative AI solutions for different professions, including law and advertising.

Adoption of AI has more than doubled since 2017, according to consultancy McKinsey. It is set to grow exponentially now that AI has become more accessible through large language models and easier to deploy in different settings.

What are the most common business uses of AI?

AI is commonly used in businesses to boost efficiency, by automating tasks such as data analysis, improving the speed or consistency of services, or using data from clients or customers to inform decision-making and make predictions.

In infrastructure planning, for example, city engineers can use AI to detect when a pothole is likely to emerge in a road, using cameras and sensors to monitor conditions over time. 

In retail — given the huge amount of data collected on consumer spending patterns — AI is being rapidly adopted to optimise the supply chain, by predicting the amount of stock required. AI is also being used to communicate with customers through virtual assistants.

How will AI change business interactions with customers?

AI-powered customer service is not a new development — the technology has been used in retail settings for more than a decade. Chatbots usually have set questions and responses to triage requests and limit the interactions needed with human customer service agents. However, the significant advances in generative AI — the type of AI that is capable of producing large paragraphs of conversational text — mean customer service functions are likely to be transformed.

AI can readily answer frequently asked questions and provide personalised advice. Meta, which owns Facebook, WhatsApp and Instagram, has recently spoken about its efforts in advancing so-called AI agents for users and businesses on the platform.

“We believe it’s important to have a diverse array of AI Assistants to represent people, creators, and businesses interests,” said Nicola Mendelsohn, head of Meta’s global business group. “For example, a person could get customer support from their favourite small business without having to wait on the phone.”

Speech recognition is another area where AI has been deployed to recognise and respond to words.

Can AI make decisions for business?

AI’s ability to identify objects in images or video, often called computer vision, has been used in various sectors, including in social media content to detect and ban harmful content, in healthcare to identify possible tumours or features on scans, and in the automotive industry for parking assistance or in the development of self-driving vehicles.

In the latter example, AI often makes a decision based on the information provided by the images, such as whether to perform an emergency stop or change course. 

AI is also widely used in the supply chain to identify inventory, help to monitor stock, reduce waste and reduce delivery times by grouping orders from the same place.

The more data an AI collects in a specific use case, for example, the more orders are completed, the more effective AI’s predictions will be.

Do humans still need to be involved in AI decision-making?

Yes. At the moment, AI can get things wrong, especially generative AI, which can state falsehoods as facts, a phenomenon known as “hallucinating”.

Another problem is algorithmic bias, where AI has been found to discriminate against people based on factors such as race or gender. For example, an analysis of AI-driven mortgage lending decisions found some programs were less likely to approve black applicants.

Which business sectors are most exposed to AI?

As the current trend for AI is in large language processing — which is the ability to process huge amounts of language data from text, audio or video — the sectors most likely to be affected are those involving the written or spoken language: advertising, journalism, consulting and legal industries.

All of these have already adopted generative AI technologies.

Which business sectors are least likely to be affected?

Research from the University of Pennsylvania and OpenAI in March found that AI driven by large language models, such as generative AI, was likely to impact about 80 per cent of the US workforce. But jobs in areas such as agriculture, mining, and manufacturing were least likely to be affected, the paper concluded.

Analysis by Goldman Sachs in March also grouped craft and related trades as the sectors least likely to be automated by AI.

However, these industries already use some forms of AI. For example, farmers use sensors to monitor crops and detect disease, pests or issues with soil. 

In many of these industries, AI using data analytics or robotics may play a greater role, and become able to replace manual workers.