One in three people have used Gen AI

Generative AI (Gen AI) tools like ChatGPT, Gemini, DALL-E and Midjourney are still young – they were only thrust into mainstream conversation 18 months ago.¹ Since then, companies across industries have raced to rebrand as AI-centric, c-suites have crafted Gen AI into their strategies and roadmaps, and some people have been experimenting and creating code, reports, art, contracts, music and more, using a growing range of ever more powerful Generative AI tools.

Deloitte surveyed a nationally representative sample of 4,150 UK citizens between 16-75 as part of its annual Digital Consumer Trends research, in April 2024.

Three in five (60%) people in the UK are now aware of a Gen AI tool, up modestly from 52% in May 2023 (see Figure 1). One in three (36%) have actually used a Gen AI tool – the equivalent of 18 million people between 16-75-years-old. This is a 35% increase from the previous survey (May 2023), when 26% or 13 million people had used Gen AI. It may be that the rate of adoption is slowing down, but this is a natural part of the progression for new technologies. Consumers who wanted to try Generative AI may have already done so: the most popular tools are easy to access via web browser or app. Sign-up takes minutes, and most offer a free version.

Awareness and use are not split evenly between genders. In terms of awareness, 67% of men have heard of Gen AI, while just 52% of women have. In usage, 43% of men have used Gen AI, but just 28% of women have. Generative AI tools can produce biased responses, typically as a result of bias that exists in the corpus of training data which is itself a reflection of the biases that have long existed in wider society.²  How to deploy Gen AI safely and ethically is a crucial challenge, and the industry’s collective response undoubtedly requires more women at the table, using and giving views on these tools.

Among the one in three people to have used Gen AI, frequency of use varies greatly. One in ten users (10%) do so daily (this is equivalent to 4% of the UK). One in four users (26%) do so weekly. The aggregated number using it at least weekly is 36% of users (equivalent to 13% of the UK). Conversely, 41% of Gen AI users claim to use it less than monthly, or to have only used it once or twice. This group may have experimented with Gen AI but found little utility from either personal or work perspectives. When asked why usage was infrequent, 23% of this group reported it was not helpful, 19% were not satisfied with the answers, and 13% said it gave inaccurate answers. Some were also concerned about cyber and legal problems, with 17% of the low-frequency group worried about data privacy, and 13% thinking it infringes on copyright and intellectual property of others.

Seven million have used Gen AI for work

Generative AI tools have a broad range of capabilities. In our personal lives, uses might include creating diet plans,³ learning about topics of interest or creating bedtime stories.⁴ However, consumers have long since become accustomed to the idea of most internet services being free, with ads enabling monetisation. Gen AI providers are still working out how to incorporate ads (or if they even should). Therefore, at present it is business use of Gen AI that has the best chance of monetising first.

Of Gen AI users, 39% claim they used it for a work purpose. This is equivalent to 14% of the UK (see Figure 3), and 7 million people between 16-75. This is up 66% from last year (May 2023), at which point 4 million in the UK had used Gen AI for work. 

It is notable, however, that a large proportion, possibly the majority, of those using Gen AI for work purposes may be doing so without official endorsement by their employer.⁵  A substantial number of companies (nine in ten) still lack a policy on whether use of Gen AI is acceptable, or a governance structure for if it is.⁶  Generally, the employee is moving faster than the employer.

Of those using Gen AI for work, the most common tasks are generating ideas (44%) and looking up information (41%) (see Figure 4). This suggests that those using Gen AI in work may still be in an exploratory stage of using the technology: they have still not ascertained how to get the best out of Gen AI’s current capabilities. As Gen AI is a predictive tool based on historical data sets, it may not be ideal for generating brand new ideas. Also, Gen AI can ‘hallucinate’, or get things wrong, and it may not be optimal for looking up information.⁷  That said, the third, fourth and fifth most popular tasks are all apt uses for Gen AI – creating written content (39%), writing emails (38%) and summarising text (37%).

Workers using Gen AI are positive about its impact on their productivity. One in four of them (28%) claim it makes them a great deal more productive, and nearly half (46%) claim to be a fair amount more productive (see Figure 5). That said, this productivity boost may be limited to a narrow set of tasks at present. However, despite this, a vast majority of workers do not feel encouraged by their employer to use Gen AI – just 27% who are aware of Gen AI claimed that their employer encouraged its use. These figures suggest that there is latent productivity and worker satisfaction that employers can unlock if they invest in Gen AI tools, create governance structures and empower their employees. Unlocking that latent productivity can initiate a transformation of workflows and processes.

Some users may not be aware of Gen AI’s deficiencies 

Gen AI is unlike most technology that enterprises buy and use today: it has the potential to hallucinate, so to be imprecise, or get things wrong. And it has the potential to create outputs which are biased, based on the available information on the internet over the last few decades that has been used to train it. Most people are aware of this, but a significant number still are not. 

Of those who have heard of Gen AI, 25% believe, wrongly, that it is always factually accurate and 26% think it is unbiased (see Figure 6). Among respondents that use or have used Gen AI, that rises to 36% and 36%, respectively. Gen AI deployment needs to be accompanied by a major learning and development programme – including training on ethics and responsible use, and how to extract the most value from tools. 

For one thing, many knowledge workers – and occasionally their bosses – continue to conflate AI with Gen AI, and many are concerned about their role being automated. Deploying software without this layer of human support could have limited, and possibly adverse effects.

Customers may initially undervalue Gen AI outputs

One business application of Gen AI may be customer service chatbots, trained on data and transcripts from previous interactions. 

Consumers might initially be hesitant about the hypothetical usage of Gen AI in such services, but it may be possible to overcome anticipated aversion to this new technology. Early deployments of Gen AI in customer service have been shown to increase satisfaction,⁸  but if consumers realise that Gen AI enables a better experience, with higher quality answers, they are likely to be more welcoming of Gen AI upgrades to chatbots.

When asked about Gen AI in customer services, over half (56%) said they would be less inclined to use it if they knew they were conversing with Gen AI (see Figure 7). In some instances, such as when lodging a complaint – there may be no substitute to talking to a human. Customers may also have had sub-par experiences with existing chatbots, and may not be aware of how Gen AI would be better.

What is more, three in five (59%) would be less inclined to trust an email if it was created using Gen AI. This could be a customer receiving an email, or it could be peer-to-peer emails in a workplace. If companies are forced to label AI-generated outputs as such in future, they should be aware of the potential discredit this might cause to their outputs.

The bottom line: Gen AI is here, and employers should move quickly to support their workforce

For 18 months, companies have been exploring what they could do with Gen AI. A popular approach has been to conduct proof-of-concepts, or partake in early access programmes for enterprise software vendors. However, deploying at scale is very different to a trial.

Companies ought now to ask what they should do, not what they could do, with Gen AI. Increasingly, the CFO is coming to the table, and the criteria for funding a deployment at scale tends to be a measurement of return on investment and worker productivity – something which can be challenging to do in nuanced roles and knowledge work.⁹

Companies should also be aware that a number of their employees may be using Gen AI tools, whether or not they are officially endorsed. Employees may even be using personal devices to avoid detection. Improving enterprise fluency is key here. Employees should be given access to a comprehensive education program before, during and after a Gen AI tool is deployed. 

Given that 7 million people in the UK are using Gen AI at work, a robust governance structure and employee training are important to mitigate risks. The employee has moved faster than the employer so far – it is time to catch up.