Al technology is not new, but has become much more accessible since OpenAI’s ChatGPT chatbot release at the end of 2022. Groundbreaking advances in large language models and generative AI are allowing more automation of cognitive tasks. AI technology is developing at rapid pace and applications are broadening.
In Autumn 2023 the Bank of England’s agency network surveyed its business contacts to ask about their investment plans including how these related to AI.
What did we learn from the Autumn 2023 agency investment intentions survey?
Across all industries, almost 30% of surveyed agency business contacts had made ‘new significant or transformative AI investments in the UK’ over the previous 12 months, and over 40% planned to do so in the future (typically in two to three years’ time). But 25% said they didn’t know if they would make any (additional) meaningful future AI investments, and 30% said they hadn’t invested materially in AI in the last year and had no plans to do so (Chart 1).
When split by industry sector, AI investment was most common in business services (mostly within finance, law and accountancy). And the proportion of business services firms investing in AI was expected to rise from 40% in the past year to around 50% in the future. Other sectors had seen lower AI investment over the past, but a sharp pickup was expected across all sectors over coming years (Chart 1).
Chart 1: Investment in AI (a)
Footnotes
- (a) Survey respondents could answer (Yes/No) to the question: ‘In the past 12 months have you made any new significant or transformative AI investments in your UK business?’. In answer to the question: ‘Do you have any plans for future significant or transformative AI investments in your UK business?’, respondents could answer (Yes/No/Don’t know) and there were three possible answers for Yes: (Next Year/Next 2–3 years/After 3 years).
Fourteen per cent of companies that had invested significantly in new AI technology in the past year had experienced an immediate productivity gain, 15% expected a productivity increase within a year, while 48% expected higher productivity within 2–3 years (Chart 2). Immediate, and within a year, productivity gains were mostly arising from specific work task automation (such as writing programming code and using chatbots for customer services enquires) and process optimisation (such as inventory distribution and delivery routes). Whereas for productivity gains expected to take 2–3 years to accrue or longer, one key reason cited by companies was the need for AI technology pre-integration, for example updating data and/or digital IT systems. The survey didn’t ask about the size of any anticipated productivity gains, however.
Chart 2: Timing of productivity gains (a)
Footnotes
- (a) Chart 2 shows responses to the question: ‘How quickly do most of the productivity gains from the new AI investments accrue to your UK business?’ from respondents who had answered ‘Yes’ to the question: ‘In the past 12 months have you made any new significant or transformative AI investments in your UK business?’.
In this post we supplement those survey results with agency intelligence gathered since then, and some additional research, to provide a fuller picture of how UK businesses are using AI and what the implications might be for their future labour demand.
How are UK businesses increasingly using AI?
The use of AI by UK businesses is likely to be higher now than suggested by the Autumn 2023 agency investment intentions survey.
Software developments in the past year have made AI use increasingly accessible to companies. Because our contacts can increasingly buy off the shelf AI powered solutions, they are able to use AI, even if they have not themselves invested in developing an AI tool, so the survey results probably understate AI adoption. In other cases, some companies have made incremental AI investments in their UK business over several years, as part of a wider multi-year ongoing digitisation strategy. And some firms are using AI in the UK, but the technology investments were made elsewhere, for example overseas, outsourced or by their parent company. Despite this when companies filled in our Autumn 2023 survey, they did not consider investment over the previous year to be ‘significant’ yet; but the impact of AI use on productivity could be more substantive going forward given the very broad range of potential AI uses highlighted in our interviews with business contacts.
Some potential AI business use examples are outlined below:
A: Automating routine, repetitive, tasks. AI software robots can deliver financial reporting; tax returns; client on-boarding; and document recognition, retrieval and processing.
B: Using AI to answer customer service enquiries more quickly and effectively. For example, by using voice recognition; prioritising customer phone calls; providing real-time scripts for customer service staff to answer questions and to sell products and services; and taking notes of customer calls. These improve the speed, quality and reliability of calls handled by customer service staff. Chatbots are also used by some agency business contacts, where the handling of queries has been (almost) fully automated.
C: Optimising business operations in real-time, to improve performance and efficiency:
- Supply chain management, such as predicting changes in customer demand; automatic ordering of new stock; better determining product selection and mix. These help to minimise waste and avoid large stock overhangs so more stock can be sold at full price.
- Workflow management. For example, routing deliveries/journeys; predicting and scheduling production/service downtimes and machine maintenance; and best use of space and time.
- Workforce management and monitoring, such as rota schedules based on staff availability, skills sets and preference, and business needs. And analysing worker activity and performance, for remote working and working from home for example.
- Pricing, eg dynamic/surge pricing determined by location and/or time and/or customer type; tailored discounts/offers; estimating price elasticities and evaluating market prices.
- Facilities management. For example, using sensors to control building temperature and air flow to reduce energy consumption, and to anticipate repairs (eg to the lifts) based on estimated wear and tear from flows of people.
D: A variety of other business uses:
- Producing computer code in software development.
- Writing legal contracts, news articles, job adverts and consultancy reports.
- Work on large data sets (such as cancer screening of x-ray photos, finding a specific face in CCTV footage, cleaning and improving archive video footage).
- Language services (eg translation, producing subtitles, transcripts and meeting notes).
- Personalisation of products and services (for example, a phone App that is automatically tailored to an individual’s circumstances).
- Creating new designs (in advertising and architecture for example).
- Discovering new products and services (such as drugs and therapeutic applications).
What does this mean for UK labour demand?
In general, many businesses that we have interviewed have steered that AI adoption is already having an impact on their businesses. But few expect AI adoption to significantly reduce their labour demand imminently. Impacts are likely to come over coming years, but the size and timing remain uncertain.
Part of the reason for that is regulation. Some tasks might be fully automatable in theory, but there may be good reasons why it does not happen in practice. For example, although a plane could now in theory be piloted automatically, regulations require two pilots in each plane for safety reasons. As AI technologies continue to advance, companies, their customers and regulators will need to agree their risk tolerances for AI use.
Automating time-consuming routine tasks can relieve some workers from mundane parts of their job, freeing them to spend more time on interesting and higher value-added tasks. Agency business contacts say that a key benefit from AI use would be more motivated workers, who gain time to do a better job, perhaps improving output quality and/or customer service for example, rather than efficiency savings.
Integrating new AI technologies will not be quick for many businesses. Since our Autumn 2023 survey, a few companies have found their desired AI use will take longer to execute than originally planned, with more upgrades to their data and technology infrastructure being required or the AI integration proving to be more complex. Workers also need time to learn and adjust to using AI effectively. Business contacts expect some staff will be needed to monitor and correct AI generated outputs to ensure accuracy and consistency, as the technology is not yet sufficiently reliable. And many workers who are redeployed after AI led automation will likely need retraining and upskilling, which will also take time.
For some business contacts, the likely job displacement impacts from using AI technology fall mainly on more routine back-office functions which are often outsourced overseas now, for example in shared service centres in Eastern Europe or Asia. That means that increased AI use may reduce current or future demand for offshore outsourcing, rather than directly impacting UK labour demand.
Some of our business contacts who use AI most innovatively, including new generative AI, aim to grow both their market share and the total market size, through offering new products and services, that would otherwise be too labour intensive without AI automation.
Contacts suggest that greater AI use by firms requires more workers with certain skills, like data scientists and technology workers for example. That would change the skills mix and not necessarily the total number of people employed at those businesses.
This post was prepared with the help of Lai Wah Co.
This article is based on findings presented to the Monetary Policy Committee in October 2023 and April 2024.
Share your thoughts with us at BankOverground@bankofengland.co.uk