4th September 2026
For decades, central bankers have had to make decisions about the British economy using information that is inevitably looking backwards.
Inflation figures arrive after prices have already changed. Employment statistics take time to compile. GDP figures are revised. Business surveys provide valuable information, but they only capture what businesses are asked and what they are willing to report.
Meanwhile, the real economy does not wait.
Businesses change their prices every day. Consumers change what they buy. Companies advertise vacancies, cut staff, expand investment or postpone orders. Firms talk about rising costs, falling demand and shortages. Financial markets move by the second.
The Bank of England has long recognised that its ability to make good decisions depends upon having good information. What is changing now is the technology available to gather and interpret that information.
The Bank is increasingly using artificial intelligence, machine learning, natural-language processing, high-frequency data and web scraping as part of a much broader transformation of its data systems.
This could eventually change something that affects every household in Britain.
It could change what the Bank of England knows about the economy when it decides what to do with interest rates.
he old problem with economic information
There is an unavoidable problem with traditional economic statistics.
They take time.
When the Office for National Statistics publishes an inflation figure, it is describing what happened to prices over a period that has already passed. When GDP is published, it is telling us what happened to economic output rather than necessarily what is happening today.
That is not a criticism of official statistics. They are enormously valuable and have to meet demanding standards of accuracy and consistency.
But policymakers increasingly want something else as well.
They want to know what is happening now.
The Bank of England has therefore been developing ways of supplementing traditional statistics with much more frequent information. Its own description of monetary policymaking now explicitly refers to big data, high-frequency indicators, unstructured data, artificial intelligence and intelligence gathered directly from businesses and markets.
That represents a significant change in the way a central bank can observe an economy.
The internet has become an economic sensor
One of the most interesting tools is web scraping.
In simple terms, web scraping means using software to collect information from websites in a structured way.
A human looking at one hundred websites would take hours.
A computer can potentially examine thousands or millions of pieces of information much more quickly.
For an organisation such as the Bank of England, that opens up an entirely different source of economic intelligence.
Prices advertised online can provide clues about changing inflationary pressures. Job vacancies can provide information about labour demand. Company websites can reveal changes in products, services and business activity. Other publicly available information can provide clues about what businesses and consumers are doing.
The Bank says web scraping is already being used for some data collection and that its new Enterprise Data Platform should allow more sophisticated integration of different sources of information.
That is potentially a major development.
Instead of waiting for a traditional survey to tell policymakers that prices are rising, technology can help provide a much more continuous stream of information about what is happening.
It does not replace official statistics.
It adds another layer of information.
Then comes AI
Collecting the information is only half the problem.
The other problem is making sense of it.
If you suddenly have millions of pieces of information, you have not necessarily solved your problem. You may simply have created a much bigger one.
This is where artificial intelligence becomes important.
The Bank's AI strategy says its new cloud-based Enterprise Data Platform is intended to make it easier for staff to work with the Bank's data and use AI tools. The Bank says the platform will support machine learning and AI applications and allow secure access to both structured and unstructured data.
Unstructured data is particularly interesting.
A spreadsheet containing inflation numbers is structured.
A company's report, an economic briefing, a regulatory document or thousands of written comments from businesses are not.
Historically, extracting useful information from those documents required people to read them.
AI can increasingly help identify themes, classify information, summarise documents and detect patterns.
That does not mean handing economic policy over to a chatbot.
The Bank is quite clear that AI has to be reliable, secure and properly governed. But it does mean that economists can potentially examine vastly more information than they could previously.
Perhaps the most interesting data still comes from people
There is an irony here.
For all the excitement about artificial intelligence, some of the Bank's most valuable economic information still comes from people talking to other people.
The Bank has a network of regional Agents across Britain who speak directly to businesses and community organisations.
They hear about what is happening on the ground.
Are customers spending less?
Are wages rising?
Are businesses struggling to recruit?
Are companies investing?
Are orders falling?
Are costs increasing?
These conversations can provide information that does not yet appear in official statistics.
The Bank says its Agents operate from 12 agencies across the UK and have thousands of discussions with businesses and community organisations every year. Their intelligence is used by the Bank's policy committees, including the Monetary Policy Committee that decides interest rates.
Now AI is being brought into that process.
The Bank is developing generative AI to help its Agents extract insights from the information gathered during company visits.
That is a fascinating combination of old and new technology.
The human talks to the business.
The human records what has been learned.
AI helps analyse thousands of those conversations.
Economists then use that information alongside conventional statistics and their own judgement.
It is not the replacement of human economic intelligence.
It is an attempt to make much more use of it.
The Bank is already discovering things about AI itself
There is another reason this development is particularly interesting.
The Bank is not simply using AI to study the economy.
It is using its network of businesses to find out what AI is actually doing to the economy.
Its July 2026 Agents report looked at businesses adopting AI and found that some firms were using it to automate routine work and accelerate knowledge-based activities. It also reported that some companies were reducing demand for certain junior and administrative roles while demand was increasing for people with AI, data and analytical skills.
That information matters enormously to monetary policy.
Suppose AI makes companies more productive.
That could allow them to produce more without increasing prices as much.
That could reduce inflationary pressure.
But suppose AI also causes businesses to invest heavily in computing infrastructure, data centres, electricity, semiconductors and specialist equipment.
That can create additional demand and higher costs.
The Bank's latest work is already highlighting this tension.
AI can therefore be both a source of productivity and a source of new inflationary pressures.
Without better information, it would be much harder to understand which effect is becoming dominant.
[v]Could this mean faster interest-rate decisions?[/b]
This is where the technology becomes relevant to ordinary households.
When the Monetary Policy Committee decides whether interest rates should rise, fall or remain unchanged, it is essentially trying to judge where the economy is heading.
It does not simply ask what inflation is today.
It has to consider what inflation is likely to be tomorrow.
That means identifying turning points.
If the economy is weakening, policymakers want to know before the weakness becomes obvious in official statistics.
If inflationary pressures are building, they want to know before those pressures become embedded.
More timely information could therefore improve what economists call "nowcasting" and "nearcasting", giving policymakers a better picture of current economic conditions and where they may be heading. The Bank explicitly identifies these approaches as part of its monetary-policy toolkit.
But there is an important warning.
More information does not automatically mean better decisions.
The danger of believing the machine
The internet contains an enormous amount of information.
It also contains an enormous amount of rubbish.
A price on a website might be temporary.
A job advertisement might remain online after a position has been filled.
A company might publish optimistic statements that bear little resemblance to its actual financial condition.
Social media can amplify extreme opinions.
AI-generated material is becoming increasingly common.
The Bank itself has warned about inaccurate AI-generated material and the growing volume of what it calls "AI slop".
There is therefore a paradox.
AI can make it possible to process far more information.
But the more information becomes available, the more important it becomes to determine which information deserves to be trusted.
That is why the Bank is not simply buying an AI system and allowing it to make economic decisions.
It is building data infrastructure, governance and analytical systems around the technology.
This could be particularly important for places like Caithness
There is another question worth asking.
Does better data mean better information about Britain as a whole, or can it eventually give policymakers a clearer picture of Britain's individual regions?
That matters enormously.
Britain is not one economy.
London is not Caithness.
The South East is not the Highlands.
A technology company in Cambridge does not face the same economic conditions as a small engineering business in Wick.
A central bank making national interest-rate decisions inevitably needs national statistics. But regional information can help policymakers understand why the same interest rate can have very different consequences in different parts of the country.
The Bank's Agents already provide some of that regional intelligence.
AI could potentially make it easier to analyse that information at scale.
That could be particularly valuable for rural economies where national averages can hide what is really happening.
We are moving towards an economy that can be watched almost in real time
This may ultimately be the biggest change.
For much of the twentieth century, governments had to wait for information about the economy to be collected, processed and published.
The digital economy is different.
Prices can change instantly.
Financial transactions generate data continuously.
Companies publish information online.
People search for products and services.
Businesses advertise jobs.
Millions of transactions leave digital traces.
The technology now exists to collect, process and analyse enormous quantities of this information.
The Bank of England is beginning to build the infrastructure to make much greater use of it.
Its 2021 plans for transforming financial-sector data collection already recognised that digitisation and automation were changing the way economic information was produced. Its more recent data strategy and Enterprise Data Platform take that process considerably further.
The direction of travel is clear.
The central bank of the future will not simply receive statistics.
It will increasingly observe the economy.
But seeing more does not necessarily mean predicting better
That distinction is important.
An AI system might tell the Bank that thousands of businesses are raising prices.
It might detect that job vacancies are falling.
It might identify that consumers are spending less.
But policymakers still have to decide why these things are happening and what they mean for the future.
That requires economic judgement.
It also requires understanding things that may not appear in the data at all.
A small business owner may postpone investment because they are worried about the future. A household may stop spending because it is frightened about its mortgage. A manufacturer may delay recruitment because it cannot obtain a particular component.
Numbers can reveal the symptoms.
People often explain the cause.
That is why the most interesting development may not be AI replacing economists.
It may be AI allowing economists to combine far more evidence with human judgement.
Britain is entering a new era of economic measurement
The Bank of England is therefore doing something much more significant than adopting a fashionable new technology.
It is attempting to modernise the way one of Britain's most important institutions gathers and understands information.
Web scraping can provide new sources of real-time information.
Machine learning can identify patterns in large datasets.
Generative AI can analyse documents and written reports.
The Bank's Agents can continue providing intelligence from businesses around Britain.
And the new data platform can bring many of these sources together.
If it works, policymakers may eventually have a much richer and faster picture of what is happening in Britain before the traditional statistics catch up.
That could improve decisions about interest rates and financial stability.
But it also raises a fascinating question.
If the Bank of England can see the economy more clearly than it ever has before, will it also become better at acting before problems reach the household?
That is the real test.
Because ultimately the purpose of all this technology is not to create better databases or more impressive artificial intelligence.
It is to make better decisions about the British economy.
And for millions of people paying mortgages, renting homes, running businesses, saving money or trying to cope with the cost of living, those decisions matter enormously.