8 minute read 19 Nov 2019
Politicians talking in government building

How AI can help governments manage their money better

8 minute read 19 Nov 2019

Show resources

AI is helping governments improve efficiency, detect irregular financial activity and deliver better value for money through augmented human decision-making.

Artificial intelligence (AI) and related technologies have the potential to solve many of the problems revenue and finance agencies face. By optimizing processes, they can make public finance management (PFM) and tax administration more efficient – and help agencies to meet the constant demand to do more with less.

By improving controls and spotting anomalies in large volumes of data, they help reduce accounting errors, identify risks and prevent tax fraud and financial crime. And by automating routine processes, they allow agencies to focus human input on where it counts – using insights from data to make better decisions.

The technologies that deliver these results – such as machine learning, robotic process automation and natural language processing – are fueled by large amounts of data. For their results to be valuable, the data behind them (including macro-economic statistics, audit reports, source data provided by taxpayers, data collected from banks and other intermediaries) needs to be reliable, consistent, transparent and available in a timely manner.

We’re also seeing governments experimenting in blockchain-based solutions alongside AI-enabled technologies. These give managers a clear picture of, for example, where money is being spent, goods are being moved and transactions are being closed.

Six ways governments could use AI in public finance management

1. Making revenue agencies more efficient and effective

Revenue agencies face the arduous task of collecting volumes of information from and about taxpayers. They also identify risks, calculate taxes, monitor tax collections, verify tax returns and address tax queries.

To improve compliance, tax authorities across some countries are using AI models along with analytics to predict those at higher risk of not paying their taxes. They’re also predicting which of those taxpayers are likely to react positively to certain tax interventions. The result is increased revenue collection and lower tax avoidance.

AI-enabled technologies can also help revenue agencies to move more and more of their communication with taxpayers onto digital platforms. This gives citizens the fast, convenient experience they expect, while freeing up employees’ time for more impactful work.

Tax authorities are showing particular interest in virtual assistants or “chatbots”.  Skatti, a chatbot for the Swedish Tax Agency, handles around 15,000 citizen queries about tax returns a month, making services more accessible, personal and efficient. Some chatbots also “learn” from the questions citizens ask. This builds up taxpayer guidance over time, so the underlying systems can handle enquiries more accurately. 

Tax authorities across some countries are using AI models along with analytics to predict higher-risk tax payers. The result is increased revenue collection and lower tax avoidance.

One way of doing this is to examine the relationship between people and computers to understand how machines can boost the effectiveness of tax professionals, resulting in a collective intelligence. To that end, high-performing revenue agencies should seek a solution that will develop a governance model of AI systems for tax purposes, and help governments better understand the externalities of tax policy with AI systems.

2. Preventing financial crime and predicting corruption

Financial crime continues to plague governments and the financial services sector. According to UN estimates, money laundering and related crimes alone cost between US$1.4 trillion and US$3.5 trillion globally a year

Regulators are encouraging government agencies and financial institutions to use AI technologies to detect money laundering. In doing so, they will help to cut funding for human trafficking, narcotics, arms sales, and terrorism.

Governments can also use AI to predict corruption and spot fraud. Researchers at the University of Valladolid in Spain have created a computer model that calculates the probability of corruption taking place in Spanish provinces. The model, which is based on neural networks, also identifies the conditions that encourage corruption.

Meanwhile, AI is also being used to predict and prevent fraudulent health insurance claims. By analyzing the massive amount of data created in the healthcare system, the technology can flag likely fraud before it starts and can create an evidence-base for fraud examiners and law enforcement to prosecute cases.

Governments elsewhere are using AI for disbursement as well as to control leakages in subsidies and welfare payments. Denmark is using AI and blockchain for faster and more effective processing of welfare payments to beneficiaries. And the UK’s Department for Work and Pensions has rolled out and tested AI algorithms to track down large-scale corruption of its benefits and welfare programs.

3. Transforming procurement to deliver better value for money

For years, technology has been taking some of the load in this labor-intensive area. But it’s only been able to do this partially.

AI has the potential to not only streamline procurement processes, but to transform them – saving time and money. It can do this by:

  • Making procurement processes smarter – for example, alerting government and suppliers to supply chain disruption, recognizing and flagging supplier compliance issues and identifying fraud
  • Automating menial procurement tasks, particularly in invoice processing and approving proposed purchases
  • Identifying opportunities to reduce government spending

Technologies under the AI umbrella can also improve the procurement process for government entities, as well as for suppliers bidding for the government work. For example, the US Air Force recently announced plans to use AI to make sense of complex acquisition regulations and speed up the buying process for goods and services.

AI has the potential to not only streamline procurement processes, but to transform them – saving time and money.

And the rebuilt BidSync™ supplier application (from Periscope Holdings) uses machine-learning technology to filter out irrelevant government opportunities and deliver the most winnable ones for each user – saving valuable time.

4. Identifying anomalies for public sector audits

Auditing and accounting in the public sector involve collecting and using huge volumes of data. As a result, the scope for human error – and the ability of AI to mitigate the risks – is higher than in most areas of government.

AI can automate a big chunk of the work humans have traditionally done, freeing the auditor up to focus on areas requiring human judgment. It can prove particularly useful in identifying anomalies or errors and finding unusual transactions that could indicate fraud or unintentional errors. It can also spot whether any transactions are missing from accounting records.

Proprietary technology can be used to detect anomalies and transform the way audits are conducted. Some governments are using MindBridge Analytics’ award-winning Ai Auditor to fulfill a similar purpose. The tool uses a mix of techniques, from decision-based rules and statistical methods to AI and machine learning, to detect anomalies in financial data. It can also analyze 100% of the transactions in a data set, resulting in analysis that’s more effective and efficient.

5. Improving macro-economic forecasting

AI and machine learning are maturing to become powerful tools for carrying out macro-economic forecasting. This real-time, multidimensional economic modeling can yield many benefits, including:

  • Identifying potential market risks and predicting situations that could harm the economy
  • Collecting big data to formulate and implement procedures in line with government agency policies
  • Predicting the impact of policy reforms in various scenarios

The Estonian government, for example, has been carrying out a pilot project to create analysis models that will monitor and predict the economy (pdf). It will also use big data generated by businesses, along with modern modeling techniques, to assess the impact of policies.

6. Making sense of vast amounts of unstructured data

As digital technologies take hold, the volume of data that revenue and finance agencies handle will grow. Much of this data will be “unstructured,” which means it will be found in emails, videos and social media posts, rather than organized in a traditional database.

Most finance and revenue agencies are yet to use AI in this way. But by tapping into previously untouchable data, they’ll be able to make radically better decisions in the future.

It would be impossible for a human to analyze thousands of pages of this data without error. But AI technologies, including text analytics and natural language processing (NLP), can do just that. For example, the Australian Tax Office used NLP techniques, among others, to review the 13.4 million files (pdf) that emerged in the Paradise Papers leak and extract names of interest.

Most finance and revenue agencies are yet to use AI in this way. But by tapping into previously untouchable data, they’ll be able to make radically better decisions in the future.

Trust is crucial to success

AI will find a firm foothold within revenue and finance agencies as they move to supplement human intuition with data-driven insight. But like all areas of government, they’ll need to prepare properly to mitigate the potential risks and build the culture needed to support these new technologies.

Revenue and finance agencies handle large amounts of sensitive financial data. So, their officials will need to reassure the public that the right level of data protection and ethical standards are in place if they’re to build trust and legitimize their use of AI. Agencies will achieve better buy-in when citizens understand the logic of the system, particularly where an administrative decision needs explaining. 

They’ll also need to be sure that the taxpayer and citizen data the algorithms use is of a good quality, to avoid unintended bias and mass-scale mistakes. Continued monitoring will be needed to make sure errors or unintended consequences are detected and quickly corrected. The systems will require ongoing supervision by knowledgeable humans.

Finally, embedding AI into existing working practices will be a key change management issue for all organizations. Success requires the trust and buy-in of finance professionals who are ready to embrace new methods of decision-making.


AI has the potential to change the way governments’ finance functions operate by creating a reliable system built on digital trust, transparency and cost efficiency. But to reap the full rewards, revenue and finance agencies need to trust their own data as well as the AI system they deploy. And they need to build the same trust among citizens.

About this article