“Can AI really save me money?” At TBI, we know through our work in more than 45 countries that this is a question finance ministers are beginning to ask. Behind it lies a deeper pressure: how to unlock fiscal space quickly without relying only on tax rises, borrowing, spending cuts or reforms that take many years to pay off.
In low- and middle-income countries, governments face difficult fiscal landscapes: slower growth, tighter debt markets, lower aid flows and citizens’ rising expectations. Raising taxes, cutting investment or reducing services can be politically costly and economically damaging. Using technology to improve processes comes with inherent political risk; a failed project can cost millions. Finance ministers therefore face a dilemma: whether or not to accept additional delivery complexity in pursuit of faster fiscal gains.
However, a new path has emerged: using artificial intelligence, data and digital tools to identify revenue loss and spending leakage that can be acted on quickly, while continuing to move forward with longer-term reform of public financial management (PFM).
We call this a dual-track approach, and it is a practical way for leaders to start delivering results now. It means adding a fast track to the PFM-reform agenda: a focused portfolio of AI-, data- and technology-enabled solutions that can raise revenue, reduce leakage and protect spending before foundational systems are fully modernised. This fast track should run in parallel with longer-term work to strengthen financial management, tax administration and digital-identity systems, rather than replacing it.
This approach is feasible because AI and digital tools are now capable, affordable and accessible enough to operate on top of imperfect existing systems. For example, they can analyse invoices, payrolls and registries to detect anomalies, prioritise cases, support document review and enhance audit capacity. The point is not to wait for perfect data; governments should use the data they already hold to identify the highest-value use cases and act on them quickly.
The opportunities across expenditure and revenue are clear. Governments can identify ghost workers, ineligible subsidy claims, suspicious payments and inflated procurement prices. They can target tax non-compliance, customs fraud, risky VAT refunds and discrepancies between taxpayer activity and declarations.
But a reliance on technology alone will not deliver these gains. Fast-track initiatives will succeed only when they are deployed as a delivery reform, not a conventional IT project.
This paper provides examples of fast-track projects that have generated substantial fiscal benefits. It then sets out five tests for governments to use to identify high-value opportunities and outlines an implementation model that can be applied to ensure projects are delivered successfully.
It explains how to deliver dual-track PFM reform – using AI, data and digital tools to create fiscal space more quickly while continuing to drive long-term change.
Why Governments Can Act Now
Political leaders are under pressure to deliver more with less: more spending on education, health, jobs and resilience, but with limited fiscal space and, often, declining public trust. In this environment, targeted interventions that can recover leakage, improve compliance and generate visible gains in months are unusually valuable.
Fiscal pressure itself is not new, but now the technology frontier has moved: AI, better data integration, cloud-based tools, digital verification and remote sensing make it easier to add an intelligence layer over existing fiscal systems. A government no longer needs to complete a full integrated financial management information system (IFMIS) replacement, tax-system overhaul or payroll rebuild before it can begin finding anomalies, ranking risks and targeting interventions.
This matters because many governments already hold enough usable data to start. Payroll files, payment records, procurement systems, customs declarations, tax returns, beneficiary registries, company registers, geographic information system (GIS) data and digital-ID records are incomplete and fragmented, but they can still reveal tractable problems. AI and analytics can help make these imperfect data sets more useful by matching records, identifying outliers, prioritising cases and reducing the cost of review.
The AI-enabled capabilities most relevant to fast-track PFM include:
Agent-supported casework: Read files, query systems, compare records and prepare case summaries for officials.
Messy-data analysis: Find patterns across spreadsheets, PDFs, scans, invoices, declarations and registries.
Document and contract review: Check contracts, tenders, invoices, audit reports and customs descriptions for risks or inconsistencies.
Cross-system entity matching: Link people, firms, accounts, tax IDs, suppliers and beneficiaries across fragmented databases.
Explainable risk scoring: Rank taxpayers, suppliers, refunds, payments or declarations by likely fiscal risk.
GIS and image verification: Use satellite imagery, maps and geotagged photos to verify projects, assets, mining or agricultural activity.
Revenue intelligence: Compare declarations with invoices, manifests, trade data and company records to flag under-reporting or misclassification.
Workflow triage: Channel low-risk cases into faster routes and send high-risk cases to the right official.
Intervention simulation: Estimate likely fiscal yield, workload and false positives before scaling an intervention.
Learning loops: Improve targeting from confirmed recoveries, false positives and enforcement outcomes.
The dual-track model therefore lets governments act with “good enough” data, prove value within months, and use these wins to strengthen the political and fiscal case for deeper reform. Early gains do not remove the need for foundational reform. They make it easier to sustain and can help convince sceptics that reform is worth pursuing.
International Examples Show This Can Work
Evidence from both expenditure- and revenue-linked reforms shows that targeted technology-enabled interventions can generate substantial fiscal benefits. The strongest cases share a pattern: governments focused on a clear fiscal problem, used data they already had or could assemble quickly, linked the technology to a specific operational decision and followed through institutionally.
Payroll and wage-bill controls: Nigeria and Ghana have shown the value of linking payrolls to verified identities. By integrating biometric or digital-identity systems with payroll databases, governments can remove ghost workers and duplicates. Reported gains are substantial: by removing 50,000 ghost workers from its payroll in 2016, Nigeria generated monthly savings of 13 billion naira ($50.5 million, based on the annual average exchange rate for 2016); Ghana removed around 26,589, with annual savings of about $103 million. These are focused interventions to improve the integrity of public administration that have generated large fiscal returns.
Procurement savings: Bangladesh and Ukraine, together with broader cross-country evidence, have shown what happens when tendering, contracting and visibility move onto digital platforms. A study covering 11 countries, including these, found that e-procurement reduced bid prices by roughly 6.75 per cent on average, with average annual savings of about $637.9 million for low-income countries and $5.2 billion for middle-income countries. Estimated benefit-cost ratios ranged from roughly 8:1 in low-income settings to as high as 473:1 in the case of lower-middle-income countries.
Subsidy targeting: India has demonstrated how technology can convert broad, leaky subsidy systems into more precise instruments. It removed millions of bogus liquefied petroleum gas accounts, saving 738.5 billion rupees (approximately $8.5 billion, based on the Q1 2025 average), and stopped 58.7 million false ration cards, saving roughly another 1,850 billion rupees ($21.4 billion), while protecting real households.
Revenue and customs analytics: Tanzania, the UK, India and Brazil have shown that better use of administrative and third-party data can strengthen revenue collection. Tanzania’s use of GIS data for property taxation, together with AI and data-mining approaches in the UK and India, illustrates how a more complete picture of the taxpayer can improve risk selection, broaden the base and make enforcement more targeted. Brazil’s AI-powered SISAM customs system is especially relevant for the fast-track agenda. It applies machine learning across 100 per cent of Brazilian import declarations. According to Brazil’s customs authority, reviewing just the top 1 per cent of declarations ranked by SISAM risk score identified 22 per cent of fiscal-classification errors, while inspecting the top 20 per cent identified 81 per cent. AI can make routine declaration review far more targeted.
Tax-administration digitalisation: Rwanda and South Africa have shown how easier compliance and more automation can lift performance. Rwanda’s e-Tax platform moved registration, filing and payment online at system level. South Africa’s automatic assessments demonstrated how digital tax filing can raise compliance and improve collection by reducing friction for taxpayers and administrators alike. Sweden showed the complementary efficiency upside very clearly: digital filing, AI and robotic process automation cut company registration times by up to six days and reduced costs by 28 million Swedish krona, or about 16 per cent of the process cost.
These cases demonstrate the logic behind the fast track: focused technology interventions can generate real fiscal gains when they are tied to clear operational decisions. AI now changes the cost, scale and speed of that opportunity. It can help governments detect anomalies, review documents, score risks, match entities, verify projects remotely and analyse data across systems at lower cost, making fast-track PFM reform more feasible than it was even a few years ago.
The Benefits of a Dual-Track Model
Finance ministries do not need to choose between long-term foundations and near-term gains. They need both. The long-term track strengthens the digital and fiscal infrastructure of the state, which includes areas like financial management, tax administration, payroll, procurement, customs, digital identity, payments, data standards and institutional controls. These systems remain essential.
The fast track is different. It is a portfolio of AI-, data- and technology-enabled solutions that can integrate with existing systems and deliver visible fiscal value quickly. This track should focus on a small number of projects where the fiscal prize is large, the data signal is good enough, officials can act on the output and political sponsors are willing to protect implementation.
The two tracks reinforce each other. Foundational systems provide rules, transaction records and institutional channels for action. Fast-track initiatives generate savings, improve data quality, expose weaknesses and build confidence that deeper reform is worth sustaining. The fast track is not a substitute for foundational systems. It is the second track governments need while those systems are being strengthened.
This is also why AI should be treated as an intelligence layer, not merely as a feature inside future systems. Currently, governments can use AI and analytics to comprehensively analyse payroll, procurement, tax, customs, payments and registry data. Over time, those same interventions can help specify which foundational upgrades matter most.
Fast-Track Use Cases: Five Tests Leaders Should Apply
The best place to start is where money, data, ownership and feasibility intersect. There should be enough potential fiscal value to justify political attention; enough usable data to test the idea; enough operational control to act on the findings; and enough delivery capacity to move quickly. The objective is not to find the perfect reform. It is to find a credible top-five list that aligns with the country’s specific context, choose one to three initial use cases and build confidence through measurable results.
In practice, governments should apply five tests:
Potential fiscal yield: Is the leakage or untapped revenue large enough that even a partial improvement would matter fiscally and politically?
Data readiness: Is there enough usable data to identify risks, cases, anomalies or payment errors, even if the underlying systems are imperfect?
Delivery complexity: Can the intervention be delivered with limited integrations and manageable behaviour change, rather than requiring a wholesale institutional overhaul?
Political sponsorship: Will senior leaders actively back the reform, especially when it threatens rents, profits or patronage networks?
Coalition for change: Is it realistic to align the agencies, officials and external stakeholders needed to implement the reform and defend it when resistance appears?
These tests should be applied during a short diagnostic sprint, producing a simple delivery plan with owners, milestones and expected gains.
A Model for Successful Delivery
The operational model matters as much as the technology. The most promising fast-track initiatives will fail if they are managed as conventional IT projects with diffuse ownership, long procurement cycles and little connection to frontline decisions. They should be run as focused delivery programmes by finance ministries, supported by the clear vision of a finance minister, and be built around the following steps:
Establish a small delivery task force, anchored close to political authority, with seconded staff from finance, revenue, procurement and relevant line ministries, and with a mix of operational, policy, data and technical expertise.
Use political authority deliberately: clear obstacles, signal priorities, protect reformers and communicate why the reform benefits citizens rather than vested interests.
Operate on a short delivery rhythm, with regular problem-solving sessions and a simple scorecard that tracks fiscal value, milestones and named responsibilities.
Assign a senior responsible owner for each fast-track project, accountable for outcomes rather than for process compliance alone.
Design the reform with collaboration across technology, people and processes. The tool must fit the workflow, the workflow must fit the institution and the institution must be able to act on what the tool finds.
Use partner contracts and procurement structures that avoid lock-in, support open standards and allow governments to retain control over data, models and core operational decisions.
Reinvest part of the gains into scaling what works and into strengthening the underlying systems that make future fast-track initiatives easier to deploy.
Fast-track does not mean governance-free. Governments need enough guardrails to choose use cases responsibly, avoid vendor-led problem definition, validate data and model outputs, protect sensitive information and keep officials accountable for decisions. A comprehensive discussion of AI governance and technology procurement is beyond the scope of this paper. The key point is that fast-track interventions need proportionate safeguards from the start, with governance capabilities expanding as the portfolio scales.
Every fast-track initiative should also have a small metrics dashboard. Fiscal metrics should track savings identified, revenue recovered, leakage reduced and compliance improved. Operational metrics should track cases reviewed, inspections prioritised, duplicate records removed, payments blocked and audits completed. Data and model metrics should track match rates, false positives, risk-score precision and data gaps resolved. Institutional metrics should track owners appointed, decisions acted on, process changes adopted and scale-up decisions made.
The purpose of these metrics is not only accountability. It is to help decide whether to scale, redesign or stop. This protects governments from expensive technology experiments and enables funding for what is proven rather than what is merely promised.
How Donors Can Support Dual-Track PFM Reform
For donors, the task is to modernise how PFM reform is funded. Long-term support for core systems remains essential, but it should be complemented by flexible funding for diagnostics, data work, embedded delivery capacity, narrow high-value use cases and rapid scale-up of successful projects. The reason is practical: diagnostics help governments identify where fiscal gains are most realistic, data work tests whether the signal is strong enough and embedded delivery support helps turn analysis into action. Done well, this can produce early wins, improve delivery and build the business case for continued investment in deeper PFM reform.
The key choice is not between foundations and fast wins. It is to find the right balance to fund both in a way that helps governments deliver meaningful fiscal results within the current political window, while supporting long-term reforms. If funded well, this is an agenda where governments can plausibly show successful results within a minister’s term, not only beyond it. Donors should focus on supporting partners who can bridge technical, institutional and delivery challenges rather than funding those elements separately.
Priorities for Governments
For governments, the immediate task is to create the fast track. That means selecting a small number of fiscally meaningful problems, backing them politically and treating them as priority delivery challenges. Visible gains in leakage reduction, subsidy integrity, procurement savings or compliance can create fiscal space now while building momentum for broader reform. The evidence suggests that these early wins are often more within reach than governments realise. A practical first step is a diagnostic sprint to build a top-five list of opportunities, test the data signal and select one to three first use cases with clear owners and metrics.
The practical opportunity now available to finance ministries is clear: continue building the foundations, but do not wait for them to be perfect. Use AI, data and digital tools to create a fast track that raises revenue, plugs spending holes and builds the confidence needed for deeper reform.