The hidden cost of unclassified spend: Why finance and procurement teams are turning to AI

Every CFO has asked the question: Where exactly is our money going? And across industries, from logistics and healthcare to manufacturing and insurance, the answer is often unclear.

That answer is in your spend data, but if your data is uncategorised, it is drowned out by the noise. For finance leaders and AP teams, uncategorised spend is not just a reporting issue; it’s a strategic blind spot. It hides risks and inefficiencies, preventing insights into procurement performance and optimisation of your supply chain. Even with significant investments in ERP systems, shared services, and e-procurement platforms, most organisations fail to classify, analyse, and manage their spending effectively. Without accurate classification, it’s impossible to answer fundamental questions like:

• Should we diversify or consolidate our supplier base?
• Can we negotiate better terms?
• Are we leaking cash through off-contract purchases or duplicate vendors?

The root cause is simple: the data is messy. Company names and structures are inconsistent. Invoice line items are coded abbreviations that require subject matter expertise and context to be decoded. The result? Data that is difficult to work with makes strategic procurement challenging and introduces risk into budgeting, sourcing, and supplier management.

Why classification fails

Spend data is notoriously difficult to work with:

• Inconsistent records for the same organisation across platforms, business units and time, with different abbreviations and wholesale name changes due to acquisitions and rebrands
• Continuous churn in both the members of your supply chain and the suppliers themselves
• A massive volume of data at a line item level that exceeds the human attention span and mental capacity to manage manual updates
• Classification of suppliers by broad-brush company SIC codes rather than examining the details of the goods or services being delivered
• Purchases, occurring outside procurement systems, for speed or to circumvent red tape
• Indirect spend categories like IT, HR, or professional services that are under-structured or under-managed

The team managing this data often has other priorities. Worse still, historic attempts to use rule-based systems have failed to scale, suffered from high error rates, and the volume of change makes manual administration extremely difficult and expensive. The result is slower, and less accurate metrics that inhibit agile decision making and bury opportunities for optimisation,

The business impact of poor metrics

When data about spending is poorly managed, the downstream consequences ripple across the organisation. Finance loses control, procurement loses leverage, critical insights are missed, and opportunities are lost. Some of the most common and costly ways that a lack of spend transparency undermines performance include:

1. Off-contract and maverick spend
   Purchases made outside of preferred vendors and negotiated terms are difficult to detect without structured categorisation at the line-item level. This leads to lost savings and weaker procurement control.
2. Fragmented supplier chain network
   Redundant vendors across regions or departments erode pricing leverage and increase vendor management complexity and overheads.
3. Missed commercial incentives
   Without visibility into payment timing and terms compliance, businesses can miss out on early payment discounts, virtual card rebates, and squander their working capital.
4. Inadequate oversight of indirect spend
   Spending on software, consulting, consumables, or facilities often escapes scrutiny.
5. Difficulty with ESG, risk, and compliance reporting
   Modern procurement strategies require insights into supplier risk, diversity, and emissions. That’s nearly impossible without a well-classified, accurate spend dataset.

How AI is changing the game

While many talk about AI in procurement, few deliver the combination of capabilities required to make spend truly visible and usable.

Precise line-item categorisation and an understanding of the diversity in a supplier’s product lines are necessary. Then your business can move beyond simply grouping transactions with ambiguous keywords and rules to accurately classifying products and services, even when descriptions are sparse or unstructured. Simultaneously, supplier records must be consolidated and verified, as standalone entities or network structures, to properly understand their capabilities. Achieving both of these requires a blend of machine learning, AI, and third-party enrichment.

This allows:

• Spend to be categorised across all levels — category, sub-category, and function — with minimal human input
• Every transaction can be linked to a clean, deduplicated supplier profile with commercial attributes
• Policy exceptions, term mismatches, or off-contract spend can be surfaced in real time
• Transactions can be tagged with insights for ESG reporting, risk management, and compliance auditing

This is not possible with generic AI, applied loosely; it needs to be purpose-built to handle the fragmented, noisy, and high-volume data that finance and procurement teams struggle with daily — the result: faster, trusted, and context-rich spend insights that enable better decisions at scale.

An illustration from the healthcare industry

Historically, healthcare providers attempting to leverage spend analytics solutions have faced a slow and resource-intensive process. Data had to be manually exported from multiple ERPs, often on-site and distributed across various hospitals and business units. Once acquired, the data classification used rigid rules or GL codes, with widely varying levels of specificity and accuracy. Given scarce resources and the initiative-driven nature of these projects, organisations often brought in consultants to manually process and report on the data. This human-capital-intensive approach typically takes months and usually yields insights that are already outdated by the time they reach decision-makers.

Today, SmartSpend from Previse eliminates almost all manual effort and can operate continuously in the background. Invoice line items are automatically classified, even with vague or missing descriptions. Supplier records are deduplicated, normalised, and enriched with third-party data, providing a clear, unified view of spend with individual suppliers and their familial networks. Off-contract purchases, duplicate vendors, and compliance issues are surfaced in real time. This allows procurement and finance teams to focus on value creation, optimising contracts, supporting General Ledger classification adherence, and enabling smarter, faster decisions across both clinical and non-clinical categories.

SmartSpend leads the way in how far AI-powered categorisation has come. In a recent case study:

• 95% of supplier records were enriched with accurate, publicly sourced company and contact information
• Over 80% of records were matched to third-party data sources like D&B or Moodys 
• Accurate supplier data, coupled with multi-model AI algorithms, enabled over 95% of invoice line items to be classified without significant manual intervention

This level of automation frees up procurement teams and provides immediate, timely insights.

Why it matters now

Procurement and finance teams are being asked to deliver more visibility, control, and savings, without increasing headcount. At the same time, pressure is growing to track and report on risk, ESG, supplier diversity, and working capital metrics with greater precision. Without clean, classified spend data, both are nearly impossible to achieve.

Previse’s SmartSpend isn’t just about operational efficiency; it’s about enabling smarter decisions, faster responses, and better outcomes.

Generative AI has unlocked new possibilities in financial operations that represent a significant leap forward from rule-based systems and static taxonomies. However, generic AI models fall short when the dataset is huge and the data is messy, due to fragmented and duplicated variations of supplier entities and little to no context. SmartSpend understands who your suppliers are, no matter how complex or incomplete the data.

We love to show how the solution:

• Can resolve fragmented supplier identities
  • Consolidates variations across ERPs, invoices, and payment systems
  • Links vendors to their global parent and enriches with D&B firmographics, ESG flags, and risk signals
• Will classify invoice line items in context
  • Even when descriptions are vague or unstructured
  • Leverages supplier behaviour, historical spend, and third-party data
  • Assigns multi-level taxonomy tags down to commodity or service level
• Identifies and flags anomalies in real time
  • Off-policy spend, payment term mismatches, and duplicate vendors
• Attaches ESG and risk metadata at the transaction level
  • Carbon impact, supplier diversity, and payment terms that are mapped and explorable using multiple facets

If you want to negotiate from strength, optimise supplier relationships, and deliver financial value, your data has to work harder. With the right tools, it finally can.