Vehicle leasing companies rarely run on a single source of truth. They rely on multiple, non-integrated systems that each hold a different slice of the picture, and getting those slices to agree is a persistent, manual chore.

To explore a better way, Enablis built a prototype that uses AI to reconcile disparate data sets - not just to solve the reconciliation problem itself, but to prove out how AI can help build and integrate the system that solves it.

Jacob Williams, Senior Consultant at Enablis, walks through a prototype for using AI to reconcile two or three different data sets in a vehicle leasing business.

The problem: data that doesn’t naturally talk

As Jacob explains, so many businesses implement so many different systems and hold so many different data types that fragmentation becomes the default state. In a vehicle leasing business, different areas of the operation each own their own data set:

  • Broker systems - the businesses and platforms relied on to bring the cars onto the books in the first place.
  • Internal customer and vehicle records - the organisation’s own system for managing customers and the vehicles being sent out to them.
  • Maintenance logs - data covering the servicing and upkeep of those vehicles.
  • Fuelling data - records of how and where those vehicles are being fuelled.

The result is lots of different data sets that are quite disparate and don’t natively or naturally talk to one another. That is exactly the problem space this prototype targets.

The approach: AI-powered fuzzy matching

Rather than forcing every system into a single rigid integration, the prototype uses AI to perform fuzzy matching across the different data sets. Records that describe the same vehicle, customer or transaction rarely line up perfectly across systems - formats differ, identifiers are inconsistent, and small variations creep in. Fuzzy matching lets the AI recognise those near-matches instead of demanding an exact one.

The system then produces a report that makes the state of the data plain:

  • What percentage of the data sets match
  • Which specific records match
  • Which records don’t match - and, crucially, why they don’t

Human in the loop

The AI does the heavy lifting, but people stay in control of the verdict. By adding human-in-the-loop verification, the report becomes a shortlist to review rather than a black-box decision. As Jacob puts it, you might be left with ten records to check: on review, one record does actually match with another, while a second genuinely doesn’t line up with X, Y and Z.

That combination - AI handling the scale of comparison, and a person confirming the edge cases - is what makes reconciling disparate data across brokers, internal systems, maintenance and fuelling both faster and trustworthy.