From 20 to 22 August 2026, three teams of officers from the Kingston and St Andrew Municipal Corporation and Tax Administration Jamaica walked Half Way Tree street by street. Hagley Park, Eastwood Park Road, Red Hills Road, Hope Road, Constant Spring Road, Waterloo Avenue, Half Way Tree Road, Maxfield Avenue, Winchester Road, Ruthven Road. They were chasing 432 properties with tax arrears going back six years, worth $162 million, plus a further $48 million owed by 195 businesses in unpaid trade licence fees.
That is a serious, well-run operation. A temporary customer service desk opened daily at Constant Spring Arcade so nobody had to travel to a central office. Kingston Mayor Andrew Swaby called it "an exercise in fairness and accountability," not harassment, and TAJ's Deputy Commissioner General for Operations, Dave Jeffery, set a target of at least 85 percent compliance on the outstanding arrears by March 2027.
What it is not is a scalable model. Half Way Tree is one commercial district in one parish. KSAMC and TAJ have already run this same door-to-door approach in St James, Portmore, and St Catherine, in each case reportedly lifting compliance from as low as 25 percent to 75 percent. That is real progress bought with real field hours, three zones of officers walking every street in the order the streets happen to run, engaging one property owner at a time. Jamaica has 14 parishes. The Caribbean has dozens of municipalities carrying the same six-year arrears problem, and none of them can staff a permanent field operation the size of this one indefinitely.
The Numbers Behind the Knock
Break down the $162 million and the shape of the problem becomes clearer. Of the 432 delinquent properties, 151 are tied to businesses, accounting for $96 million of the total. The remaining 281 properties, carrying $66 million between them, are presumably residential or mixed-use. Add the $48 million in trade licence arrears across 195 registered businesses and the combined shortfall in this single district passes $200 million, money the municipal corporation is statutorily owed and has not collected for up to six years.
None of that arrears appeared overnight. A property does not fall six years behind without someone, at some point, failing to flag it. That is the part a field operation cannot fix retroactively. KSAMC and TAJ can recover what is owed today, but the same gap that let 432 properties drift six years behind will keep producing new delinquent accounts next year, and the year after, unless something changes in how the arrears get spotted in the first place.
What AI-Powered Tax Compliance Actually Looks Like
The tools that close this gap without tripling headcount already exist, and several revenue authorities outside the Caribbean are running them now.
Satellite and Aerial Property Detection
Satellite and drone imagery, cross-referenced against a jurisdiction's cadastral and valuation roll records, can flag a property whose footprint has grown, a structure that was never registered, or a commercial unit operating from an address the valuation roll still lists as vacant land. A computer vision model trained to spot the difference between what a satellite photograph shows on the ground and what a government database has on file does this across an entire parish in the time it takes a three-person team to walk one street.
Risk-Scored Delinquent Accounts
Not every arrears account is equally worth chasing. Machine learning models built on payment history, property value, business registration status, and prior compliance patterns can rank 432 delinquent properties by how much they owe and how likely they are to pay once approached, so a limited field team spends three days at the doors most likely to close, rather than working the full list in street order regardless of which accounts are actually recoverable.
Automated Change Detection Between Assessment Cycles
Jamaica's valuation roll, maintained by the National Land Agency, already underpins property tax billing. An AI system that periodically re-scans imagery against that roll, flagging anything that has changed since the last recorded assessment, catches new arrears the year they start rather than the sixth year a field operation finally reaches that street.
Predictive Compliance Nudges
Behavioural data on who tends to fall behind, and when, lets a revenue authority send a targeted reminder before an account reaches arrears rather than a field officer after. TAJ already runs online payment registration; pairing that channel with a model that flags an account drifting toward delinquency turns a reactive door knock into a preventable notification.
The Core Argument
KSAMC and TAJ cannot staff a permanent three-zone field operation in every parish. They can build a system that tells three field officers exactly which doors in any parish are worth knocking on, and that system does not need six years to notice a property has gone quiet.
The Precedent: France Found 20,000 Undeclared Pools From a Screen
France's tax authority is the clearest working example of this at scale. Working with Google and the IT firm Capgemini, French officials built a system that cross-checked satellite imagery against declared property records across nine regions in a 2021 trial, looking specifically for swimming pools that had never been registered for tax purposes. The system found more than 20,000 of them and recovered an estimated 10 million euros in a single year, at a tax rate of roughly 200 euros annually per typical pool. France then took the same computer vision approach nationwide and extended it beyond pools to undeclared extensions, verandas, and permanent pergolas, with officials projecting up to 40 million euros in additional annual revenue once the full rollout was complete.
Nobody walked a single street to find those 20,000 pools. A model looked at imagery the government already had access to, compared it against records the government already maintained, and produced a list of addresses worth a second look. That is the same architecture Half Way Tree's 432 delinquent properties would need: existing satellite coverage, an existing valuation roll, and a model built to notice where the two disagree.
Why This Is a Caribbean-Wide Opportunity, Not Just a Jamaica Problem
Four days before KSAMC's teams walked into Half Way Tree, regional tax administrators were sitting in Georgetown making the same argument from a different angle. The Caribbean Organisation of Tax Administrators opened its 27th General Assembly and Technical Conference in Guyana on 27 July 2026, and Guyana's Finance Minister, Dr Ashni Singh, told the assembled tax authorities to embrace artificial intelligence, big data, machine learning, and analytics tools as standard operating practice, not experimental add-ons.
Singh pointed to the Guyana Revenue Authority's own electronic filing app as evidence of what digital adoption can do when a Caribbean revenue authority commits to it, describing an "astronomical take-up" rate after launch and framing AI-driven data mining as offering a speed and sophistication tax administrations across the region have never had. That is not a hypothetical pitch. It is the head of one CARICOM finance ministry telling a room of his regional counterparts, weeks before KSAMC sent field teams door to door in Kingston, that the manual model most of them are still running has a ceiling.
Jamaica's own arrears history proves the point. St James, Portmore, and St Catherine each needed a dedicated field phase to lift compliance, and each phase consumed weeks of officer time that a satellite-and-risk-scoring layer could have front-loaded into a target list before anyone left the office. Extend that gap across every parish in Jamaica, then across every municipality in every CARICOM member state running the same manual model, and the aggregate revenue sitting uncollected because nobody automated the detection step likely runs into the billions, not the hundreds of millions, across the region.
What Should Happen Next
Three moves would turn Half Way Tree's field operation from a one-off recovery exercise into a repeatable system.
First, KSAMC and TAJ should pair the next phase of this 11-month revenue drive with a pilot satellite and change-detection layer run against the National Land Agency's existing valuation roll and iMapJamaica imagery. The infrastructure is largely already in place. What is missing is the model that compares the two and flags what has drifted out of sync, which is a far cheaper build than a fourth or fifth parish-wide field deployment.
Second, TAJ's target of 85 percent compliance by March 2027 should come with a parallel target for how many of Jamaica's 14 parishes have a risk-scored delinquent account list before their own field phase begins. A ranked list turns three days of officers walking every street into three days of officers visiting the addresses most likely to pay, without reducing the headcount doing the visiting.
Third, COTA should treat its own July 2026 call to action as a mandate rather than a closing remark. A shared regional model, or at minimum a shared framework, for satellite-based property change detection would let smaller CARICOM states without Jamaica's field capacity access the same detection layer without each building it alone, the same logic that already underpins regional cooperation on customs and electronic filing systems.
The Knock That Should Not Take Six Years
Half Way Tree's $162 million did not accumulate because KSAMC and TAJ were not trying. It accumulated because the only tool available to notice a property had gone quiet was a person eventually walking past it. That person showed up in August 2026, three years, four years, six years after some of these accounts first fell behind, and did excellent, necessary work once they arrived.
The fix is not replacing that officer. It is making sure the next Half Way Tree gets noticed in year one, by a system that is already watching, instead of year six, by a field team that finally reached that street.
Frequently Asked Questions
Why did KSAMC and TAJ send field teams into Half Way Tree in August 2026?
The Kingston and St Andrew Municipal Corporation and Tax Administration Jamaica ran a three-day field operation from 20 to 22 August 2026 to recover outstanding property taxes, trade licence fees, and signs and billboards approvals from commercial operators across Half Way Tree, St Andrew. Teams worked three zones covering Hagley Park, Eastwood Park Road, Red Hills Road, Hope Road, Constant Spring Road, Waterloo Avenue, Half Way Tree Road, Maxfield Avenue, Winchester Road, and Ruthven Road, engaging property owners and business operators street by street.
How much money does Half Way Tree owe in unpaid property taxes?
Excluding the current 2026 to 2027 financial year, 432 properties in the Half Way Tree area carry property tax arrears going back six years, totalling $162 million. Of those 432 properties, 151 are tied to businesses and account for $96 million of that total. A further 195 businesses in the area carry an estimated $48 million in unpaid trade licence fees, taking the combined shortfall in this one commercial district past $200 million.
How is Jamaica currently finding tax-delinquent properties?
KSAMC and TAJ are using a door-to-door field operation, three zones of officers walking Half Way Tree's commercial streets to engage individual property owners and business operators in person, backed by a temporary customer service desk at Constant Spring Arcade. This follows earlier phases of the same 11-month revenue drive in St James, Portmore, and St Catherine, where compliance reportedly rose from as low as 25 percent to 75 percent using the same in-person model.
How could AI improve property tax collection in Jamaica?
AI-based property tax collection combines satellite and aerial imagery, cross-referenced against cadastral and valuation roll records, to flag unregistered structures, undeclared extensions, and properties whose footprint has changed since their last assessment, without a single site visit. Paired with machine learning models that score delinquent accounts by likelihood and size of recoverable revenue, this lets a revenue authority send its limited field officers to the properties most likely to pay, rather than working through a list street by street in the order the streets happen to run.
Has any government actually used AI to find unpaid property taxes?
Yes. France's tax authority worked with Google and Capgemini to build a satellite imagery and machine learning system that cross-checked aerial photographs against declared property records across nine regions in a 2021 trial, uncovering more than 20,000 undeclared swimming pools and recovering an estimated 10 million euros in a single year, with a nationwide rollout projected to add up to 40 million euros more by also catching undeclared extensions, verandas, and permanent structures.
What did the Caribbean Organisation of Tax Administrators say about AI in 2026?
At COTA's 27th General Assembly and Technical Conference in Guyana, opening 27 July 2026, Guyana's Finance Minister Dr Ashni Singh told regional tax administrators to embrace technological and digital developments including artificial intelligence, pointing to big data, machine learning, and analytics tools as the route to data mining at a speed and sophistication tax authorities have not had before. He cited the Guyana Revenue Authority's electronic filing app, which he said achieved an astronomical take-up rate after launch.
Is Jamaica's tax arrears problem unique, or does it affect the wider Caribbean?
It is regional. Jamaica's own KSAMC and TAJ have already run similar door-to-door revenue drives in St James, Portmore, and St Catherine before reaching Half Way Tree, and COTA's July 2026 assembly in Guyana was convened specifically because tax administrations across the Caribbean share the same structural gap: outdated cadastral records, thin field staffing, and revenue authorities that still rely on manual canvassing to find money that better data could locate from a desk.
Would AI replace the KSAMC and TAJ field officers currently doing this work?
No. AI would change what those officers do with their time, not remove them. Satellite detection and risk-scoring narrow a list of thousands of properties down to the smaller number most likely to owe money and most likely to pay if approached, so the same three field teams that spent three days walking every street in Half Way Tree could instead spend those three days at the doors most worth knocking on, with the desk-based analysis already done before anyone left the office.