The supply chain industry faces a crisis that traditional business intelligence was never built to solve. When a regulatory shock lands — a court ruling, a tariff reversal, a new Customs and Border Protection deadline — the answers companies need are contained in spreadsheets, government portals, entry histories and tariff schedules that were never designed to connect. Faced with high-stakes decisions on a government clock, teams scramble and hope nothing slips through the cracks.
That’s where dashboards reveal their structural limit. Business intelligence is organized around predefined questions and schema, while regulatory shocks are open-ended, customer-specific, rule-bound and constantly shifting. The gap is a problem of basic architecture.
In early 2026, a major change in U.S. tariff policy triggered a refund wave worth billions of dollars and introduced new processing rules and deadlines for import entries. For customs brokers, it instantly became a high-volume math problem: compute refunds across hundreds of clients and tens of thousands of historical entries, under timelines the government was prepared to enforce.
RL Jones Customhouse Brokers has served the U.S. southern border since 1938. The firm employs more than 350 people, operates from every major U.S. southern port from Los Angeles to McAllen, Texas, and is known in the industry for trade-policy fluency and intense line-item accuracy. When the Supreme Court invalidated tariffs imposed under the authority of the International Emergency Economic Powers Act (IEEPA), RL Jones's customers needed clear answers fast: which entries were impacted, which duties could be reimbursed, whether protests were required, what deadlines applied, and how much money could potentially be recovered.
The starting procedure was almost entirely manual. An analyst would log into CBP’s Automated Commercial Environment (ACE), download entry data for a single client, paste it into a working spreadsheet, and run a sequence of cross-reference lookups against tariff schedules, paid duty amounts, entry summary dates and liquidation status. Each customer required its own workbook. Each workbook required hours of careful formula work. Layered on top was a calendar problem: Consolidated Administration and Processing of Entries (CAPE) phase windows, liquidation timing, and per-entry deadlines that varied by customer.
The biggest limitation was structural. Government information lived in one world; RL Jones's internal operational environment lived in another. There was no unified intelligence layer connecting government data, internal databases, customer entry histories, reimbursement logic and protest tracking. Critical analysis required hours or days of manual work, was error-prone in exactly the places where errors are most expensive, and the volume made keeping up impossible.
“We needed an artificial intelligence solution to provide our decision-makers with real-time insights to navigate these evolving requirements effectively,” says Michael Rullis, general manager of RL Jones.
The instinctive response, when reporting feels slow, is to add a dashboard. RL Jones evaluated that path and quickly recognized the mismatch. Traditional BI assumes the questions are stable enough to be modeled, that the schema can be agreed in advance, and that the consumer of the data knows which visualization they need. The IEEPA refund problem broke all three assumptions. Each customer's exposure was different. Each protest deadline depended on entry-level facts no analyst had seen yet. The schema had to span ACE exports, internal systems, tariff schedules and reimbursement logic. And the question coming in from a client on the phone was rarely one a BI engineer had pre-built a chart for.
What RL Jones needed was a reporting layer — a system architected from the ground up to take a natural-language question and return a sourced, decision-ready answer drawn from the underlying data.
Amberd.ai is a private, large language model-native decision platform for data-intensive enterprises. Instead of layering a language model onto legacy BI, it combines structured and unstructured data inside a customer-controlled environment and answers executive questions in real time.
For R.L. Jones, that meant a simple pattern. Sensitive ACE data moves on a controlled schedule into the broker’s own cloud environment, preserving data ownership and the privacy posture required for regulated trade information. Amberd’s private LLM operates against that data in place, performing the lookup work, tariff logic and refund calculations that analysts once stitched together by hand.
On top of that, Amberd built an intelligence workflow that unifies entry status, appeal and protest deadlines, tariff changes, reimbursement eligibility, historical customer entries, transportation impacts, and operational exceptions in a single searchable environment. Scheduled reports tied to the regulatory calendar notify staff automatically when a client is approaching a CAPE phase deadline or other CBP cutoff, with the underlying entry-level detail attached.
Day to day, the change shows up in how analysts work. They ask questions in natural language such as:
- Which customers are eligible for reimbursement based on the reversed tariff changes?
- Show all impacted entries for a specific importer.
- Which protests are approaching deadlines?
- Which entries changed status in the last seven days?
- Generate a detailed export for all affected shipments tied to a specific HTS code.
The platform correlates the required data and returns structured answers as detailed tables, exportable comma-separated values (CSV) reports, customer-specific datasets, and operational summaries, all with line-item traceability back to source ACE data. Work that took hours or days now completes in seconds, and the implementation itself went from kickoff to production in three weeks. Manual ACE downloads, per-customer Excel workbooks and lookup-table reconciliation have been retired from the refund workflow. Decisions are faster, time savings are significant, and senior brokers can focus on judgment-heavy trade-policy work rather than spreadsheet maintenance.
“The ability to access accurate operational intelligence and detailed reporting in seconds instead of hours has had a major impact on our workflows, our responsiveness, and ultimately the level of service we provide to our customers,” says Eduardo “Lalo” Acosta, vice president with RL Jones.
IEEPA was the trigger, but the problem it exposed is structural and industry-wide. Government portals are disconnected from internal systems. Compliance workflows are manual. Operational data is fragmented across enterprise resource planning, transportation management systems, customer files and regulator submissions. Reimbursement math is complex, volumes are high, and the records that matter are structured in ways that resist conventional BI.
Every regulated supply chain will face its own version of this moment, whether in pharmaceuticals, food, energy, automotive or cross-border logistics. The companies that pull ahead will be the ones that stop trying to retrofit dashboards around regulatory whiplash and instead build a decision layer designed for it: customer-owned data, a private model that reasons in place, line-item traceability, and answers in natural language at the speed of the question.
The lesson from RL Jones is that AI built for decisions is a different category of system — and the supply chain industry’s next round of regulatory surprises is when that distinction will start to matter.
Resource Links:
RL Jones Customhouse Brokers: https://www.rljones.com/
Amberd.ai: https://amberd.ai/
.png?auto=format%2Ccompress&fit=crop&h=141&q=70&w=250)
