Zone & Co · Treasury Light
5 months · native inside NetSuiteA conversational CFO, grounded in NetSuite ledgers
Finance teams ask Treasury Light questions like “what if we delay payables 45 days?” and get back a quantified forecast, a cited chain of reasoning, and a path to export. Under the hood: Claude inside the Agent SDK, orchestrating forecast tools, reconciliation data, and a transparency layer Zone’s leadership called the “golden goose.”
“Golden goose.”
— Zone & Co, on the transparency layer
Finance teams running on NetSuite were forecasting cash flow in Excel. Every week, someone manually exported transactions, built a spreadsheet, and hoped the numbers were right. Zone had the reconciliation data — they just had no way to turn it into forward-looking intelligence. The opportunity: become the only NetSuite partner with AI-powered treasury management built directly into the ERP.
What We Built
Liquidity forecasting that actually works
30, 60, and 90-day cash flow forecasts with day-by-day breakdowns. The system combines pattern recognition on recurring payments with AI models for unpredictable transactions — giving finance teams forecasts they can trust, not just hope are right.
Filter by subsidiary, bank account, vendor, or transaction type. Toggle between forecasting methods. Every prediction comes with a confidence score so you know exactly how much to trust it.
Scenario planning in plain English
A conversational interface where finance teams ask questions like “what if we delay payments 45 days?” and get instant, quantified answers. Not a generic chatbot — it runs the actual forecast comparison and shows the real impact.
Thread-based history means CFOs can build on previous scenarios, test cascading decisions, and compare outcomes side by side — all without touching a spreadsheet.
Working capital that manages itself
The system watches payment patterns and proactively surfaces recommendations: which invoices to chase, which early payment discounts to take, how to prioritize payments to hit your target number.
It categorizes vendors by risk, identifies customers with slipping payment behavior, and tells you what to do about it — before it becomes a problem.
Full transparency — the “golden goose”
Every forecast shows exactly how it was generated — which data sources, what methodology, which transactions contributed. Finance teams and auditors can trace any number back to its origin.
Confidence scoring on every prediction. Historical variance analysis so the system gets better over time. This is what Zone called the “golden goose” — the transparency layer that made AI forecasting trustworthy enough for regulated industries.
The Claude Stack
Orchestrator
Agent SDK · scenario loop
A CFO-facing supervising agent holds thread state, decomposes questions into forecast calls, and keeps a citation trail that auditors can walk.
Reasoning Tier
Opus · extended thinking
Used when a scenario branches into cascading decisions — the model that earns its keep on ‘what if we delay payables and also pull forward AR.’
Bulk Tier
Sonnet · primary workhorse
Most forecast summaries, vendor categorization, and confidence-score explanations run on Sonnet, with the forecast engine providing the numbers.
Tools
NetSuite · forecast engine
Tool use reaches straight into NetSuite saved searches, the deterministic forecast engine, and Zone’s reconciliation store. Claude reasons about the answers; it doesn’t invent them.
Grounding
Structured outputs + citations
Every answer comes back with the transactions, saved searches, and assumptions that produced it — the transparency layer Zone called the ‘golden goose.’
Caching
Prompt caching on ledger context
Multi-turn CFO threads stay cheap because the account structure, subsidiary map, and chart of accounts are cached across turns.
The Scenario Loop
How “what if we delay payables 45 days?” turns into a defensible number
Intent parser
Claude reads the CFO’s natural-language question and classifies it: horizon, subsidiaries, bank accounts, vendors, filters. Emits a structured scenario spec.
Data fetcher
Pulls the actual ledger slice from NetSuite — saved searches, transaction history, open AR/AP — through a typed tool layer. No ledger data is hallucinated; it’s quoted.
Forecast invoker
Calls the deterministic forecast engine with the structured spec and receives 30/60/90-day projections, confidence bands, and vendor-level breakdowns as JSON.
Scenario comparator
For branching questions, spawns a subagent that runs the same scenario under alternate assumptions and diffs the outcomes — this is where extended thinking earns its cost.
Explainer + citer
Writes the answer back in plain English, with every number linked to the transactions, assumptions, and confidence scores that produced it. This is the transparency layer.
Why Claude
Finance won't sign off on a black box
Every number in a CFO’s scenario needs a walk-back path. The thing that makes Treasury Light shippable to regulated industries isn’t the forecast accuracy — it’s the explainability. Claude produces answers that stay grounded in the actual tool outputs, and calibrates its language to the confidence the engine reports, rather than confidently filling gaps.
Tool use reliability on Claude is what let us make the “no invented numbers” rule enforceable at the system level. If the model can’t cite a transaction, it won’t render a number.
Outcomes
Market Position
Zone became the only NetSuite partner offering AI-powered cash flow forecasting natively inside the ERP. Not a third-party integration — a native capability that lives where finance teams already work.
Revenue Expansion
Treasury management functionality added to existing Zone Reconcile licenses, creating upsell opportunities across the entire installed base without separate licensing complexity.
Excel Eliminated
Finance teams stopped exporting transactions to spreadsheets. Forecasting, scenario planning, and working capital management all happen inside NetSuite now — with better accuracy than manual methods.
Compliance Ready
Built-in audit trails, forecast explainability, and multi-region data residency. The transparency layer satisfies regulatory requirements that block AI adoption in financial services.
From bank reconciliation tool to the only AI-native treasury platform in NetSuite.
Built on Claude. Designed to compound.
LightCI partners with companies shipping production agents on the Claude Agent SDK.
Talk to LightCI