Global – October 27, 2026 -- Routing deterministic reconciliation work to an Alteryx One workflow cut token consumption by up to 95%, a 20x reduction, in a benchmark run by financial services consultancy NextWave, an Alteryx partner.
NextWave tested a 120-trade Product Control reconciliation two ways
The workload was a daily front-office-to-back-office reconciliation. It validated open trades, valuations, risk and cash balances across 120 trades, then explained differences and proposed next steps.
In the first run, an LLM received the front-office and back-office files directly and compared them unaided. In the second, the LLM called the Front Office Back Office Data Reconciliation workflow on Alteryx One through the Model Context Protocol (MCP). Alteryx executed the comparison logic, and the LLM interpreted the output and recommended next steps.
Workflow approach delivered 100% accuracy while the unaided LLM missed 2 breaks
The Alteryx One plus LLM setup consumed 20x fewer tokens than the purely stochastic approach. It reconciled with 100% accuracy, while the unaided LLM missed 2 breaks out of 120 trades.
The companies also cite governance, explainability and auditability built into the deterministic workflow layer, which they say a pure LLM comparison cannot guarantee.
Alteryx internal tests show token reductions of 83% to 93%
Alteryx points to its own internal testing, published alongside its launches of Ask Alteryx, Agent Studio, an MCP Server and Alteryx Skills. On tasks using raw, ungrounded data, it found up to 93% token reduction and an 85% speed increase. On tasks using clean, already-governed data, it found up to 83% token cost reduction and a 65% speed increase.
Alteryx ties these results to its VURA framework, which makes business logic visible, understandable, repeatable and auditable, so AI agents draw on governed logic rather than re-deriving it each time.
Office of the CFO use cases drive the pattern for financial services clients
NextWave and Alteryx are applying the split to reconciliation and other Office of the CFO problems for financial services clients. Deterministic workflows handle comparison and validation, and LLMs are reserved for reasoning, summarization and next-best-action recommendations. The partners also use LLMs to speed adoption of the Alteryx One platform itself.
Alteryx directs readers to its Month End Close Acceleration AI-Ready Starter Kit as a reference implementation.