What happened
American Banker reports that Bank of Montreal (BMO) is piloting a machine learning workflow that analyzes customers' payments and accounts-payable data to generate recommendations for salespeople. American Banker says BMO is using a technology from Codat to pull payments and AP information through APIs and feed machine learning models that surface opportunities for relationship managers. American Banker quotes Rose Grande, head of North American corporate card product and programs at Montreal-based BMO: "We like to be that customer's trusted advisor, and this allows us to play that role in a more consistent way." American Banker reports the pilot will start in Canada in a couple of months and could roll out by the end of 2026.
Editorial analysis - technical context
Industry-pattern observations: integrating payments and ERP data requires connectors into accounting platforms, normalization of vendor and transaction entities, and deduplication across PDFs and spreadsheets. Bradley Leimer of Leimer One Advisors told American Banker that much of commercial-data is fragmented across ERPs, accounting platforms, PDFs, spreadsheets, and CRM notes, and that cleaning that data is a central engineering task. Production-grade recommendation systems in banking typically need strong inference explainability and audit trails to satisfy relationship managers and compliance teams.
Context and significance
Industry context: Commercial banking stores high-value signals in cash flow, payments, and supplier networks, making it a natural domain for ML-driven sales recommendations. American Banker frames BMO's effort as part of a broader push in which 66% of bankers say AI is a strategic priority this year. For practitioners, consolidating AP and payment feeds into a normalized dataset is often the highest-effort component before model training or heuristic scoring can yield reliable opportunities.
What to watch
Observers should track:
- •the pilot's integration with BMO's CRM and salesperson workflows
- •measurement of recommendation precision and conversion lift
- •data-privacy and consent handling for customer accounting data
- •how connector coverage across accounting platforms scales. Industry observers will also watch whether the implementation emphasizes human-in-the-loop review and explainability when recommendations reach relationship managers
Key Points
- 1Banks tapping payments and AP feeds can transform fragmented cash-flow signals into actionable sales leads, improving conversation relevance for relationship managers.
- 2Reliable recommendations require normalization across ERPs, spreadsheets, and PDFs, so integration and data-cleaning effort often dominates model work.
- 3Production rollout depends on CRM integration, explainability, and privacy controls; these operational elements determine practitioner adoption and measurable ROI.
Scoring Rationale
This is a notable industry deployment showing how banks operationalize ML on payments and AP data. The pilot is relevant to practitioners building enterprise ML systems, but it is not a frontier-model release or sector-wide shift.
Practice with real Banking data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Banking problems
