Duco says agentic AI cuts bank reconciliation work 76%

Duco said 12 banks and financial institutions cut manual post-trade reconciliation work by an average of 76% in production trials using its agentic AI tools.

The group included BBVA, Blackstone Credit & Insurance and CIBC Mellon.

The results came from Duco's Pacesetters programme, in which 12 clients tested the company's agentic workspace in live operational settings. The average reduction covered a range of manual reconciliation tasks within post-trade operations.

The figures showed the biggest reductions in the time needed to build reconciliation processes, which fell by more than 91%. Time spent on process optimisation dropped by 84%, while break workflow management and investigation tasks recorded savings of more than 73%.

The findings come as banks and asset managers prepare for a shorter securities settlement cycle in the UK and the EU. The move to T+1 will give firms less time to identify and resolve reconciliation breaks after trades are executed, increasing pressure on back-office teams that still rely heavily on manual processes.

Post-trade reconciliation involves checking that records match across counterparties, custodians, internal books and external systems. Breaks can arise from timing differences, missing data, formatting issues or genuine trade discrepancies. Resolving them often requires a mix of repetitive checking and human judgement.

Duco said its setup combines an agentic workspace with its existing rule-based reconciliation platform. Clients can also access the system through their own agents in a headless model via MCP.

Operational shift

Duco presented the trial results as evidence that operations teams could spend less time on routine processing and more on exceptions and decisions. The view reflects a wider push across financial services to test where generative and agentic AI tools can be deployed within controlled operational processes rather than customer-facing functions.

Christian Nentwich, Chief Executive Officer, Duco, said: "We believe that operations departments will move from heavy task based work to mainly decision making in a short period of time, using agentic AI. With agentic workspace, and through the headless model, clients get the best of both worlds."

Nentwich added: "Duco's proven rule-based platform keeps data processes predictable, repeatable and transparent, while agentic AI determines what needs doing and acts. Operations teams move from data wrangling to the judgement calls that matter most. This is revolutionising post trade operations. That's not just what we think, but what we're seeing, and what our clients are telling us."

The reference to predictable and transparent processing points to a central issue in operational AI adoption at banks: whether newer AI-driven tools can be used without weakening auditability and control. In reconciliation, firms typically need clear records of why a task was performed, what data was used and how exceptions were handled.

Client view

CIBC Mellon was among the institutions in the trial group and commented on the results. Its remarks focused on the potential to shift employee time away from manual activity and towards problem-solving work.

Jeremy Hulme, Vice President, Operational Excellence & Transformation, Client Service Delivery, CIBC Mellon, said: "What excites us about Agentic Workspace is its potential to help organizations move beyond task-based automation and rethink how operational work gets done."

Hulme added: "By combining operational expertise with agentic capabilities, we see an opportunity to reduce time spent on manual activity and give teams more capacity to focus on problem-solving, decision-making and continuous improvement."

Duco is based in London and said its platform is used by more than 10,000 users across more than 30 countries. Its customer base includes banks, investment managers, exchanges and insurers. The company said its agentic workspace is now available to new and existing clients.

The production trial figures offer one of the clearer early data points on how agentic AI is being measured in financial operations, an area where firms have often discussed potential gains in broad terms but disclosed little numerical evidence from live workflows. Duco said the numbers were drawn from production trials rather than isolated laboratory testing, with the average time reduction across the 12 participating organisations standing at 76%.