Smartstream’s Affinity: Closing the Reconciliation Gap

At Sibos 2025, Thomas Steinborn, Chief Product Officer at Smartstream explains why its AI-led reconciliation product, Affinity, is gaining traction. In large financial operations, rules and exact matching clear most items automatically, but even a tiny leftover rate can mean thousands of records for humans to resolve. Affinity targets that tail: across live clients, it’s handling roughly half of the previously unresolved records. In one deployment, that cut manual effort from about 40–45 hours to roughly three, close to a 10× productivity gain.

Before leaning into AI, Steinborn stresses the basics: firms need a data strategy first. Smartstream has invested in pulling in both well-structured financial data and the data customers increasingly rely on PDFs and other unstructured documents. The aim is to clean and prep this information so the models can actually deliver the kind of lift described above, rather than producing unusable outputs.

Even with strong automation, exceptions don’t disappear. Many cases still require outreach to third parties and follow-up investigations and which is where Smartstream sees “agentic AI” making a step-change, software that can take initiative across workflows, not just classify or predict. SmartStream is building capabilities to orchestrate that work, drawing on shared knowledge across users and, where appropriate, across customers on the platform, to shave off more hours from exception handling.

It’s not just reconciliations. Payments operations, where billions of transactions flow daily, still see a small but meaningful slice (around 3–5%) that need external resolution. Steinborn highlights Smartstream’s Smart Payments solution for advanced payment controls and exception management, automating the back-and-forth with counterparties and ingesting responses automatically or semi-automatically to speed investigations.

Looking ahead, Steinborn expects AI adoption to deepen through 2026 and beyond, with agentic patterns becoming mainstream and also flags distributed ledgers as an area to watch. One example under review: representing collateral as on-chain assets governed by smart contracts. The direction of travel is clear, the operations stack is getting more automated, more data-aware, and increasingly capable of acting on its own to resolve the hard cases that used to require a queue of people.