Evaluating AI vendors in financial services, on the record.

These two sell to banks from opposite ends, and on the cost question only one has published an answer. Cotribute grows deposits, loans and membership at credit unions and community banks by layering account opening, lending applications and three AI Growth Agents onto the core an institution already runs, while Bottomline moves more than $16 trillion of business payments a year and sells banks a fraud platform that can hold a payment in flight. Neither is an AI company at heart. Without the models, one is still a working origination platform and the other a payments utility. On cost, Cotribute publishes its whole basis of charge: an annual platform fee tiered by assets, modules priced as line items, core integrations included and no metering of applications, with two list prices public. Bottomline publishes no price, unit or tier for any of its four businesses. Cotribute also documents its customer evidence and integrations in detail, with named credit unions reporting quantified outcomes and real time connections named down to each core product, where Bottomline documents both in part. Bottomline's standing is scale: more than 800 financial institutions, over a million businesses on its payments network and 23 years of audited public company reporting. Its weakest point is liability. A wrongly held payment has a cost, and nothing published says who bears it.

Only one of these is customer facing, and the security question leans the other way. Personetics reads transaction data for more than 150 million active monthly banking customers and turns it into proactive insights and offers. Titan works behind the counter, running agents across compliance, underwriting, risk and operations with people keeping the final decision. A buyer looking for AI that speaks to customers has one candidate in this pair. A buyer weighing data handling gets a clearer picture from the other. Titan reaches foundation models through a private interface, and what its customers share is banking domain knowledge rather than institution records. Personetics says little about privacy or data handling, while assembling a detailed financial picture of ordinary people, open banking data included. Personetics answers with scale, documenting customer results and segment coverage across 30 markets, where Titan, less than a year out of stealth, names no customer. Titan documents oversight in detail, with step level reasoning, universal logging and a stated human decision point. AI is central to both, and neither publishes a security attestation or trust center.

Two long running credit decision engines, and in both AI is a layer rather than the core. They execute the lender's own rules and scorecards, and the models sit above an engine that came first. Zoot has run hosted decisions since 1992, GDS Link since 2006, and in both the lender owns the logic. Past that, Zoot publishes more. It connects hundreds of live data sources built over three decades, runs its main data center in a building it owns, holds PCI DSS certification among other security credentials, and quotes three European customers by name. GDS Link brings more than 200 data sources with attributes already defined and material for five kinds of lender, and publishes no commercial terms at all. Zoot's own gap comes from its best feature: business users can change live credit rules without engineering, and nothing published describes the change control around that.