RedCompass Labs releases AI models fine-tuned for payments

Trained on the largest collection of payments rulebooks and scheme documents in the world and fine-tuned on decades of knowledge from leading payments experts, AnalystAccelerator 2.5 delivers cutting-edge applied AI models to help banks accelerate payments transformation.

Banks worldwide are managing complex, multi-year, multi-million-dollar modernization projects, including ISO 20022, SEPA Instant in Europe, and FedNow and RTP in the US. As payments evolve, AI is becoming integral to these efforts, with 54% of banks planning to use generative AI to support the shift to instant payments, while 42% are actively exploring its potential. However, many rely on general AI models designed for speed rather than accuracy, undermining the precision needed for these pivotal initiatives.

AnalystAccelerator 2.5’s fine-tuned AI outperforms leading general AI models, including those developed by OpenAI. It performs 13% better on complex payments-related tasks than GPT-4o and achieves a perplexity score that is 10x lower than open-source AI models.

For banks, payments software providers and system integrators, AnalystAccelerator enables them to accelerate the analysis phase of the software development life cycle for payments modernization projects by 45%. It slashes project costs by 25% and ensures projects benefit from comprehensive, accurate insights drawn from every relevant rulebook globally.

AnalystAccelerator 2.5 is model-agnostic, enabling the applied solution to operate on any available or future foundation model and ensuring it remains state-of-the-art while being optimized specifically for the payments domain.

RedCompass Labs expert knowledge curators continuously annotate payments rulebook and scheme documents with expertise gathered from decades of project work, enabling the model to improve and deliver accurate responses to payments challenges as they emerge. It also saves computing power and time spent on prompts by precisely targeting retrieval to specific payment documents rather than on all information available.

RedCompass Labs has partnered with a tier one North American bank for an AnalystAccelerator proof of concept engagement. The firm is also in late stage talks with large global system integrators, consulting firms and a range of large international banks about deploying AnalystAccelerator to help turbocharge their modernization programs.

Tom Hewson, CEO of RedCompass Labs commented: “AI is the biggest shift in payments modernization since the introduction of the computer. What once took an army of consultants weeks to solve can now be fixed by a handful of experts in a fraction of the time using AI. However, what isn’t changing is the complexity of payments modernization.

“With the global move to instant, interoperable payments and ISO 20022, the stakes have never been higher for banks. General AI tools lack precision and in payments, 99% accurate is not good enough. Payments experts ask complex, nuanced questions that need exact answers, answers that broad AI products, like ChatGPT, cannot guarantee.

“To support banks with payments modernization, we have created the world’s most comprehensive payments AI, trained on the largest collection of payments documentation in the world. Armed with AnalystAccelerator, analysts can work faster and smarter than before, while banks save millions as a result of the productivity gains.”