While creating our Smart Assistant alongside the IT team, one thing was clear from the beginning: Conversational AI works great across many sectors, but some banking apps that have integrated it suffer from the infamous hallucinations of LLMs. What’s missing?
In this post, I’ll walk you through it.
Juan Carlos López
Co-founder & CPO of Coinscrap Finance
Conversations in digital banking are changing the user experience
Since tools like ChatGPT came onto the scene a few years ago, we’ve embraced a new way of interacting with technology. And it’s not just me saying it! We’re seeing it with every single person who has access to the internet – around 5.56 billion people, according to the latest We Are Social: “Digital Report 2025”.
Today, digital interaction means talking to AI. It’s something natural, yet many banks still haven’t included it in their services. Those that have – check out the top ones here – face challenges like long processing times, high costs, and inaccurate outputs due to data scarcity.
What does generative AI-powered conversational banking offer?
It’s not just about chatting with a virtual assistant. At Coinscrap Finance, we want users to feel like their bank truly understands them. And even better – we make sure that understanding translates into real opportunities for upselling and cross-selling.
If you know what your customer is looking for, it’s only natural that – during the conversation – you suggest a service they need, alert them to opportunities to improve coverage, adjust a current product, or access something better. All based on their financial behavior.
We’re talking about interfaces where people find exactly what they need, receive smart recommendations, and get nudges to grow their savings. But how can banks actually implement these advanced intelligent assistants? Let me tell you more about the importance of properly training the algorithm.
The power of AI trained on millions of real transactions
COCO, our proprietary engine, has spent years learning from the massive volume of banking data it categorizes and enriches every day. It’s the heart of our Smart Assistant and the key differentiator from other conversational AIs in the industry.
Thanks to COCO, our intelligent assistant can understand user intent, not just literal text. When someone asks, “How many subscriptions do I have active?”, Smart Assistant knows they’re really asking about recurring expenses that could be optimized (a savings opportunity), and it returns graphs, insights, and even personalized suggestions based on their preferences.
Now, users can talk to their bank just like they’d talk to a friend who’s great with money. No endless menus, no complicated commands, they type a question, and boom – done! That’s how our assistant works: it understands the digital banking user and delivers answers, context, and suggestions in milliseconds.
All with precision, thanks to a leading engine, with 95% accuracy in categorization. Transactions are its fuel, and it draws from financial history – meaning no hallucinations or errors. Satisfaction guaranteed within your app. What else could you ask for?
Upselling & cross-selling opportunities with every query
Our tool goes one step further: If it detects the user might benefit from a related product or service, it suggests it naturally – never forced.
So when someone asks, “How many insurance policies do I have and how much do they cost me annually?”, not only does it respond instantly with a breakdown, it also highlights potentially interesting services at that moment.
For example, if the assistant detects several transactions at pet stores or veterinary clinics, it might suggest pet insurance. These automatic recommendations, along with summaries and visuals, empower the user to make smarter money decisions.
This is all grounded in behavioral finance. Smart Assistant uses the kind of “nudges” that made Richard H. Thaler famous – and earned him the Nobel Prize in Economics in 2017.
Behavioral science tells us people tend to choose short-term rewards over future gains, and are prone to cognitive biases that limit their financial potential. Banks have the chance to address this – and gain a competitive advantage – by embracing smart tools like this.
🔎 Do you know that…
“Smart Assistant is able to understand the user and deliver responses, context, and suggestions in milliseconds – with 95% accuracy, thanks to our leading categorization engine”.
And for tech teams… seamless implementation, zero headaches
Integrating Smart Assistant into banking environments is as smooth as placing the final puzzle piece. Our solution is designed to fit like it’s always been part of the system. Its flexibility and ease of connection make it the perfect companion for any financial institution aiming to deliver a first-class digital experience – for both clients and internal teams.
A frictionless experience in the banking app or chat
Smart Assistant can be embedded as a widget directly within the bank’s app, in any section you choose, and without complex development. Of course, it’s fully customizable: colors, logos, fonts… everything can match your corporate visual identity.
If your institution already has a chat system, the assistant can be integrated into that channel, handling frequent queries, banking operations, and even product onboarding – all in a single conversational thread.
Prefer a more technical route? No problem. Smart Assistant is also available via REST API, making it easy to embed its features into custom workflows or connect it with your bank’s proprietary systems in a modular, agile way.
Smart Assistant turns bank employees into efficiency machines
The integration isn’t just for customers. Internal teams also benefit – the assistant can connect directly to the bank’s dashboards and internal tools. This way, any employee can access user information in real time, view relevant transactions, or explore key insights – all without leaving their usual workspace.
Normalized, ready-to-use responses
One of Smart Assistant’s core strengths is that all its responses are delivered in a structured, normalized format: JSON. This allows for direct reading by other systems, whether on the app’s frontend or internal platforms. This standardized format provides:
- Agility: technical integration is much faster.
- Simplicity: no need to transform the data for system compatibility.
- Total Compatibility: fits any existing tech architecture.
- Scalability: easy to maintain and constantly evolving.
Moreover, all enrichment logic is pre-processed: the AI doesn’t touch raw data or access the full database, ensuring regulatory compliance and total privacy.
True personalization and operational cost savings for the financial industry
We’ve seen it firsthand: when a conversational interface is well-designed, the impact is real – support calls drop, tickets decrease, and there’s no need to visit a bank branch. Everything becomes smoother and faster.
We’re talking about convenience for both users and financial institutions: customers solve their issues with a simple question to Smart Assistant, and banks finally gain a real understanding of their users’ needs – in real time.
By combining generative AI with enriched transactional data, banks can now offer recommendations that adapt to each person’s financial behavior. It’s a technology set to completely transform the relationship between banks and their users.
Don’t just take our word for it,
Check out the keynote
from David Conde, CEO and Co-Founder of Coinscrap Finance, together with Alberto Ceniz, Head of Financial Services at Google Cloud Spain:
About the Author
Juan Carlos López Díaz is Chief Product Officer and co-founder of Coinscrap Finance. In 2016, together with David Conde and Óscar Barba, he created Txstockdata and Coinscrap Finance. After the tremendous success achieved, the business pivoted towards B2B, in partnership with EVO Banco and its “Smart Piggy Bank.” As a developer, he has over 8 years of experience leading major projects.
Along with his team, he is capable of creating the best tools for the financial world. From the product department, Juan Carlos has delivered projects for major companies in the sector: Evo Banco, Santander, Caser, Mapfre, and Bankia. He holds a degree in Electrical Engineering from the Central University of Venezuela and an iOS App Development certification from U.N.E.D and U.C.A.M.
