Banks must push AI beyond pilots to boost SME support and grow revenues
SMEs are eager to sign up for digital services that use AI to improve the efficiency and profitability of their businesses. But many banks are keeping AI applications, which might support SMEs, on the back burner. They’re reluctant to move successful pilot projects into production.
In partnership with
Ernst & Young
At EY, the insights and quality services we provide help build trust and confidence in the capital markets and in economies the world over.
Executive summary
SMEs are eager to sign up for digital services that use AI to improve the efficiency and profitability of their businesses. But many banks are keeping AI applications, which might support SMEs, on the back burner. They’re reluctant to move successful pilot projects into production.
Delays in scaling AI applications put SME banking revenues at risk. Scores of tech-savvy fintechs are already offering small businesses a wide range of innovative products and services. And competition is rising.
To explore how banks can scale AI and better support SMEs, the Qorus SME Banking Community and EY hosted an online event that featured experts from across the financial services industry. Alongside industry specialists James Sankey and Matt Cox from EY speakers included Bruno Rosevics of Banco Bradesco, Michel Branco of Banco do Brasil, Pavel Januška of ČSOB in the Czech Republic, and Bruno Reggiani of Italian fintech Tot.
Banks making the greatest progress with AI are already focusing on the financial returns they can generate from these new technologies, says EY’s Sankey. This value focus unlocks AI’s transformative capability.
Key points:
- Banks risk SME revenue by leaving AI in pilots while fintech competitors roll-out innovative digital services.
- Business strategy should set the agenda for the development of AI applications, not technology priorities.
- Establish governance frameworks early by building guardrails, engaging risk teams, and enforcing data controls from day one.
- Banco Bradesco has run more than 400 AI experiments and implemented over 60 applications while Banco do Brasil’s Ari AI assistant serves 100,000 SMEs and delivers 60 million recommendations each month.
- EY reports that 90% of CFOs would use an AI financial advisor and 86% an AI treasury assistant.
Check out the live poll results!
Banks need to stop treating AI as the responsibility of their technology departments and instead make it part of their business strategies, says James Sankey, EMEIA Corporate, Commercial and SME Banking Leader at consulting firm EY.
“Banks that are making the greatest progress with AI are thinking about what return they can get from it. They’re focusing on the value potential of AI. That helps unlock ideas about AI’s true transformative capability.”
The main obstacles stalling the scaling of AI applications are legacy systems and infrastructure limitations together with regulatory or risk management concerns. That’s according to executives polled at the event. The complexity of managing disparate and distributed data is also a common barrier, although some banks are beginning to use AI to remediate credit data and loan agreements.
EY research shows that business leaders are increasingly willing to embrace AI. Around 90% of company CFOs and treasurers say they'd be comfortable working with an AI financial advisor, while 86% would welcome working with an AI treasury assistant.
“A couple of years ago we surveyed CFOs and treasurers and just 15% said they were considering using AI. That’s changed dramatically,” says Matt Cox, Global Corporate, Commercial and SME Banking Consulting Leader at EY.
Eight initiatives to push AI pilots into production
The consulting firm points to eight initiatives that can help banks scale AI.
1. Pivot to impact: Shift from back-office tasks to customer-facing experiences.
2. Empower the business: Make business leaders owners of the AI strategy.
3. Measure what matters: Track ROI, even with directional metrics.
4. Build for scale: Create reusable AI capabilities on a common platform.
5. Fix data: Use AI to improve data quality and reduce manual work.
6. Rethink infrastructure: Rebalance cloud and on-premises infrastructure as generative AI (GenAI) matures.
7. Future-proof talent: Define and onboard critical skills.
8. Manage the risks: Engage risk teams early and update controls as models evolve.
Integration is key to maximizing the potential of AI, says Sankey.
EY’s Cox adds that banks should look to consolidate AI pilot projects onto a common platform before attempting to put them into production.
“Look for use cases with the same data model, the same LLMs (large language models), the same decisioning logic and then refactor to use those commonalities on a standard platform.”
Cox adds that banks should centralize the design of AI systems and align development with the organization’s business needs.
“You don't want to stamp on or trample innovation. You want teams to go out and try different things. The key is how do you then pull that back into a core platform. That has to be done centrally. It can't be done by a bunch of discrete groups.”
Brazilian banks are successfully scaling AI
Banco Bradesco in Brazil has successfully implemented a centralized approach to scaling AI that builds applications on a core internal platform with well-defined guardrails. The bank has embarked on more than 400 AI experiments and developed close to 60 applications that serve over 30 business units.
He adds that Bradesco has created an in-house AI development environment, open to all employees, that gives users access to technologies such as Gemini, DeepSeek, and Llama to create applications or enhance their current systems. It has enabled the development of the bank’s AI-guided digital assistant, BIA, that provides product guidance and process support to customers and employees across multiple digital channels.
At Bradesco, a central governance team creates and monitors guardrails to ensure correct data controls while the business units are responsible for implementing them for their applications.
“Once we create a guardrail for a given AI case, it makes it safer and easier to use in other areas in the bank.”
Michel Branco, Product Manager at Banco do Brasil, says it’s important to establish governance at the beginning of the development of AI applications.
During the development of Ari, its AI-guided digital assistant for SMEs, the bank ran more than 4,000 risk detection tests following responsible AI frameworks and LLM best practices. The tests identified risks such as prompt injection, data repetition, and misinformation, and enabled the bank to implement guardrails to mitigate them. The application now operates with input/output sanitization, automatic redaction of sensitive data, and continuous monitoring to ensure it complies with Brazil's data regulations.
“Ari already produces over 60 million recommendations a month and helps more than 100,000 small business clients with tailored insights on credit, cash flow, and sales,” says Branco.
Customer service top GenAI use cases
While the arrival of GenAI three years ago triggered a wave of pilot projects, lots of banks had already begun implementing production systems that incorporated earlier AI technologies. These traditional AI systems tend to analyze patterns in data to make predictions, classifications, or decisions rather than create new output like GenAI.
Many banks serving SMEs have implemented traditional AI in core back-office functions such as risk management and transaction processing.
ČSOB in the Czech Republic has deployed AI in back-office functions that include fraud detection and the processing of financial statements to reduce human intervention, improve accuracy, and increase processing speeds.
Beyond back-office automation, ČSOB is implementing AI systems to support its employees. The bank has developed an AI-powered lead generation system, Kate, that analyzes client data and behavior to provide recommendations across multiple product lines.
The bank is also deploying GenAI to improve employee efficiency and to develop MI Kate, a retrieval augmented generation (RAG) system that helps staff navigate policy instructions through a chat interface.
"AI for bankers is really important for us. It will improve how our bankers work with our biggest clients," says Januška.
Early adopters of GenAI have been eager to apply those technologies to front-office functions such as relationship management and customer experience.
Executives polled at the online event reported that GenAI systems for customer service and support were their organization’s most common AI application. They ranked them well ahead of AML and fraud prevention, customer data aggregation, and product recommendations.
Challenger banks are unfettered by legacy constraints
Challenger banks have been quick to see the potential of AI to enhance their customer service. Italian fintech Tot, which is working to secure a banking license, aims to use AI to streamline the onboarding of clients and to provide value-added services such as tax calculations.
Bruno Reggiani, Co-founder & COO at Tot, says the firm plans to use AI to enhance efficiency as well as improve customer service.
Reggiani adds that Tot as a new entrant to the banking market has an advantage over traditional finance providers because it is not encumbered by legacy systems. Governance, however, requires careful navigation.
"Italy is pretty strict but things are changing. We are starting to get direct dialogue with the Central Bank, so we think things are opening up."
Banks that have successfully moved from piloting AI projects to scaling production systems have pursued their business goals rather than technology objectives, established governance frameworks early, and built core platforms that speed up the roll-out of proven applications. Such progressive banks will gain an edge on competitors by providing SMEs with products and services that are convenient, tailored, and quick to respond to changing needs.
Want to know more about scaling AI in your organization?
Join more than 150 leaders, from over 80 banks and finance providers in close to 35 countries, as a member of the Qorus SME Banking Community. Gain access to best practices shared by community members through podcasts, virtual events, articles, and our extensive digital library of more than 2,000 banking innovations. Discover from leaders in SME banking how business banking is evolving and embracing new technologies to meet the needs of small businesses and open new revenue opportunities.
SME Banking community
With Qorus memberships, you gain access to exclusive innovation best practices and tailored matchmaking opportunities with executives who share your challenges.