In the global banking sector, artificial intelligence (AI) has transitioned from the exploratory phase to strategic integration.

The 2026 Gartner CIO and Technology Executive Survey underscores this shift, with 55% of the 2,300+ banking CIOs and tech executives polled in 2025 reporting that their enterprises had already deployed generative AI, and 26% planning to deploy it within 12 months.

As banks accelerate their AI strategies, Gartner released in January a report identifying the sector’s top AI trends for 2026, highlighting increased investment in AI application development platforms, multi-agent ecosystems, and domain-specific language models (DSLMs) among the most prominent developments in the space this year.

AI application development platforms

As AI efforts scale and move deeper into core operations, CIOs will increasingly recognize the need for a shared foundation that can support the safe, secure and repeatable delivery of AI across use cases, Gartner forecasts. This will drive a shift towards modular AI application development platforms with reusable components. These platforms allow banks to build, govern, and expand AI capabilities with the speed and consistency that isolated tools cannot provide.

Leading financial institutions are already leveraging AI application development platforms. Wells Fargo has developed the Enterprise Open Source Data Science Platform, which features reusable AI building blocks; and BNY has built Eliza, an enterprise AI platform with a solutions marketplace as well as approved datasets and models to build upon.

Gartner predicts that this year, more than 40% of banks globally will make investments in AI application development platforms to scale AI securely and compete in emerging agent-to-agent ecosystems.

Multi-agent systems

As banks increasingly adopt AI agents, they are starting to experiment with multi-agent systems (MAS). MAS are networks of AI agents that collaborate or operate independently to achieve specific objectives.

Currently, adoption is concentrated among Tier 1 banks that are building multi-agent workflows for complex processes such as quantitative investing and lending. Relevant examples include MACAW, a multi-agent workflow for car buying and financial assistance by Capital One; ASK DAVID, a multi-agent workflow for quantitative investing and research launched by JP Morgan; and Project Coral, an autonomous multi-agent orchestration system for engineering workflows introduced by Commonwealth Bank of Australia (CBA).

According to Gartner’s 2026 CIO and Technology Executive Survey, 17% of banking CIOs have already deployed AI agents, and 41% plan to do so within the next 12 months.

By year-end 2027, the firm forecasts that at least 30% of day-to-day banking decisions, including loan pre-approvals, transaction anomaly detection, and dispute resolution, will be made autonomously by multi-agent systems.

Domain-specific language models

Banking CIOs recognize that while genAI and agentic AI offer significant potential for differentiation and revenue generation, building these on generic LLMs present risks of hallucinations and knowledge dilution.

DSLMs mitigate these risks by being trained and fine-tuned on curated, high-quality banking data. These models offer superior accuracy, quality and domain specificity in business-critical tasks while reducing risk and optimizing costs for banking workflows.

For banks, the specialization of DSLMs provides a defensible competitive advantage, and several institutions are already leveraging such models. These include Westpac, which has adopted Kai‑GPT, a model fine-tuned on its proprietary data specifically for financial services; Wells Fargo, which has developed a DSLM and small language model (SLM) for personally identifiable information (PII) detection integrated into its conversational AI assistant; and State Bank of India (SBI), which is creating its own DSLM.

Gartner states that by 2028, more than 60% of the genAI models used by banks will be domain-specific, up from 30% in 2025.

Physical AI

The rise of frontier‑scale foundation models, more capable embedded AI, and significant improvements in edge computing are reducing the historical constraints associated with physical automation. These technologies now empower robots and intelligent devices to operate more reliably in semi-structured environments such as branches, ATMs and service lobbies, while interacting more naturally with customers and staff.

As AI maturity improves on both the supply and demand sides, Gartner expects physical AI to play an increasingly important role in modernizing physical banking operations, a trend that’s already reflected in early deployments. China Construction Bank has deployed humanoid robots in its flagship branches to guide customers, answer queries, and assist with account opening; HSBC has introduced Pepper humanoid robots in branches across the US and Canada to greet customers, answer FAQs, and direct them to staff; and Caixa Bank has launched AI-powered ATMs equipped with facial recognition authentication.

By year‑end 2027, about 10% of banks will deploy physical AI in branches or operations, up significantly from about 3% today, according to Gartner. Adoption is projected to rise sharply afterward as proven use cases and customer familiarity drive confidence in the technology.

AI security platforms

As banks scale AI and agentic AI solutions across internal and external environments, AI security platforms are becoming indispensable. According to Gartner’s 2H 2025 Financial Services Leaders’ AI Survey, the top two risks in agentic AI investments are regulatory compliance and data privacy and security breaches, underscoring the necessity for centralized AI security platforms (AISPs).

AISPs consolidate fragmented AI risk controls, AI governance, and cybersecurity into a single architecture. They provide a unified way to secure third-party and custom-built AI applications, centralizing visibility, enforcing usage policies and protecting against AI-specific risks, including prompt injection, data leakage and rogue agent actions.

Gartner forecasts that by 2028, 75% of data breaches will be a result of AI and agentic AI adoption, resulting in stalled adoption unless banks adopt AI security platforms.

Featured image: Edited by Fintech News Switzerland, based on image by thanyakij-12 via Magnific