Banks are embracing AI to drive smarter processes and better customer experience – but research suggests 55 per cent still see the limitations of their core system as an obstacle to achieving their business goals. It doesn’t have to be that way. For the frontier banks rearchitecting themselves around AI, the core is still central to their ambitions.
“Operating models are changing as work is redesigned around outcomes instead of processes,” says Tyler Pichach, Microsoft’s global head of AI strategy and go to market, payments and banking. “Banks can now modernise around the core, expose reliable capabilities through APIs and use agentic orchestration to turn fragmented processes into end-to-end execution. The core remains the trusted system of record, but it is no longer the only place where innovation can happen.”
Banks build their business on trust, and protecting it is a key principle of this transformation. That means navigating a shifting trust boundary, as autonomous technology changes the way data, users and transactions move between systems and services.
“When an agent can learn, call other agents and act through systems, the trust boundary ceases to be the login screen and becomes every decision made by the agent on behalf of the bank,” says Emmanuel Baviere, financial services frontier industry advisor at Microsoft. “Three elements move within this boundary that were once outside: the data the agent reads, the reasoning it applies and the actions it triggers. Protecting all three starts by treating agents as first-order identities with rights and duties. Know Your Agent principles are essential. In addition, Microsoft Entra Purview and a control layer such as Agent 365 exist to provide security, compliance and oversight for an entire fleet of agents.”
Pichach points to corporate and commercial banks as the industry’s clearest examples of frontier firms. In these institutions, a single transaction can cross multiple teams and systems – while their clients experience just one bank.
“Corporate banking work rarely fits inside a single application,” says Pichach. “A banker may need to review meetings, compare documents, analyse account activity, consult policy, coordinate approvals and produce a client-ready deliverable. This is where Microsoft Copilot Cowork delivers real value. The banker delegates a body of work while still guiding the process, validating the sources and approving the outcome. Cowork and specialised agents can help prepare the client context, identify the right experts, surface risks, draft options and monitor next actions. I call it macro-delegation with micro-steering: a human still owns the judgement and relationship, but the bank can respond with more speed, consistency and relevance.”
Below this layer, a comprehensive Microsoft stack handles orchestration.
“Microsoft 365 Copilot is where most banks start, in the daily workflow,” says Baviere. “Copilot Studio and Microsoft Foundry allow them to create low-code or pro-code custom agents for their own processes. Fabric IQ provides a semantic layer over the bank’s data, while Foundry IQ bases responses on cited and authoritative knowledge such as product terms and policies. Work IQ supports execution like creating records, orchestrating steps across systems with an audit trail and keeping humans informed. Meanwhile at the trust boundary, Microsoft Entra and Purview, with a control layer such as Agent 365, provide security, compliance and oversight for the entire fleet of agents. This allows staff to spend less time assembling information and focus on judgement and relationships, where their value to the customer really lies.”
To see this orchestration – and the process reinvention it enables – in action, look at how banks are responding as digital assets reshape the way businesses interact with money, markets and each other.
“Digital assets promise a currency that circulates instantly, ceaselessly and programmatically: stablecoins, tokenised deposits and, in some markets, central bank digital currencies,” says Baviere. “The challenge for corporate and commercial banks is that this happens on top of, not in place of, the existing rails. Treasurers expect a seamless experience as the bank juggles cards, account-to-account, real-time and digital currency, each with different fees, timelines, settlement requirements and compliance obligations.”
AI can become an orchestration layer across settlement rails – one that interprets context, applies policy, monitors obligations, supports reconciliation and helps explain payment choices.
“Microsoft has already explored proofs of concept with ecosystem partners to validate stablecoin-compliant payment flows, while ensuring fiat settlement in line with corporate treasury and regulatory requirements,” adds Baviere. “In the targeted state we describe to customers, every transaction is tokenised and independent of rails, and the digital currency simply becomes an additional option accessible at the orchestration layer.”
Meanwhile, for retail banks, preparing for digital currencies brings an opportunity to further enrich the customer experience.
“Digital currencies will only become mainstream in retail banking if the customer experience becomes simple,” says Pichach. “Consumers want the payment to be safe, fast, fairly priced and easy to understand. The strategic opportunity for retail banks is not simply to add a new form of money. It is to make new forms of money usable in a way that strengthens trust, protects consumers and fits naturally into existing digital banking experiences.”
This is just part of an overarching challenge for the retail banking sector: to meet customer expectations shaped by slick, AI-powered experiences across other industries, from retail to media streaming.
“Retail customers judge their bank by the best digital experience they have, and that bar is set by assistants who take care of things, not menus that list options,” says Baviere. “In this sector, the ongoing shift is from rule-based chatbots to agent assistants who understand intent and context, are aware of profile and consent and can take action: move money, split a purchase into instalments, arrange a payment or resolve a dispute.”
A solid foundation – and the right operating model – is essential to deliver on those expectations while maintaining trust.
“For retail banking, the right model is a permissions ladder, not a binary choice between manual service and full autonomy,” says Pichach. “Personalisation must also remain customer-controlled, so users can understand why an action was recommended, change preferences, and know when an agent is supporting a service experience. Value is best measured through end-to-end outcomes: time to resolve, first-contact completion, quality, control effectiveness, customer effort and employee productivity. Those measures encourage the front office and back office to improve the same journey together.”
“The value of AI in retail banking accumulates when the back office and customer experience are built on the same foundation,” says Baviere. “We advise banks to sequence adoption in stages. First, give every employee a copilot and automate back-office workflows, with agents acting as digital colleagues directed by humans. Next, transform those capabilities into customer-facing assistants that the bank fully controls and integrates into its mobile and online channels. Because the two stages share a layer of governed data and knowledge, an agent can pass a conversation to a contact centre colleague with all the context, so the customer never has to repeat themselves. In the longer term, value shifts to new models including fully agentic operations, agentic commerce, digital currency rails and an AI concierge for every customer relationship.”
Microsoft provides the foundation for banks to follow that path with confidence: Copilot in the workflow, an open agent platform in Foundry and Copilot Studio, governed data, and the security and compliance of Entra, Purview and Defender. But it’s the industry expertise of its vast partner ecosystem that brings the frontier bank vision to life.
“Our partner ecosystem has three layers,” says Baviere. “First, core and payment software providers like Temenos, Finastra and FIS are modernising their platforms on Microsoft Azure and opening them up to agents, so banks can add intelligence without dismantling the core. Second, the networks – including Mastercard, Visa and PayPal – are preparing for the age of agents. For instance, Mastercard’s Agent Pay is collaborating with Microsoft, using Azure OpenAI and Copilot Studio to make agent-initiated payments secure and dependable, with agent tokens as shared credentials. Third, global systems integrators such as Accenture, BCG, KPMG and EY bring the process and regulatory expertise to get these solutions into production safely.”
Mastercard’s Agent Pay is collaborating with Microsoft to make agent-initiated payments secure and dependable
Commerzbank provides a standout example of the intelligent, scalable and replicable banking services that are now possible. It developed Ava, an AI banking assistant, on Azure AI Foundry Agent Service and the Microsoft Agent Framework. Ava already handles more than 30,000 customer conversations per month and resolves about three-quarters of them autonomously. Importantly, the bank sees this as a blueprint for safely adopting AI in a regulated environment.
“The most valuable partner work is the creation of reusable patterns that can move from one bank and one process to many,” says Pichach. “Banks should look for ecosystem programmes with clear ownership, shared controls, measurable business outcomes and a path from pilot to production. That is what turns AI from a promising experiment into an institutional capability.”
Commerzbank’s Ava banking assistant, developed with Microsoft agentic capabilities, provides a blueprint for safely adopting AI in a regulated environment
The genius of the frontier bank lies in integrating AI to orchestrate existing systems and new capabilities, providing real-time intelligence and behaviour that legacy systems could not deliver. By wrapping the core in a layer of intelligent orchestration, banks can transform the way they do things while giving a new lease of life to this trusted platform. “The same governed data and agents can now serve a customer conversation, execute the operational steps behind it and update the case,” says Baviere. “This will blur the traditional notions of front, middle and back office, which define the way work traditionally flowed between systems and teams. It doesn’t make the core system obsolete though – instead, we will see it transform as part of an AI-orchestrated fabric.”
With agentic AI able to carry context, policy and intent across the full journey, Pichach says customer and business outcomes, not system processes, will become the best organising principle for banks.
“The opportunity is to redesign products and operating models for a world where customers, employees and enterprises increasingly use agents themselves,” concludes Pichach. “The core remains essential as the trusted system of record, but intelligence, orchestration and experience move into a more flexible layer around it. That is the future direction: human-led, AI-operated and outcome-oriented.”
Partner perspectives
Microsoft partners Coretek and Moody’s tell Technology Record how they are helping banks harness AI-powered tools to optimise processes, innovate at pace and empower workers to deliver more insightful customer outcomes
Coretek helps banks and financial institutions use Microsoft AI to modernise payments operations, enhance risk visibility and unlock greater value from the vast amounts of data generated across their business. By combining Microsoft Fabric, Azure AI, Microsoft 365 Copilot and intelligent AI agents, we help organisations connect information across core banking, payments, lending and customer service systems to surface insights faster and reduce operational friction. These capabilities enable teams to identify exceptions more quickly, streamline research and analysis, and make more informed decisions while maintaining strong governance and security. The result is a more agile financial institution that can improve operational efficiency, accelerate responsiveness to customer needs, and empower employees to deliver more personalised, data-driven financial experiences.
Brian Barnes, Chief Technology Officer, Coretek
Banks exist to help entrepreneurs and communities access capital so they can fulfil their ambitions, but that only works if bankers can trust the data behind every decision. Through our partnership with Microsoft, we’re embedding our decision-grade intelligence – credit, risk, entity and ownership data – directly into the Microsoft 365 workflow. That means less time reconciling data across systems and more time advising customers. Banking has always been a people business, and our focus is on giving bankers back the time they need to engage with people, not more data and spreadsheets.
Rob Fauber, CEO, Moody’s Corporation
Read all this and more, including news, interviews, case studies and insight from industry experts, in the Autumn 2026 issue of Technology Record. To get content like this on a regular basis, subscribe to the print edition or free digital edition of the quarterly Technology Record magazine.