How AI Is Rearchitecting Lending

Lending Is Reaching An Inflection Point

More than 80% of financial services (FS) AI decision‑makers plan to increase investments in both predictive AI and generative AI (genAI), with most firms expecting double‑digit growth. The immediate focus for the majority of FS leaders remains pragmatic: scaling origination, reducing friction caused by handoffs, and improving risk control. But the leaders aren’t the only ones who optimize for middle- to back-office workflows; lenders who embed intelligence from the very first moment a customer expresses interest and intent, all the way through to helping them achieve their goals, will outpace those who don’t.

Efficiency And Risk Mitigation Remain The Primary Drivers

Despite growing interest in customer‑facing use cases, most lenders continue to focus on efficiency gains and risk mitigation, such as:

• Identity verification and fraud prevention. They achieve this by combining ML, graph analytics, behavioral biometrics, and liveness detection to counter deepfakes and synthetic identities. HSBC used graph ML to map the “network” of an individual or entity, which uncovers insights or relationships to other individuals, events, or entities that might indicate fraudulent behaviors.
• Fraud and anti-money laundering monitoring. This is achieved by using multimodal analysis of documents, transactions, and unstructured data to detect evolving threats and reduce false positives. BCU Bank leveraged network, forensic, semantic, and preceptor detectors — as well as relational links — to uncover fraudulent documents (e.g., a repeated balance on a bank statement), preventing $5.6 million in losses in the first nine months of 2025.
• Credit memo generation. They achieve this by leveraging retrieval-augmented generation to extract key insights from submitted documents, financial statements, and internal taxonomies. As well as third-party news, market data, and financial filings, lenders can use genAI to synthesize this data into a concise credit memo that provides a borrower overview, financial and risk analysis, covenants, and a recommendation.

Customer‑Facing GenAI Adoption Is Advancing, But Selectively

While interest in customer-facing and domain-specific applications is growing, most financial institutions continue to proceed cautiously due to regulatory scrutiny, integration complexity, and the need for high levels of accuracy and explainability. Forward-looking lenders are experimenting with and scaling AI across:

• Personalized marketing campaigns. Using ML, deep learning, predictive AI, natural language processing, and genAI, lenders can map and generate marketing content with language that aligns with specific emotional triggers for an individual.
• Customer help and support. Rocket Mortgage has built a customer-facing genAI assistant, called Rocket AI Agent, with eight domain-specific agents invoked via a centralized orchestration layer to help with a number of tasks, such as providing answers to product questions like rates, options, and processes; guiding borrowers through preapproval forms; and helping borrowers schedule payments.
• Customer engagement. Built on an agentic AI architecture, Lendi Group launched Guardian: a mobile-first, AI-powered companion that serves as the conversational front door to the group’s agentic platform. Under the hood, Guardian is a multiagent system powered by more than a dozen specialized agents, which continuously support customers at key moments by surfacing insights that matter, such as interest rate saving alerts, equity changes, and suburb price growth. 

Use AI To Unlock The Next Phase Of Value Creation

The reality is most lenders will struggle to unlock new growth vectors. Fixated on using AI as a tool for business-as-usual efficiency, not as a strategic lever, many lenders will face diminishing marginal gains. To unlock the next phase of value creation, lenders need to leverage AI as a driver of growth, differentiated experiences, and competitive advantage. Over the next 12 months, we expect:

• Conversational banking to become the cornerstone of customer engagement. The next iteration will replace menus with a single conversational bar, where borrowers are guided from initial intent through to loan management — with the overall goal of debt reduction and wealth creation. This is a paradigm shift, not an incremental change or a UX upgrade.
• Agentic AI will act on behalf of people and software, but it won’t be 100% autonomous. Early agentic AI lending implementations focus on middle- and back-office operations; however, truly transformative outcomes only emerge when agentic AI is embedded in the experience layer: for example, an adaptive system that guides borrowers through their end-to-end journey by integrating data from two-way conversations (between borrowers, AI agents, and brokers) with the data fabric. These combined insights build borrower preferences and context knowledge, helping agents plan, select the right tools, self reflect, and improve their decision-making.

Read my latest report, The State Of AI In Lending, 2026, for a deep dive into these insights. Got a question? Book in a guidance session with me.

Categories

Blog

Salesforce’s New AgentExchange Store: Easier AI Helpers But Still Inside Their Walled Garden

Salesforce’s reimagined marketplace makes buying AI and applications inside Salesforce dramatically easier and makes leaving Salesforce later even harder. Salesforce has merged three separate “stores” for apps and AI tools into a single marketplace called AgentExchange. It is live now with roughly 13,000-plus ready-made apps, AI agents, and automation components. The stated goal is simple: […]

Blog

Tech Leaders, Brace Yourselves: AI Costs Will Only Go Up

Anthropic gave leaders a blunt signal about planning for LLM charges when it introduced blended pricing for enterprise users of Claude recently. Enterprises with over 150 seatholders will now pay a blanket per-seat charge ($20 per seat at the moment), plus a variable charge based on API usage and which Claude model they’re using. When […]