TD Bank U.S. Launches AI-Powered 'Credit Assist' to Slash Commercial Underwriting Time

TD Bank U.S. has launched TD Credit Assist, an AI-driven tool designed to compress the commercial credit analysis process from three days to under 30 minutes. For fintech leaders, this represents a significant shift in middle-office automation, moving generative AI from experimental chatbots into high-stakes, document-heavy underwriting workflows where speed and accuracy are critical.

What was announced

TD Credit Assist is a practical application of AI designed to support the bank’s 300-plus credit professionals, who collectively handle more than 10,000 credit memos every year. The system addresses the complexity of commercial lending, where a single review often requires the synthesis of 20 to 30 distinct deal documents across 50 different document types. Instead of beginning with a blank page, credit professionals receive a cited draft of six key credit memo sections that they can then question, verify, and refine. This allows them to focus more on understanding client businesses and assessing risk rather than manual data entry.

The technology uses an orchestration layer to coordinate specialized AI agents that process, extract, and analyze information in parallel. Each section of the memo undergoes its own process of analysis and refinement before being presented to the human reviewer as a cited narrative. The development was led by TD’s Analytics, Intelligence & AI team in collaboration with Layer 6, the bank’s dedicated AI research and development center, and involved experts across technology, risk, architecture, governance, and controls. In testing, the tool was evaluated against 220 curated deals with input from 70 domain experts, resulting in 95% of the AI-generated outputs passing subject-matter expert reviews. The tool was unveiled at the 2026 Evident AI Symposium in New York, following TD’s rise to the tenth spot in the 2026 Evident AI Index, which evaluates the AI maturity of 50 of the world’s largest banks.

"Credit Assist is a great example of how we're simplifying and optimizing the way we work,"

Jill Gateman, Head of Commercial Banking, TD Bank U.S.

The companies involved

TD Bank U.S. operates as a major commercial banking entity, supported by internal specialized units for technological advancement. A key driver of its technical strategy is Layer 6, TD’s AI research and development center, which focuses on translating advanced machine learning research into practical banking applications. The bank’s AI initiatives are overseen by the Analytics, Intelligence & AI team, which includes leadership such as Ted Paris, the Head of Analytics, Intelligence & AI at TD Bank U.S., and Sumee Seetharaman, who serves as the Head of Artificial Intelligence and Machine Learning Practice & Center of Excellence at Layer 6. Jill Gateman serves as the Head of Commercial Banking at TD Bank U.S., overseeing the division where Credit Assist is being deployed.

This internal structure allows the bank to develop proprietary orchestration layers and AI agent frameworks rather than relying solely on off-the-shelf third-party solutions. By integrating these teams with commercial underwriting experts, the bank maintains a focus on "human-led" AI, where the technology handles the administrative burden of data extraction while credit professionals retain final accountability for lending decisions and risk assessment. The collaboration between the technology, risk, architecture, and governance teams ensures that required controls remain in place throughout the AI-enhanced underwriting process. This reflects the bank's broader strategy of applying AI to specific business problems while maintaining transparency and human accountability for all final lending recommendations.

What this means

This launch marks a transition in the banking sector from general AI experimentation to specialized, agentic workflows. By automating the most labor-intensive part of the credit cycle—the initial data synthesis—TD is putting pressure on competitors who still rely on manual document review and traditional underwriting templates. The 95% expert pass rate suggests that AI is becoming reliable enough for regulated financial environments, provided there is a "human-in-the-loop" to verify citations. The broader industry question is no longer whether AI can write a memo, but how quickly banks can integrate these tools into their core risk frameworks without compromising governance or increasing model risk. As more institutions move toward "AI-enhanced" models, the competitive advantage will likely shift to those who can prove the accuracy of their automated citations to regulators and maintain the rigor of their lending decisions.