Practical AI for modernizing commercial lending operations
Setting the pace in a redefined landscape
Artificial intelligence (AI) adoption, as a key enabler for delivering transformation and optimization results for commercial lending operating models, has moved decisively into the mainstream. Banks are embedding generative AI into technology delivery as well as into front, middle, and back-office operations, with the delta between early adopters and the rest of the market continuing to widen. The practical challenge has moved beyond experimenting with AI — to embedding it in ways that generate demonstrable value while remaining defensible under regulatory scrutiny.
The debate over AI adoption for commercial lenders has largely moved past whether AI will shape commercial lending. The more pressing question is how deliberately institutions define their role — rather than allowing market norms, vendor defaults, or fragmented execution to define it for them. How this shift plays out varies significantly by institution, shaped by risk posture, operating maturity, and regulatory context.
Discussions with commercial lending leaders consistently surface two themes. First, the ongoing importance of governance, risk management, and appropriate controls. Second, a shared recognition that AI‑enabled capabilities present material enhancement opportunities that institutions are proactively incorporating into their respective growth and cost reduction plans.
A joint study by The University of Melbourne and KPMG LLP highlights this dynamic: AI usage continues to increase, while confidence — particularly in high-stakes and judgment-heavy decisions — continues to build.¹ In regulated environments, this measured posture reflects healthy attention to accountability, supervisory expectations, and enterprise risk.
At the same time, industry estimates suggest generative AI could unlock $200B–$340B in annual banking value, driven primarily by productivity gains.² For regulated institutions, the opportunity lies in advancing both value creation and trust together — for example, by designing human‑in‑the‑loop quality review processes that enable early productivity gains while preserving oversight, then expanding automation as confidence, accuracy, and control maturity increase. In this model, trust becomes an enabler of scale rather than a constraint on value.
What the future looks like for leading institutions
As AI becomes embedded across commercial lending, leading institutions are converging toward a future state where institutional knowledge persists across the lending lifecycle, execution friction declines, and delivery quality improves earlier rather than later.
In this future state, AI acts as connective tissue, linking decisions, documentation, and rationale across roles and phases, so teams build on prior work instead of repeatedly reconstructing it. The result is faster execution, greater consistency, and outcomes that executives and, where appropriate, regulators can evaluate with confidence.
Why do some commercial lenders see AI value sooner than others?
Early indicators already show divergence. Institutions that embed AI into delivery and operational workflows often experience shorter cycle times, lower rework volumes, and reduced late‑stage remediation, particularly in documentation‑heavy and manual-intensive processes.
Organizations that delay building these capabilities continue to absorb execution drag in the form of elongated timelines, inconsistent outcomes, and rising quality‑control overhead. Over time, this gap can widen, creating meaningful differences in speed, cost efficiency, and delivery reliability.
What delivery shift is required to scale AI effectively?
As AI becomes embedded into commercial lending operations and transformation delivery, institutions are encountering a predictable inflection point: existing delivery and operating models were built for work with stable requirements and predictable outputs.
Most commercial lending programs assume that requirements can be defined early, artifacts remain fixed, and quality is validated at discrete stage gates. AI‑enabled work introduces a different execution pattern. Outputs evolve as context improves, quality depends on how well inputs are grounded and reviewed, and oversight becomes an ongoing activity rather than a one‑time checkpoint.
A key challenge in scaling AI is balancing continuous refinement, traceability, and human‑in‑the‑loop validation with appropriate controls and execution speed.
Addressing this scale challenge requires an execution model designed for AI‑enabled work — one that embeds continuous refinement, traceability, and human‑in‑the‑loop validation directly into delivery and operational workflows. Institutions that adopt this model move beyond isolated use cases and build repeatable, defensible AI capabilities that scale across programs and operations.
The real constraint in commercial lending: the relearning tax
This delivery shift exposes a structural constraint that many commercial lending organizations face today. If delivery teams and operations leaders are asked where commercial lending programs lose momentum and accumulate risk, the answer is consistent: relearning.
Technology transformation initiatives frequently slow because information is repeatedly captured during discovery, reinterpreted during design, redocumented during build, revalidated during testing, and re‑explained during production readiness and post‑implementation stabilization.
.svg){kind=icon}
The relearning tax refers to the repeated loss, reinterpretation, and reconstruction of institutional knowledge across delivery and operational phases.
Each handoff introduces friction and becomes duplication. Documentation loses alignment with original decisions and intent. Outcomes follow a familiar pattern: stretched timelines, late‑stage surprises, increased delivery risk, and fatigue across business, technology, and operations teams.
Operational environments experience similar strain. Exceptions, policy interpretations, credit decision rationale, and booking logic are processed repeatedly across roles and queues, driving inconsistency and avoidable quality‑control burden, particularly during volume spikes or staff capacity pressure.
The relearning tax creates three executive‑level challenges:
- Cost: Capacity is consumed by rework rather than value‑adding change
- Risk: Loss of decision rationale and assumptions drives late‑stage quality issues that are costly to resolve and difficult to defend
- Speed: Operational cycle times remain extended as documents, decisions, and exceptions are repeatedly re‑processed across stages and roles.
AI’s most immediate impact in commercial lending can lie in reducing this relearning tax — by keeping institutional knowledge, decisions, and artifacts connected end‑to‑end. When implemented with appropriate controls, AI can support earlier validation, smoother handoffs, and stronger execution discipline.
AI-enabled transformation delivery: Four execution principles
Executives seeking to scale AI across commercial lending transformations, without increasing risk, can anchor on four execution principles that directly reduce rework, preserve accountability, and compress delivery timelines.
to reduce downstream reinterpretation and documentation drift. High value information should be captured directly from authoritative documents and working sessions, then structured immediately. Earlier structure reduces downstream reinterpretation and drift.
to lower late-stage remediation and approval risk. AI can help surface structured requirements, test scenarios, documentation drafts, and control evidence earlier in the lifecycle, lowering risk by moving scrutiny upstream.
to directly counter relearning and context loss. Teams benefit from being able to search and reuse assumptions, mappings, and rationale across roles and phases. Reuse directly counters relearning.
to enable delivery at scale without diluting ownership or decision accountability. AI accelerates analysis and artifact creation, while accountable leaders and SMEs retain ownership for validation and approvals at defined checkpoints. This model allows programs to scale speed and reuse without eroding accountability, auditability, or risk ownership.
How AI is embedded into commercial lending delivery and operations
This is where the scale challenge is solved in practice. Real progress comes from embedding AI into delivery and operational models, rather than layering it around them. This requires tooling aligned to workflows, auditability, and controls, as well as clear ownership and measurable outcomes.
When AI-enabled capabilities like these are embedded across discovery, delivery, and operations, institutions can scale AI execution with continuity of intent - reducing rework, late-stage risk, and documentation overhead to drive materially stronger execution and fewer surprises.
At KPMG, we deploy AI capabilities designed to support this objective.
KPMG Blaze
Our AI-enabled solution that accelerates technology modernization, including discovery, requirements mapping, artifact generation, testing, and remediation while preserving traceability and structured review points.
KPMG Ignite | Commercial Lending Navigator
Designed to help preserve institutional context across commercial lending workflows by anchoring outputs in a tailored domain corpus, including regulatory guidance, accelerators, and bank‑specific policies where permitted, supporting consistency without flattening expert judgment.
Banking and Capital Markets
KPMG is proud to receive 1st for most authoritative firm in financial services by clients, according to analyst firm, Source
What leaders should do now
AI adoption in commercial banking benefits from deliberate execution. Institutions that scale successfully focus on embedding AI in ways that executives and regulators can evaluate with confidence.
A practical path forward includes:
- Select use cases where relearning and handoffs create clear, quantifiable strain (cycle time, exceptions, rework, QC demand).
- Establish clear definitions of success for traceability, review, and oversight as scope expands.
- Assign accountable ownership and milestones across business, operations, technology, and risk.
- Design adoption directly into workflows through training, checkpoints, and escalation paths.
- Translate measurable outcomes into an executable delivery roadmap, sequencing work against time, quality, cost, and risk to guide scaling decisions and investment.
AI adoption across commercial lending continues to accelerate, and institutions that take a deliberate approach - selecting use cases tied to real operational strain, designing human in the loop agentic workflows, and measuring outcomes in cycle time, quality, cost, and risk - are better positioned for achieving unprecedented levels of efficiency that allow them to scale more profitably.
Footnotes
1 Gillespie N, Lockey S, Ward T, Macdade A and Hassed G, Trust, attitudes and use of artificial intelligence: A global study 2025, The University of Melbourne and KPMG, 2025.
2 McKinsey & Company, Capturing the full value of generative AI in banking, McKinsey & Company, 2023.
Meet the team
