A senior banking executive walked out of a quarterly earnings call recently with a different feeling than usual. Accordiong to Tom Collins, senior partner and national commercial banking lead at West Monroe, shared that the executive told him this was probably the last call where he could say the bank was "still trying to figure it out." The next call would need a real plan, or the market would punish the stock price.
That moment captures where commercial banking sits in 2026. Boards have stopped accepting "we're prioritizing use cases" as an answer. They want proof.
This piece breaks down what Collins told Backbase's Banking Reinvented podcast: the five trends reshaping commercial banking right now, the AI use case most banks are still missing, and the five concrete recommendations he gives banks trying to move from pilot to production without breaking the relationship model the business runs on.
Five trends are converging on commercial banking at once
Speaking on Backbase's Banking Reinvented podcast, Collins laid out what he's seeing across his firm's commercial banking client base. Five forces are hitting at the same time, but not in sequence.
Client onboarding is finally treated as a competitive differentiator instead of a back-office chore, a shift Backbase has covered in depth. Business ownership is transferring generationally, and the new owners expect banking on their own terms. Banks are chasing relationship primacy, hunting for white space to sell beyond the loan. AI investment has accelerated faster in the last two quarters than most banks planned for. Last but not least, small and mid-sized business banking is getting renewed attention as a deposit growth engine.
Throughout Collins' conversation, .he recommends banks actually do, starting with the use case that gets the least attention.
The AI use case hiding in plain sight: continuous relationship intelligence
Collins described a use case his clients are already getting results from. He called it "a forensic sort of portfolio management capability" that identifies emerging risk, fraud, and revenue opportunity from transaction and behavioral patterns, before any of it reaches a human's inbox.
A sudden spike in line utilization. An unusual transaction from a role that never does that kind of activity. A pattern that looks like early-stage attrition. All of it becomes an insight pushed to the RM instead of something buried in a system nobody checks.
Two versions of this use case, and why the distinction matters
Collins is explicit that this internal, forensic version of the use case is deliberately not client-facing. The value goes to the bank's own people first, so they can intervene before a client raises the issue.
He also describes a second, lighter version: in-app nudges that surface directly to the client, like flagging a transaction pattern that suggests a working capital problem with a prompt to learn more. That's a genuinely different, client-facing use case worth its own treatment. This piece focuses on the internal version, since it's the one with the least existing coverage and the clearest near-term ROI.
Why this belongs to Relationship Intelligence, not another point tool
It's tempting to file continuous relationship intelligence under fraud detection or under RM productivity, but it's really neither. This is Relationship Intelligence: continuous, proactive signal detection that puts next-best-action insight directly in front of the banker managing that relationship.
The fraud dimension only becomes customer-facing once a signal turns into an actual case, at which point it hands off to a resolution workflow that contacts the client for proof of payment or funds. The everyday portfolio monitoring stays internal, which is exactly why most banks haven't built for it. It doesn't show up in a client demo.
According to the 2026 AFP Payments Fraud and Control Survey, 76% of organizations experienced attempted or actual payments fraud in 2025, and treasury teams catch most of it before anyone else: 83% of organizations cite their treasury function as the first to discover attempted fraud. That's the same internal-first logic behind continuous relationship intelligence. The earlier a signal reaches the right person inside the bank, the less of it ever reaches the client.
Diagnosis isn't the hard part. Here's what banks are recommended to do
During the conversation with host Tim Rutten, Collins tried to answer the harder question: how banks execute continuous relationship intelligence without breaking the RM model that still drives the business.
1. Data governance is table stakes, not a nice-to-have
Collins was blunt about credit operations specifically. Guarantor financials and supporting documents mostly still arrive as unstructured PDFs that never get digested properly. His warning: "clients expect us to know this information." When a bank asks a client something the client thinks it should already know, it damages credibility instead of building it.
2. Choreograph how digital and physical interactions coexist
This is a distinct point Collins made, separate from RM training. As automation takes over transaction coordination, banks need to deliberately design how digital self-service and human RM conversations hand off to each other. Get this wrong, and clients experience a disjointed relationship instead of a coherent one, even if each channel works fine on its own.
3. You cannot flip a switch and call an RM a trusted advisor
Automating the tactical side of the RM role does not automatically turn that person into a strategic advisor overnight. "It's going to require some human capital management," Collins said, and it takes real time.
Practically, that means training RMs to hold a different kind of conversation: asking about new product lines and working capital needs 12 months out, instead of chasing last quarter's financials for a renewal. Not every RM in the role today will make that transition successfully, Collins added, which is why banks need a deliberate plan for who moves into the advisor seat.
4. Fewer, more skilled RMs beat more, less-equipped ones
Collins expects the RM model to shift, not shrink for its own sake. As mundane tactical work gets automated, RMs can responsibly cover a broader portfolio, which means fewer RMs overall but each one carrying more relationships at higher quality.
This tracks with the wider data on where RM time actually goes. BCG research shows RMs in US commercial banks can spend up to 60% of their time on internally-facing administrative tasks with little effect on client satisfaction or revenue.Free up that time responsibly, and coverage capacity rises without adding headcount. It's the same math behind how commercial banks fix RM productivity without adding headcount.
5. Invest now, or become someone else's acquisition target
Collins closed with his starkest recommendation. Banks that keep investing in AI, platforms, and data programs will have the efficiency to stay independent and acquisitive. Banks that don't, in his words, "are probably setting themselves up to be acquired." He named Backbase directly as one of the investments he sees separating what he called the industry's coming haves and have-nots.
This isn't hypothetical urgency. Cost-to-income ratios in corporate and investment banking, the category that includes large commercial banking arms, averaged 54% in 2024, according to McKinsey's 2026 state of the industry report. Efficiency is the clearest lever left for banks trying to close the gap to top performers.
Why this breaks without a unified foundation underneath
None of the five recommendations above work if the signals feeding them come from a disconnected system. Continuous relationship monitoring needs a real-time, shared view of the client across lending, treasury, payments, and credit, not a new dashboard bolted onto the fifth system an RM already has open.
This is the case for connecting Relationship Intelligence and Customer Operations into one shared context. Running them as separate, disconnected tools defeats the point. Fraud, risk, and attrition signals only compound in value when they draw from that same context. It needs to reach the RM's next-best-action, the credit team's exception queue, and the compliance team's audit trail at once. The Banking OS is what makes that possible, the shared foundation powering both solutions instead of splitting them apart. Fragmented architecture is what keeps cost-to-income ratios stuck above 60% industry-wide, even after years of front-end investment.
Governance matters as much as the data model underneath. Gartner predicts over 40% of agentic AI projects will be canceled by 2027. The drivers are escalating costs, unclear value, and poor risk controls, not model quality. A monitoring system that flags fraud or credit risk needs a clear, auditable authority trail before it acts.
What this means for your commercial banking operating model
- Build continuous, internally-facing relationship monitoring before another client-facing feature.
- Fix data governance first. Bad data undermines the trusted advisor pitch faster than no AI at all.
- Design the digital-to-human handoff deliberately, don't let it happen by accident.
- Budget time and training for RMs to become advisors. The tech transition and the human transition are separate projects.
- Plan for fewer, higher-capacity RM roles, not a flat headcount cut.
- Treat platform investment as a defense against consolidation, not just an efficiency play.
Frequently asked questions
What does a modern commercial banking operating model need from AI?
It needs a unified view of the client across lending, treasury, and payments so AI-driven signals, whether fraud, credit risk, or attrition, reach the right person with full context, governed by a clear authority trail rather than sitting in a disconnected point tool.
What is continuous relationship intelligence in commercial banking?
It's ongoing, AI-driven monitoring of a commercial client's transaction and behavioral patterns to surface emerging credit risk, fraud, and attrition signals directly to the relationship manager, before the client raises an issue.
Is this the same as AI-powered onboarding or RM productivity tools?
No. Onboarding automation speeds up account opening. RM productivity tools remove administrative work. Continuous relationship intelligence is a separate, ongoing signal-detection layer running on the existing portfolio.
Why do banks need data governance before deploying this kind of AI?
Signals from incomplete or unstructured data create false positives and erode client trust. Tom Collins points out that clients expect their bank to already know basic information about the relationship.
How does this connect to Backbase's platform?
Relationship Intelligence runs on the Banking OS, giving RMs proactive, next-best-action insight from a unified view of the client, with Customer Operations resolving any signal that escalates into an actual case.
