Blog
What Happens to Junior Lenders When AI Does Their Training Work
Everyone wants to talk about whether AI should be making lending decisions. I understand why. It is the headline question, the one that gets people nervous in a boardroom and gets clicks in a trade publication. But after spending time with underwriting teams, credit committees, and heads of lending over the past year, I have come to believe it is not the most important question in front of us. The more consequential question is quieter, and it is this: what happens to how lenders learn their craft once AI starts doing the foundational work that used to teach them how to think?
The capacity AI creates is real
In consumer lending and small business lending, automation is already handling meaningful portions of the underwriting process. Document collection is faster. Income verification is largely automated. Initial credit screening is heading the same direction quickly. For experienced lending teams, this is unambiguously good news. It removes hours of repetitive prep work and frees up time for exceptions, complex credits, and the judgment calls that actually require a trained eye. Any operator who has watched a credit team drown in file assembly during a volume spike knows exactly how valuable that capacity is.
I am not writing this to argue against that shift. Automation belongs in document prep, data reconciliation, and initial file assembly. That is precisely where it should be doing the heavy lifting. The problem is not the automation itself. The problem is what we are not paying attention to while it happens.
Judgment was never taught in a classroom
Almost every experienced credit officer I know did not learn judgment from a training manual or a certification course. They learned it by doing the early analytical work themselves, badly at first, and getting corrected. Spreading financials by hand. Reading an application closely enough to notice something inconsistent. Making a recommendation and then sitting across from someone more senior who challenged every assumption in it. That repetition, uncomfortable as it was, is how an analyst develops the instinct for when something does not look right even though the numbers technically check out.
This is a distinction that matters enormously in lending. There is a difference between an analyst who can confirm that a debt service coverage ratio meets policy and an analyst who understands why that ratio might be misleading for a particular borrower in a particular industry at a particular point in a cycle. The first is a calculation. The second is judgment. And judgment is built through volume of exposure, not through instruction.
If AI takes over the early analytical work, it removes exactly the repetitions that built that judgment. It creates capacity at the senior level while quietly hollowing out the training pipeline underneath it. Nobody designed it this way on purpose. It is simply what happens when you automate the entry-level tasks without redesigning what replaces them as a learning mechanism.
The gap will not show up right away
This is the part that makes the problem easy to ignore. The consequences of a hollowed-out training pipeline do not show up in this year’s numbers. They show up three to five years from now, when the analysts who came up during the automation transition are expected to run credit committees, manage portfolios independently, or train the next cohort behind them. At that point, an organization discovers that it has a team of people who are excellent at operating tools and interpreting outputs, but who struggle to explain why a credit decision was actually right, or to catch the kind of anomaly that does not trip any automated flag.
I have sat in enough credit meetings to know what that gap looks like in practice. It is the analyst who can walk you through every field in the system but cannot answer a follow-up question about why the borrower’s receivables aging matters more this quarter than last. It is not a failure of intelligence or effort. It is a failure of exposure. They were never asked to sit with a messy file and figure out what was wrong with it, because the system handed them a clean one.
Redesigning training on purpose, not by accident
The lending organizations that navigate this well are not the ones avoiding automation. They are the ones being deliberate about what automation replaces and what it does not. If AI is going to remove the administrative burden around a lending decision, that capacity has to be redirected somewhere intentional. It should not simply become more volume processed by the same number of people with no change in how those people are developed.
The organizations getting this right are using the time AI frees up to do more, not less, direct coaching. More credit discussions where a senior lender walks a junior analyst through the reasoning behind a decision instead of just approving it. More structured exposure to marginal files, the ones that do not resolve cleanly, because those are exactly the files that build judgment fastest. In other words, they are treating the capacity AI creates as a training budget, not just an efficiency gain.
This requires a shift in how leadership thinks about the automated work itself. When a file lands in front of a credit officer clean, complete, and ready for review, that is not the end of the value automation can create. It is the beginning of an opportunity. That clean file is a chance to spend time on the discussion around it rather than the assembly of it. Organizations that treat automation purely as a headcount or throughput play are leaving that opportunity on the table.
What this looks like operationally
In practice, this means being explicit about which tasks get automated and which get preserved specifically because they carry training value, even if they are no longer strictly necessary for the decision itself. Some lenders I have talked to are experimenting with keeping a rotation of files where junior analysts still do the initial spread and recommendation manually, even though the system could do it faster, specifically because that exercise is where the learning happens. The automated version runs in parallel as a check, not a replacement.
Others are restructuring credit committee meetings so that junior staff are required to present and defend recommendations on a subset of files, rather than simply routing everything through senior underwriters for speed. It takes longer. It is less efficient in the short term. But it is a direct investment in the pipeline of people who will eventually be running those committees.
None of this is about resisting technology. It is about recognizing that operational efficiency and talent development are two different goals that can pull against each other if you are not paying attention. A lending organization can become dramatically more efficient in the next two years and dramatically weaker in institutional judgment five years out, and those two outcomes can happen at the same time without anyone noticing until it is expensive to fix.
Why this matters more for complex lending operations
This issue is more acute for lenders with operational complexity, specialty finance companies, commercial real estate lenders, CDFIs, and others managing diverse portfolios where judgment cannot be fully standardized. In simple, high-volume consumer lending, a well-tuned model can carry more of the decision weight because the variance between files is narrower. But in commercial and specialty lending, every borrower brings a different capital structure, a different industry exposure, a different set of risks that do not map cleanly onto a template. Those are exactly the environments where human judgment matters most, and exactly the environments where the training pipeline needs the most protection.
For these organizations, the platforms and workflows they build around AI matter as much as the AI itself. A system that automates document collection and reconciliation but still routes the resulting analysis through a structured review process, one that surfaces the reasoning behind a recommendation rather than just the output, preserves the visibility that senior lenders need to coach effectively. A system that simply produces a decision with no visible reasoning path does the opposite. It optimizes for speed at the expense of transparency, and transparency is what makes a file a teaching tool instead of just a transaction.
The real choice in front of lending leaders
I do not think the choice lenders face is whether to adopt AI in underwriting. That decision is largely already made, and the operational benefits are too significant to ignore. The real choice is what to do with the capacity it creates. Used well, that capacity funds more mentorship, more structured judgment-building, and a stronger bench of future credit officers. Used carelessly, it simply becomes more volume pushed through the same team, with nobody accountable for whether the next generation of analysts can actually think through a complex credit or just operate the tools that used to do the thinking for them.
The organizations that get this right will not necessarily look different in their technology stack from the ones that get it wrong. The difference will show up later, in the quality of the people making decisions when the automated systems hit an edge case they were never trained to handle. That is the moment judgment matters most, and it is exactly the moment an organization either has it or does not, depending on choices made years earlier about how junior staff were allowed to learn.
Blog
Why Connected AI Agents Beat Isolated AI Tools in Lending
I have spent a lot of time lately in conversations with COOs and Heads of Lending about AI, and I keep noticing the same pattern. Almost every lender I talk to is thinking about AI the same way, and I think that framing is actually getting in the way of the real opportunity in front of them.
Here is what I mean. When a lending organization adopts AI today, it typically adopts it in one place. An AI tool for spreading financial documents. An AI tool for drafting borrower communications. A chatbot for answering customer questions. Each of these tools does its job reasonably well in isolation. Vendors demo them well, teams pilot them, and for a narrow task they often deliver. Then the tool hits a wall, and that wall is almost always the same one: the departmental boundary.
No Problem in Lending Is Actually Isolated
Lending does not happen in isolated steps. It happens as a chain of dependent events. A borrower submits an application. That single moment should trigger a dozen things at once: credit verification, income confirmation, document collection, underwriting rule evaluation, compliance screening, CRM updates, and borrower communication. All of that should move together, in parallel, coordinated.
In most lending organizations I visit, that sequence is still largely stitched together by people. Someone downloads a verification report and uploads it into a different system. Someone notices a condition was cleared and sends an email to an underwriter. Someone updates a record in the loan file, or forgets to, and three days later an auditor or a borrower finds the gap. Every one of those handoffs is a seam in the process, and in that seam lives delay, error, and cost.
This is the part that gets missed when a lender buys a point solution for AI. The tool might genuinely be good at its narrow task. But if it cannot see what happened before it in the process, and it cannot tell anything downstream what just happened, it is an island. It speeds up one step and does nothing for the twelve handoffs surrounding it. The borrower still waits. The underwriter still has to notice. The operations team still spends their day reconciling what happened across systems that do not talk to each other.
The Real Opportunity Is Agents That Talk to Each Other
The lenders who are going to pull ahead with AI over the next several years are not the ones who buy the most AI point solutions. They are the ones who architect their systems so that AI agents can coordinate with each other across the full lending lifecycle.
Think about what that actually looks like in practice, because it is a meaningfully different operating model than what most lenders run today. A loan comes in. An origination agent processes the application and identifies that the borrower’s income documentation is incomplete. In the old model, that gap sits until a human notices it, usually during a manual file review days later. In a connected model, the origination agent fires a signal the moment it identifies the gap. A document collection agent picks up that signal immediately and sends the borrower a targeted, specific request for exactly the missing document, not a generic reminder to upload paperwork. The moment the borrower responds and the document arrives, another agent routes it directly to underwriting. The underwriting agent evaluates it against the credit policy that is already configured in the system and fires its own signal back: condition cleared. The borrower receives a status update in seconds, not the next time someone happens to check a queue.
Same loan. Same staff. Same policies. The difference is that coordination no longer lives in people’s inboxes and mental checklists. It lives in the architecture of the platform itself. That is the actual unlock, and it is very different from simply adding a chatbot on top of an existing broken workflow.
You Cannot Automate Chaos
Here is the caveat that I think matters more than anything else in this conversation, and it is the part vendors tend to skip because it is not exciting to talk about. What I keep observing across lending organizations, whether they are CDFIs, private lenders, or commercial real estate shops, is that the teams doing this well are almost never the ones with the biggest AI budgets. They are the ones who started with a clean system of record and well-defined, documented workflows.
AI agents can only coordinate across a process that already exists in a structured, observable form. If your income verification step lives in one vendor portal, your document collection lives in a shared drive, your underwriting notes live in someone’s personal spreadsheet, and your CRM is only updated when someone remembers to do it, there is no coherent process for an agent to plug into. You cannot automate chaos. You can only make chaos move faster, which usually just means you find your errors sooner and your compliance exposure grows just as quickly as your throughput.
This is why I tell lenders that the AI conversation is really a data architecture and workflow conversation wearing a different hat. Before you can meaningfully connect agents across origination, underwriting, servicing, and borrower communication, you need those functions operating on a common system of record with workflows that are actually defined, not just understood tribally by your most senior underwriter. That is unglamorous work. It is also the entire precondition for everything that comes after it.
What Changes When the Foundation Is Right
Once that foundation exists, connected agents change the math in ways that are hard to replicate through headcount or point solutions alone. Closing times collapse because the handoffs that used to take days now take minutes. Error rates drop because information is not being manually re-keyed between systems, and re-keying is where most operational errors are born in the first place. Your operations team stops spending its day chasing information across systems, tracking down a missing verification, confirming a document was received, updating a record someone forgot to touch, and starts spending its day on the judgment-intensive work that actually requires a human: evaluating a marginal credit decision, structuring a complex facility, having the borrower conversation that needs empathy and negotiation rather than a status update.
That shift in where your team spends its time is, in my view, the actual return on an AI investment. Not that a machine replaced a person, but that the coordination overhead that used to consume a huge share of your operational capacity simply disappears, and the capacity gets reinvested into the parts of lending that are genuinely difficult to automate.
Why Departmental AI Purchases Keep Underdelivering
I want to be specific about why the departmental approach to AI keeps underdelivering, because I do not think it is a failure of the individual tools. It is a structural issue. When origination buys a document-spreading tool, underwriting buys a risk-scoring tool, and customer service buys a chatbot, each department has optimized for its own slice of the process. Nobody owns the seams between departments, because no single AI tool was ever designed to own a seam. It was designed to own a task.
The seams are exactly where the operational pain lives. The seam between origination and underwriting is where conditions get missed. The seam between underwriting and servicing is where onboarding errors happen. The seam between servicing and reporting is where numbers stop reconciling and someone spends a week at month end trying to figure out why. Point solutions, no matter how sophisticated individually, do not resolve seams. Only a connected architecture does, because a connected architecture is designed around the process as a whole rather than around a department’s task list.
The Question Every Lending Executive Should Be Asking
So the practical question I would put in front of any lending executive right now is not, what AI tool should we buy next. It is a more fundamental question: does our current platform architecture support agents that can actually talk to each other across the lending lifecycle, or are we buying another island?
If the honest answer is that your origination system, your underwriting process, your servicing platform, and your CRM are separate systems held together by exports, uploads, and someone’s memory of what needs to happen next, then the AI tool you are evaluating, however impressive its demo, is going to hit the same wall your current process hits. It will be faster at one task and just as blind to everything around it. Isolated bots are not the destination for lending operations. They are just a faster way to arrive at the same walls you already have.
The lenders who get this right are approaching it as an operating model decision first and a technology decision second. They are consolidating around a system of record, documenting their workflows honestly, including the messy exceptions, and only then layering in agents that can see across the process and coordinate with each other. That is a harder path than buying a point solution off a vendor’s feature list. It is also the only path that actually closes the gaps where delay, error, and cost have been living in your operation the entire time.
This is the shift I think every lending organization needs to be having internally right now, well before the next AI tool demo shows up in their inbox. The winners in this next phase will not be the lenders with the most AI subscriptions. They will be the ones whose systems were built to let intelligence move freely across the entire loan lifecycle, instead of being trapped inside a single department’s workflow.
Blog
Why Clean Data Alone Can’t Protect Lenders From Risk
I was on a podcast recently talking about technology in specialty lending, and the host said something that stuck with me. We were talking about data quality, which is the topic everyone in this industry wants to talk about, and he made a distinction I had not heard articulated quite that way before. He separated data quality from controls. Most of the conversation in lending technology circles treats those as the same problem. They are not. And the difference between them is where a lot of lenders are quietly exposed.
The Industry Is Obsessed With Dirty Data
If you have spent any time around loan operations, servicing, or a digital transformation initiative at a lender, you have heard the phrase dirty data more times than you can count. Duplicate borrower records. Inconsistent field formats. Stale collateral values. Loan officers entering the same information three different ways in three different systems. It is a real problem, and I am not going to tell you otherwise. Bad data creates bad reporting, bad reporting creates bad decisions, and bad decisions compound over a loan book that might carry hundreds of millions of dollars in exposure.
So lenders spend money and time on data cleanup projects. They bring in consultants to standardize fields. They migrate to new systems with cleaner schemas. They build dashboards that promise a single source of truth. All of that is worthwhile work. But here is the uncomfortable part. None of it prevents the kind of failure that actually causes the most damage at most lending organizations. Clean data describes what is true. It does not govern what people are allowed to do with that information, or what happens before an action becomes irreversible.
Controls Are a Different Layer Entirely
Controls are the guardrails that sit on top of data and on top of process. They are the rules that determine who can approve a loan above a certain size, who can release funds, who can see a borrower’s full financial file, and what has to happen before a wire goes out the door. Controls do not care whether the underlying data is perfectly clean. They care about sequence, authority, and verification. A lender can have immaculate data and still have someone wire the wrong amount, approve a loan that should have gone to a second reviewer, or grant system access to someone who should not have it. I have seen this happen at organizations that had genuinely good data hygiene. The data was not the problem. The absence of a checkpoint was the problem.
This distinction matters more as lending organizations grow. A five-person shop can rely on informal controls, because everyone knows everyone and someone will probably catch a mistake before it becomes expensive. That informal safety net disappears the moment a lender scales past a certain size, adds new loan products, brings on new staff, or starts operating across multiple offices or business lines. At that point, if controls are not built into the workflow itself, they exist only in policy documents that nobody consistently follows under deadline pressure. And deadline pressure is the normal operating condition in loan operations, not the exception.
A Story That Has Stayed With Me
I will tell you why this topic is personal for me. Early in my career, I was processing a wire transfer and fat fingered an extra zero into the amount field. The transfer went to a bait and tackle shop in Montana, right before their busiest season of the year. We never got that money back. There was no system in place that flagged the transaction, no secondary approval that would have caught an amount that was ten times larger than it should have been, nothing that stood between a single keystroke and a permanent loss. The data in our system was accurate right up until the moment I typed the wrong number into it. Clean data did nothing to stop that error, because the error was not a data quality problem. It was a missing control.
That experience is a big part of why I ended up building lending technology instead of just using it. Every time I talk to a COO or a head of lending operations about their systems, I am listening for whether they have thought about controls as a distinct discipline from data management. Most have not, not because they are careless, but because the industry conversation has trained everyone to think about the problem the wrong way. Vendors sell data quality tools. Consultants sell data governance frameworks. Almost nobody is selling operational control architecture, because it is harder to demo and harder to put in a slide deck. But it is the thing that actually prevents the losses that show up in board meetings.
Where Lenders Are Most Exposed
In my conversations with operations leaders across specialty finance, CDFIs, commercial lenders, and private credit funds, the exposure tends to cluster in a few predictable places. The first is funding and disbursement. Anywhere money physically leaves the organization is the highest stakes moment in the entire loan lifecycle, and it is often the least protected step in the process, because by the time a loan reaches funding, everyone assumes the hard work of underwriting and approval is already done. That assumption is exactly what makes funding dangerous. A control failure at underwriting usually gets caught somewhere downstream. A control failure at disbursement is frequently final.
The second area of exposure is access and permissions. As lending organizations grow, they add staff, add departments, and add third-party partners who need some level of system access. Permissions get granted in the moment to solve an immediate problem, and they rarely get revisited. I have sat with operations teams who could not tell me, without pulling a report and manually reviewing it, who currently had authority to approve a loan above a certain threshold. That is not a data quality gap. That is a control gap, and it is one that creates real regulatory and financial risk, particularly for lenders operating in regulated categories like government-backed lending or CDFI programs where audit exposure is constant.
The third area is the handoff between systems. Most lenders I talk to are not running one unified platform. They are running a loan origination system, a separate servicing system, a document repository, and a handful of spreadsheets that fill the gaps nobody built for. Every handoff between those systems is a point where a control can silently disappear. The origination system might require a second approval for large loans. The servicing system, being a completely separate piece of software from a different vendor with different logic, might not enforce that same rule once the loan moves into servicing. Nobody designed that gap on purpose. It emerged because the systems were never built to share a control framework in the first place.
Why This Gets Worse as Lenders Scale
The lenders who get hurt worst by this are usually not the smallest ones. Small lenders are informally protected by low volume and close oversight. The lenders most exposed are the ones in the middle of a growth curve, adding loan products, adding staff, adding geography, and doing it on infrastructure that was designed for a smaller, simpler operation. I have watched organizations double their loan volume in eighteen months while their operational control structure stayed exactly the same as it was when they were half the size. Everyone is moving fast, everyone is focused on growth, and controls quietly become the thing nobody has time to revisit until something goes wrong.
This is also where technology decisions matter more than people give them credit for. A platform that treats controls as configurable workflow logic, built into the approval chain, the funding process, and the permissioning structure, gives an operations leader the ability to enforce consistency without relying on individual discipline. A platform that treats controls as an afterthought, or leaves them to policy documents and manual review, puts the entire organization at the mercy of every individual employee getting every step right, every single time, under pressure, forever. That is not a sustainable control strategy. It is a hope strategy.
What Operational Leaders Should Actually Be Asking
If you are a COO or head of lending evaluating your own exposure, the useful question is not whether your data is clean. Ask instead whether your systems physically prevent someone from taking an action they should not be authorized to take, or whether your systems simply record that the action happened after the fact. Ask whether a large disbursement requires a second set of eyes by design, or whether it requires a second set of eyes only because someone remembers to ask for one. Ask whether permissions are reviewed on a schedule, or whether they accumulate silently until an audit forces a cleanup.
None of this is about finding a villain or blaming your team. Almost every control gap I have seen was built by well-intentioned people solving an immediate problem without thinking about the long-term structure they were creating. The fix is not a lecture about diligence. It is building an operational structure where the right thing happens automatically, because the workflow requires it, not because someone remembered to check.
Clean data is worth pursuing. I am not arguing against it. But if a lender is only investing in data quality and treating that as their risk management strategy, they are protecting themselves against the wrong failure mode. The wire transfer that never comes back, the loan approved by someone who should not have had that authority, the access granted and never revoked, none of those are data problems. They are control problems. And control problems do not show up in a data quality report. They show up in a loss statement, or worse, in a regulatory exam. The lenders who understand that distinction early are the ones who scale without a costly lesson attached to it.
Blog
Why Loan Servicing Is Now a Strategic Asset, Not Overhead
I have spent a lot of time over the past year sitting across the table from COOs and Heads of Lending at specialty finance companies, CDFIs, and commercial lenders of every size. The conversations vary, but there is a pattern I keep running into, and it is worth naming because I do not think enough people in this industry are saying it out loud. For as long as most of these executives can remember, loan servicing has been treated as the back office. The unglamorous half of the business. Originations gets the budget, the new hires, the technology investment, and the attention of the board. Servicing gets whatever is left over after those priorities are funded. It is managed like a cost center, not built like a capability.
What has changed is that a subset of lenders have quietly rejected that framing, and it is starting to show up in their results. They stopped treating servicing as an administrative obligation and started treating it as a strategic asset. That is not a branding exercise. It is a genuine shift in how they allocate investment, how they measure success, and how they think about the relationship between servicing quality and the rest of the business. I want to walk through why that shift matters, what under-investment actually costs, and how operations leaders can reframe the conversation internally.
What under-invested servicing actually looks like
It is worth being specific here because the phrase under-invested servicing can sound abstract until you see it in a day-to-day operation. Payment processing is manual or semi-manual, with someone reconciling ACH files or checks against a loan management system that was never designed to talk to the bank feed directly. Borrower communications are inconsistent, driven by whichever servicing rep happens to pick up the account that week rather than a standardized cadence. Reporting requires someone, often a fairly senior analyst, to pull data out of three or four different systems and reconcile it by hand before it can go in front of leadership or a regulator. Delinquencies surface late because nobody is actually monitoring portfolio behavior in real time. The first sign of trouble is often a missed payment notice, not a pattern that would have been visible weeks earlier if someone had been watching the right signals.
None of this is because the operations team is bad at their job. It is because the infrastructure underneath them was never built to do this well. It was built to be adequate. And adequate infrastructure produces a very specific kind of cost that does not show up cleanly on an income statement. It shows up as staff hours spent on reconciliation instead of borrower relationships. It shows up as error rates that create rework and, occasionally, real financial exposure. It shows up as borrower frustration that never gets logged anywhere but quietly erodes the relationship. And it shows up as missed early warning signals on loan performance, which is arguably the most expensive cost of all because it is the one that turns into actual credit losses.
I want to be careful here not to overstate this as purely a technology problem. It is an operating model problem, and technology is one lever among several. But it is a lever that most lenders have historically pulled last, if at all, when it comes to servicing. Originations gets the new platform. Servicing gets told to make the old spreadsheet work a little longer.
Servicing is the front door to the next loan, not the back half of this one
The angle I find most underappreciated in these conversations is the borrower retention piece. Most lenders think about servicing as the tail end of a transaction that is already won. The deal closed, the loan funded, and now servicing is just the administrative task of collecting payments until the loan matures or pays off. That framing misses something important. Servicing is not the back half of the transaction you already have. It is the front door to the next one.
A borrower who has a frictionless servicing experience, clear statements, easy payment options, responsive communication when something needs attention, is far more likely to come back for their next loan. They are more likely to refer a colleague or a business partner. They are more likely to take a call about a refinancing option when rates or terms shift in their favor. That borrower has effectively been sold on your organization a second time, without a single dollar spent on origination marketing.
The borrower who has a frustrating servicing experience does something different. They do not complain, usually. They do not file a formal grievance. They just quietly go somewhere else the next time they need capital, and you never hear from them again. There is no chargeback, no support ticket, no data point that shows up in a churn dashboard, because most servicing operations are not set up to measure churn in the first place. It just shows up, eventually, as a portfolio that is not growing the way it should be, and nobody can quite explain why.
This is the part of the servicing conversation that I think gets missed most often in budget discussions. When a CFO is deciding whether to fund a servicing platform upgrade, the conversation is almost always framed around operational efficiency, and that is a legitimate and important frame. But it undersells the actual return. A well-run servicing operation is a retention engine and a referral engine at the same time, and those are two of the cheapest sources of loan volume a lender can have. Cheaper than paid acquisition. Cheaper than expanding into a new vertical. And almost entirely unmeasured at most lending organizations because the systems were never built to capture that signal.
The compliance exposure nobody wants to talk about until the audit
Then there is the regulatory piece, which deserves its own attention because the risk profile has genuinely changed over the last several years. Compliance requirements in lending are not getting simpler, regardless of loan type or lending vertical. Documentation standards, disclosure requirements, and reporting obligations have all become more granular, and the expectation from regulators and auditors is that lenders can demonstrate consistency across their entire portfolio, not just produce a clean file when asked.
Lenders running manual compliance processes inside servicing are carrying more risk than most of them realize, and the risk is not just the possibility of a fine on an individual loan file. The bigger risk is what an audit finds when it starts pulling a sample of files and discovers that the gap in one file is actually a gap across a meaningful percentage of the portfolio. A manual process that works well enough when one person is handling forty accounts starts breaking down in ways that are invisible day to day once that same process is stretched across four hundred accounts and three team members with different habits. That is not a hypothetical. It is the most common way I have seen a servicing gap turn into a genuine institutional problem, and it almost always traces back to a process that depended on individual diligence rather than a system that enforced consistency.
The lenders who have gotten ahead of this did not do it by hiring more compliance staff to check more files by hand. They did it by building servicing workflows where the consistency is structural, where the documentation trail is automatic, and where an auditor can see a clean, uniform process across the entire book rather than a patchwork of individual effort. That is a fundamentally different risk posture, and it is one that only comes from treating servicing infrastructure as something worth building well, not something to patch together and hope holds.
The question worth asking instead
Here is the reframe I would offer any operations leader having this conversation internally, because I think the way the question gets asked determines the answer you end up with. The default question is what does our servicing platform cost us. That question almost always leads to a defensive posture, because the answer is a number on a budget line, and budget lines get cut when times are tight.
The better question is what is our servicing capability actually worth to us. Worth in borrower retention, because a borrower who stays is worth more than a new borrower acquired at cost. Worth in operational efficiency, because every hour your team spends reconciling data by hand is an hour not spent on borrower relationships or portfolio strategy. Worth in risk reduction, because a systemic compliance gap discovered in an audit costs far more than the infrastructure that would have prevented it. And worth in how fast your team can spot and respond to early signs of portfolio stress, which in a lending business is close to the whole game. Credit losses rarely arrive without warning. They arrive after warnings that nobody was positioned to see in time.
I keep coming back to a version of this conversation with executives who made this shift three or four years ago, often before it was an obvious or fashionable thing to do. They are running materially better operations today than they were then, and materially better operations than peers who kept treating servicing as the function you fund last. Their teams spend less time on reconciliation and more time on judgment calls that actually require a human. Their borrowers stay longer and come back more often. Their compliance posture holds up under scrutiny because it was built to, not because nobody has looked closely yet.
And the lenders still treating servicing as a cost center are going to feel that gap most acutely when volume comes back into the market. It is one thing to run an under-invested servicing operation when portfolio growth is slow and the team has enough slack to absorb the manual work. It is a different problem entirely when volume accelerates and the same manual processes that were merely inefficient become the ceiling on how much business you can actually take on. At that point, the cost of under-investing in servicing is not measured in staff hours anymore. It is measured in the growth you could not support.
Servicing was never actually the back office. It just got treated that way for long enough that most of the industry stopped questioning it. The lenders questioning it now are the ones setting the pace.
Blog
Why Your Lending Systems Don’t Talk to Each Other
I attended a webinar recently hosted by a lending technology firm that framed something I have been observing for a long time in a way I thought was worth sharing. Their thesis was simple, but it stuck with me: most lenders are not suffering from a lack of technology. They are suffering from a lack of coordination between the technology they already have.
The analogy they used was the human body. The heart pumps. The lungs process oxygen. The nervous system coordinates movement. Now imagine those systems technically functioning but never actually communicating with each other. The body would survive in spite of itself, not because of itself. That is exactly how a lot of lending operations function today.
Everyone Is Working. Nobody Is Working Together.
Originations has its system. Servicing has its own. Underwriting runs separate workflows. Compliance has different tools. Payments live somewhere else entirely. Each department is fully staffed, fully busy, and fully convinced it is doing its job well. And it probably is. The problem is not effort. The problem is that none of these systems were ever designed to operate as a single organism.
Most lenders evolved their technology incrementally over the last decade rather than architecturally. A loan origination system was adopted here. A servicing platform was added there. A CRM got bolted on. Verification vendors were plugged in. Compliance tooling was layered on top. At each individual step, the decision made sense. Someone had a problem, someone bought a tool that solved it, and the business moved forward. But the cumulative result is an ecosystem that is technically alive in every part, and functionally fragmented as a whole.
The Cost of Fragmentation Doesn’t Show Up on One Line Item
This is the part that I think gets missed in most technology conversations. Fragmentation does not show up as a single expense you can point to and fix. It is distributed across the entire operation. It shows up as longer closing cycles. It shows up as higher fallout rates, where borrowers who were qualified and interested simply give up somewhere in the middle of the process. It shows up as manual underwriting overhead, where analysts spend more time gathering and reconciling information than actually evaluating risk. And it shows up in the quiet burnout of operations teams who spend their days manually coordinating between systems that should be coordinating themselves.
One statistic from the webinar stuck with me. The Mortgage Bankers Association has reported that the average cost to originate a single mortgage loan is around eleven thousand dollars. A significant portion of that cost is not underwriting risk or regulatory compliance, both of which are unavoidable and necessary. It is what I would call the friction tax: the manual handoffs, the duplicate data entry, the human coordination layer that exists purely because systems do not automatically share information with each other. That tax gets paid on every single loan, whether anyone notices it or not.
Fragmented Ecosystem Versus Living Ecosystem
The framing I found most useful from the session was the distinction between a fragmented ecosystem and what they called a living ecosystem. In a fragmented ecosystem, each step in the process waits for the previous one to be manually closed out before it can begin. Someone downloads a verification report and uploads it somewhere else. Someone notices that a condition was cleared and sends an email to an underwriter. Someone updates a record in one system and forgets to update the corresponding record in another. Every handoff adds hours or days to the timeline. Every manual touch adds the chance of an error that will not surface until three steps later, when it is more expensive to fix.
In a living ecosystem, a single borrower action, such as submitting an application, triggers everything that can run in parallel to actually run in parallel. Verification launches automatically the moment it is needed. Income data flows directly into underwriting rules without a human physically moving it from one screen to another. The instant a condition clears, the borrower is notified by the system in seconds, rather than by whoever happened to check the queue after lunch. Same loan. Same vendors. Same people. The only difference is elapsed time and error rate. But that difference compounds across every loan in the pipeline, every month, every year.
The Real Question Isn’t Which Platform to Buy Next
What I took from this is a reframe that I think is genuinely useful for lending executives thinking about their next technology investment. The question most organizations ask is which platform should we buy next, or which vendor has the best point solution for this particular gap. That is the wrong starting question. The right starting question is how does information actually move through our business right now, and what is that costing us in time, in fallout, and in operational strain.
The lenders I talk to who are furthest ahead operationally are not necessarily the ones with the most sophisticated individual tools. In many cases their origination system or their servicing platform looks fairly ordinary on its own. What sets them apart is that they have designed the way information moves between their tools as deliberately as they designed the tools themselves. They have thought through what happens the moment a document is uploaded, the moment a condition is cleared, the moment a payment posts, and they have built the connective tissue so that those events trigger the next step automatically rather than waiting for a person to notice and act.
That coordination layer, the way data flows from application to underwriting to servicing to compliance to payment, is not an IT concern to be handled quietly in the background. It is an operational strategy. It determines how fast you can close a loan, how many borrowers you lose to friction, how much headcount you need to run the same volume, and how exposed you are to errors that come from humans doing work that should be handled by systems talking to each other.
A Practical Audit Before Your Next Platform Evaluation
The practical takeaway I would offer any COO or Head of Lending reading this is straightforward. Before your next platform evaluation, before you sit through another vendor demo, do a simple audit. Pick one loan and trace it from application to funding. Follow every manual handoff along the way. Every moment a human moved data from one system to another because the systems could not do it themselves. Every email that substituted for what should have been an automated trigger. Every place where work had to stop because someone had to check a queue before the next step could begin.
That audit will tell you more about where your operational ceiling actually is than any vendor demo will. It will show you whether your bottleneck is really a missing capability, or whether it is a coordination gap between capabilities you already own. In my experience, it is almost always the latter. Lenders rarely need another system. They need the systems they already have to behave like parts of the same organism rather than a shelf of tools that happen to sit next to each other.
Why This Matters More as Lenders Scale
This gets harder, not easier, as a lending organization grows. Every new product line, every new geography, every new regulatory requirement tends to add another tool or another manual workaround rather than removing one. Complexity accumulates quietly until a COO looks up one day and realizes that closing a loan requires touching six systems and coordinating four departments by hand, and nobody remembers exactly when it got that way. It happened one reasonable decision at a time.
The lenders who manage this well are the ones who treat their technology environment as a living system that needs deliberate architecture, not just a collection of point solutions acquired to solve whatever problem was most urgent at the time. That is a mindset shift as much as it is a technical one. It means asking, every time a new tool or workflow is introduced, not just does this solve the immediate problem, but how does this connect to everything else, and who or what is responsible for moving information across that boundary.
This is a large part of why platforms built natively on a single underlying system, such as Salesforce, have an inherent advantage for complex lenders. When origination, underwriting, servicing, and compliance data live in one connected environment instead of being stitched together after the fact through integrations and manual exports, the coordination layer stops being something your team has to build and maintain by hand. It becomes part of the platform itself. That does not eliminate the need for good operational design, but it removes an enormous amount of the friction tax that comes from systems that were never meant to talk to each other in the first place.
The bigger point stands regardless of which platform a lender chooses. Technology investment decisions should start with a clear picture of how information currently moves through the business, where it stalls, and what that stalling costs in time, in borrower experience, and in operational capacity. Buy the next tool if you genuinely need a new capability. But if what you actually have is a coordination problem dressed up as a capability gap, no amount of additional software will fix it. It will just add one more system that does not talk to the others.
Blog
Why Community Lenders Are Hiring Enterprise Tech Leaders
I had a discovery call a few weeks ago that stuck with me, and I have been thinking about it ever since. The organization was a Community Development Financial Institution, one of those mission-driven lenders that serves small businesses, affordable housing developers, and underbanked communities that traditional banks generally will not touch. I will not name them, because the details are not the point. The pattern is the point, and it is a pattern I am now seeing repeat itself across the community lending space with enough frequency that it deserves a closer look.
The person on the call was the organization’s new Deputy Director. She had spent years at Capital One running data and AI strategy. She is technical, she is sophisticated, and she had been at this community lender for about a month when we spoke. Her mandate was simple to state and enormous to execute: modernize the entire technology stack, top to bottom. Her words, not mine, were that ninety percent of what the organization does across the lending lifecycle is manual.
What Ninety Percent Manual Actually Looks Like
It is easy to hear a statistic like that and let it wash over you without absorbing what it means operationally. So let me describe what she described to us. A borrower applies using a Word document that gets emailed back and forth. Underwriting, including cash flow analysis and risk calculations, happens in Excel. Payments arrive through a mix of Venmo, paper checks, and ACH transfers, and someone on staff manually keys every single transaction into their system. They do have a loan management platform for servicing, but even that requires manual data entry after every origination because nothing upstream connects to it automatically.
Then there is accounting. The team exports spreadsheets out of their loan system and hands them to accounting to manually enter into QuickBooks. Her exact words, and I am paraphrasing only slightly, were that they are losing track of invoices that have been paid because the process is entirely manual. She was careful to say they are not mismanaging money. But she also said they are getting close to that risk, which is about as candid an admission as you will hear from someone one month into a new role.
This is a lender with a portfolio of roughly two hundred fifty to three hundred loans across commercial real estate development, small business lending, and some infrastructure deals. Small in absolute numbers, but growing quickly, and carrying the operational complexity of a much larger institution because of how diverse their loan products are. That combination — growth plus complexity plus manual process — is exactly where things start to break.
A Different Kind of Buyer Is Showing Up
Here is the part of the story that I think matters most for anyone paying attention to how lending technology decisions get made. This was not a scrappy nonprofit lender stumbling into a software search because someone finally got fed up with spreadsheets. This was a leadership team that made a deliberate decision to bring in an executive with real enterprise technology experience, specifically to fix this problem. That is a meaningfully different starting point.
When someone comes from an organization the size and sophistication of a major bank, they arrive with a frame of reference most community lenders have never had access to before. They are not wondering whether loan origination can be automated, whether payments can post themselves, or whether accounting integration is realistic. They already know all of that is possible because they have lived inside systems where it was standard, not aspirational. Their question is not “can this be done.” Their question is “why isn’t it done here yet, and how fast can we close the gap.”
That shift in framing changes everything about how a technology evaluation unfolds. It changes the pace. It changes the level of technical scrutiny. It changes what counts as an acceptable answer from a vendor. And it changes the internal appetite for actually driving change through an organization that may have operated the same way for a decade or more.
Why This Is Happening Now
I do not think this is a coincidence, and I do not think it is isolated to one organization. Community lenders, CDFIs in particular, have been under increasing pressure from multiple directions at once. Loan volumes have grown as more capital has flowed toward community development and affordable housing. Reporting requirements from funders, regulators, and government partners have become more demanding. And the workforce expectations of newer hires, especially anyone with private sector experience, have shifted. People who have worked inside modern technology environments simply will not tolerate re-keying the same loan data three or four times across disconnected systems. They have seen better, and they know better is achievable.
At the same time, funding has started to catch up with the need. In the case I am describing, the organization has grant funding specifically earmarked for technology modernization. That detail matters more than it might seem. Budget is not the obstacle. The timeline they described to us was a decision by the end of the third quarter, a signed contract in the fourth quarter, and implementation beginning in the first quarter of the following year. That is a deliberate, funded, executive-sponsored initiative, not a wish list.
I bring this up because for years the conventional wisdom in our industry was that community lenders and CDFIs were slow-moving, resource-constrained, and years behind commercial lenders in their appetite for new systems. That may have been true once. It is becoming less true every quarter. The organizations that are serious about growth are recognizing that their mission depends on operational capacity, and operational capacity depends on getting out of spreadsheets and manual entry.
What This New Buyer Sees When They Walk In The Door
The Deputy Director on our call described her vision in specific terms. She wants a cloud-based system. She wants real-time dashboarding so leadership can see portfolio performance without waiting for someone to compile a report. She wants clean, structured data instead of loan information trapped in dozens of separate files. And she wants a genuine two-way integration with their accounting system, not an export-and-re-enter workaround dressed up as an integration.
None of that is an unusual request from someone with an enterprise background. It is table stakes. What is unusual is hearing it articulated with that level of clarity by someone at a community lender with a few hundred loans in the portfolio. A few years ago, an organization that size might not have known what to ask for, let alone how to evaluate whether a vendor could deliver it. Now the person sitting across the table can tell within the first fifteen minutes of a demo whether what they are looking at is a real platform or a workaround with a nice interface.
That is the sophistication I mean when I say the buyer has changed. It is not just that the requirements are more advanced. It is that the person evaluating those requirements has the technical literacy to know the difference between a genuine integration and a scheduled data export, between a workflow engine and a series of email reminders, between a reporting dashboard and a static spreadsheet dressed up with a nicer font.
The Opportunity And The Responsibility
I think there is a real opportunity here for lending technology providers who take this seriously, and I also think there is a real responsibility that comes with it. The gap between where an organization like this one is today and where its new leadership wants it to be is enormous. Going from a Word document application and manual Excel underwriting to a connected platform with real-time visibility and automated accounting integration is not a small step. It is a full operational transformation, and it touches nearly every person on staff, from loan officers to servicing to accounting to executive leadership.
Get that transformation right, and you have given a mission-driven lender the operational capacity to serve more borrowers, close loans faster, and actually see their portfolio risk clearly instead of discovering problems weeks after they started. Get it wrong, and you have handed a resource-constrained organization a new set of technical debt and a leadership team that now has good reason to be skeptical of every vendor that comes calling after you.
This is why I keep coming back to the idea that lenders are not really buying software. They are buying operational capability. A CDFI with grant funding for modernization is not trying to check a box that says “we have a system now.” They are trying to build the capacity to grow their loan portfolio, serve more underbanked borrowers and small businesses, and do it without the risk that comes from manual, disconnected processes. The technology is the enabler. The capability is the goal.
What This Means If You Are Evaluating Your Own Stack
If you lead operations or technology at a community lender, CDFI, or any mission-driven lending organization, I would encourage you to take an honest inventory the way this Deputy Director did in her first month. Where does data get re-entered instead of flowing automatically? Where does servicing depend on someone remembering to update a spreadsheet after every transaction? Where does your accounting team receive information secondhand instead of through a real connection to your loan system? Those are the places where risk quietly accumulates, even when nobody is doing anything wrong.
The organizations moving fastest right now are the ones treating this as an operational priority backed by real budget and real technical leadership, not as a someday project. If your organization has grant funding, board attention, or a mandate for modernization, that is not a reason to slow down and study the market for another year. It is a signal that the moment to act is now, while the sponsorship and the funding are aligned. The gap between manual and modern does not close on its own, and the lenders who close it deliberately will be the ones best positioned to grow into the capital and the mission ahead of them.