Four tech and operations leaders discuss why the real competitive edge lies in targeted workflows, proprietary data and empowering teams to make better, faster decisions.

Artificial intelligence has moved from buzzword to boardroom priority across equipment finance, but the gap between lenders extracting real value and those still running pilots keeps widening. The technology itself is rarely the differentiator anymore. What separates the two groups is how disciplined they are about the problem they’re solving, the data they trust and the people they put around the process.

To dig into what that discipline looks like in practice, four executives who sit at the intersection of technology and operations in equipment finance came together for a roundtable conversation. Between them, they have watched the industry cycle through digitization, automation and now a wave of AI adoption that touches everything from credit decisioning to back-office servicing.

In the conversation that follows, the four discuss what actually separates lenders getting measurable value from AI from those still experimenting, where competitive advantage is shifting as speed becomes table stakes, the outdated assumptions still shaping technology decisions, how risk management has to evolve alongside automation, what the talent pipeline needs to look like as routine work disappears, and the one mindset shift they would recommend to any executive building a technology strategy.

Their answers diverge on some points and converge sharply on others, but all four keep returning to the same theme: Technology is not the strategy. It is only as good as the business problem it is pointed at.

AI is the dominant conversation in equipment finance right now, but adoption levels vary widely across the industry. In your view, what separates lenders who are getting real value from AI from those who are just experimenting with it?

Dan Levin: They started with a specific pain point, and didn’t focus purely on AI as a technology, but rather the business problem they were trying to solve. They picked something contained — for example, application intake, document review, exception triage — then set a baseline and measured whether it moved. That forces the invisible or undervalued work first, such as cleaning up data and writing down credit policy that had been living in three people’s heads for 15 years. The ones stuck experimenting did the reverse. They started with the tool, went looking for somewhere to apply it, and stood up a pilot nobody owned and nothing depended on. Keep it small enough to manage, aim it at a problem people actually complain about, and attach a number to it. If you can name the metric that moved, you’re getting value. If you can’t, you’re still experimenting.

Jeff Lezinski: It all boils down to whether lenders are using AI to solve real business problems or challenges rather than investing in AI simply because everyone else is. Lenders who identify specific areas of the business that they can optimize and make more efficient through AI tools (e.g., IDP, workflow prioritization, etc.) and apply AI to advanced predictive models that minimize risk (credit default risk models) while maximizing revenue, such as end-of-lease promotions, will outperform those who are not diving deep into specific business cases to use AI.

Shivi Sharma: The lenders getting real value start with a specific problem. They’re asking, “Where is my team spending time they shouldn’t be?” and work backward from there versus saying, “Let’s use AI.” The ones experimenting are doing the opposite, where they read about an AI tool like Claude Code and then look for a problem to attach it to. In equipment finance, the highest-friction moments are generally predictable, like document intake, business verification, financial spreading and credit memo drafting. So lenders who trained AI on those exact workflows and problems they identify are seeing measurable throughput gains.

Lara Tolland: Real value starts with picking one specific thing you are trying to move, say cycle time or approval accuracy. From there, it comes down to the rollout, getting everyone moving in the same direction and working toward that new metric in a new way, instead of running the old process through a new tool with new labels on it. Lenders getting real value from AI have done that work up front. The team understands what is changing and why, and they operate differently because of it. Lastly, these lenders are measuring their objectives, so they know if they are meeting their goals or need to further refine.

Every wave of new technology tends to shift where the real competitive advantage sits. In equipment finance today, where do you think that advantage is actually coming from, and is that changing?

Levin: Speed was the differentiator for the last decade. It’s now table stakes, and speed without discipline just creates chaos faster. On the technology side, the advantage sits in security, in the ability to actually execute and above all in proprietary data — what you’ve funded, what performed, by asset class, vendor, geography and channel. Nobody else has your portfolio history. AI can’t invent it, and no vendor can sell it to you. Relationships still matter enormously in vendor and dealer channels, but relationships increasingly go to whoever is easiest to transact with, which loops back to data, process and execution. What’s changing is the constraint. It has moved from getting the data to trusting it enough to decide on.

Lezinski: This really depends on a number of factors, but a clear one is the ticket size. In small- and micro-ticket transactions, speed is undoubtedly where competitive advantage sits, enabled by technological enhancements. The faster a deal can get approved and funded is clearly an important competitive variable in that part of the equipment finance market. On the other hand, when looking at large-ticket transactions, relationships still matter most in getting those deals done, with technology playing a smaller, indirect role there.

Sharma: Speed of funding is now a baseline expectation from vendors. The real advantage is decision quality at scale, and the ability to underwrite more deals, faster, without increasing risk exposure or cost or headcount or team burnout. Doing so requires automation that actually understands equipment finance, not just a generic AI LLM bolted onto a workflow.

Tolland: The advantage is going to lenders who run every technology decision through a simple filter: Does it increase revenue or reduce cost? Everything else is noise. On the revenue side, that means a better borrower and vendor experience, less friction to submit and close a deal, and better credit data that improves portfolio performance over time, not just faster approvals. On the cost side, it means real operational efficiency, not just moving the same manual work into a new interface. Lenders adopting technology because it’s new, without tying it to one of those two outcomes, are the ones stuck experimenting instead of getting real value.

What’s a widely held assumption about technology in equipment finance that you think is wrong or outdated?

Levin: That modernization means replacing the core system. Rip-and-replace is the most expensive way an organization can avoid deciding how it wants to operate. These programs consume two or three years and enormous capital, and they often deliver the same workflows on newer infrastructure, because nobody redesigned the process — they just migrated it. The bigger mistake is what follows: Once a project that size is underway, everything else waits. Technology moves too fast for that. Continuous, incremental improvement is what keeps you from needing the next major redesign. Most of the real gains sit in the connective tissue anyway — data access, integrations, decisioning layers and the manual handoffs between systems. Fix those, and the core system question usually gets smaller.

Lezinski: In several instances, I’ve seen the concepts of automation and digital transformation heavily focused on the front-end origination process, where the sales focus sits naturally. However, both concepts can be applied to back-office servicing. There is significant room for lenders to drive greater automation on the back end, leading to improved efficiency. Additionally, “digitizing” servicing processes can also enhance customer experience; for example, communicating end-of-term options through email, texts, bots, portals, etc., and enabling two-way communication are things customers increasingly expect.

Sharma: That AI needs perfect, structured data to work. Equipment finance runs on messy inputs. How many lenders still receive handwritten apps and scanned PDFs? The best AI systems today are built to handle exactly that messiness. The ones that require structured input are the ones that don’t actually solve the problem. If your AI tool can’t read a handwritten credit app or extract meaningful data from a scanned bank statement, it isn’t built for this industry.

Tolland: A common assumption is that digital transformation is a technology initiative, something IT or a vendor delivers and the organization simply adopts. In practice, the technology is the easy part. The harder work is whether people actually change how they operate, whether a faster system leads to faster decisions or just a faster version of the same manual habits. Real transformation is a leadership and culture problem that happens to involve technology. Whether the organization is willing to work differently is the part worth spending real time on.

Credit risk, fraud and underwriting standards are under constant pressure as deal volume and speed expectations rise. How do you think the industry should be thinking differently about risk in an increasingly automated environment?

Levin: Start by understanding how much fraud risk has changed. The same capability that automates legitimate workflows lets bad actors produce convincing synthetic identities and documentation at volume, cheaply. That is the argument for automating the noise — intake, document handling, stipulation chasing, deal prioritization — and for pushing verification upstream into intake rather than leaving it downstream as a review step. Underwriters then spend their judgment on the deals you want to take on and stop wasting it on the ones you don’t. Where decisions are automated, override rates, approval mix by channel and rule performance have to be monitored continuously, the way you already watch delinquency, not reviewed once a year by a policy committee. Every automated decision also needs a reason an examiner and a customer can both read; explainability is an architectural requirement, not a reporting feature.

Lezinski: Each lender will still need to determine its own appetite for risk. Historically, lenders have relied on hard-coded algorithms to set their risk thresholds. As AI-driven credit decisioning models evolve, lenders need to rethink their approach to risk, since these models do not operate on fixed rules set in stone. Continuous validation and evaluation of new models is essential to ensure they remain aligned with the lender’s stated risk appetite.

Sharma: We need to stop treating risk evaluation as separate from automation. They go hand in hand. The industry still runs risk as a tollbooth. Every deal stops, a human inspects, throughput suffers. That breaks when volume triples and brokers expect same-day answers. Automation without risk assessment isn’t useful enough for teams. Speed isn’t the enemy of good credit. Inconsistency is. Automation can help surface inconsistencies better, eventually helping risk teams. Automatically flagging tampered documents and inconsistencies before a human ever opens the file can change a deal’s trajectory entirely.

Tolland: Too much of underwriting still applies the same level of scrutiny to every deal, no matter the size or risk profile. That is the outdated model, not automation itself. The industry needs automation that differentiates scrutiny, not one that just speeds up a flat checklist. A deal with red flags should get surfaced for investigation right away. A straightforward $100,000 deal should not move through the same review path as a complex $1 million one. Done right, automation means oversight applied where it’s actually needed, instead of spread evenly across every deal whether it needs it or not.

Talent and workforce are a recurring concern in equipment finance. As automation takes on more of the operational workload, how do you see the skills and roles that lenders need most changing over the next few years?

Levin: In my world, we use the term “full-stack” for people who understand the whole technical stack, not just one specialty. That idea will spread to every function. Lenders will need people who understand credit and technology well enough to translate between them, who can take underwriting judgment and express it in data, rules and process. As routine processing gets automated, the human work concentrates in the hard exceptions, which takes more judgment per transaction, not less. The risk nobody is planning for — or that some are even chasing without knowing it — is the gap with entry-level people. Junior staff learned this business by doing the routine work now being automated. Remove the training ground without deliberately replacing it, and in 10 years you will have excellent automation and nobody qualified to supervise it. I still think the junior roles are important and should not be overlooked or an afterthought.

Lezinski: There will be a shift from traditional back-office roles — such as validating documents, performing reconciliations, etc. — toward roles that require human judgment and specialized domain expertise. For example, in the near term, AI agents will not be making 100% of a lender’s credit decisions or deciding on a complex restructuring plan for a large corporate lease transaction.

Sharma: It is important to keep bringing fresh talent to the industry. The lenders who will win on talent are the ones who stop hiring people to do data entry and start hiring people who can use exceptional judgment and use AI to make themselves more efficient. For example, the credit analyst role has evolved significantly. The manual work of pulling SOS data, spreading financials, verifying addresses and drafting credit memos is increasingly automatable. What isn’t automatable is judgment: the ability to read a borrower’s story, structure around a credit challenge and make a call on a deal that doesn’t fit the model. A good credit analyst now needs exceptional judgment rather than the skill to merely manage operational workload.

Tolland: Automation is going to take the mundane, repetitive work off people’s plates, and that’s a good thing, because the relationship has always been the real value in equipment finance, and that’s not going away. What changes is what people need to be good at. As more of the process runs itself, judgment becomes the thing that differentiates someone — knowing which deal needs a second look, or when a client needs a human on the phone instead of an automated update. Lenders need to hire and train for judgment on purpose.

If you could give equipment finance executives one piece of advice about how to approach technology strategy — not which tools to buy, but how to approach decision-making — what would it be?

Levin: Treat technology as an investment, not an expense. Approach it as a cost line item, or as a way to cut headcount, and that is exactly what you get. Steady investment is what keeps you competitive and gives your people room to grow. I’ll also add: Start from the pain point, not the tool. Most technology decisions go wrong because they begin with a product evaluation instead of a business position. Be specific about what “good” looks like three years out, find the smallest change that proves you are on the right path, and protect your optionality — you will replace most of your tools within five years, but never your data. Own your data and your decision logic and workflows and give every initiative an owner and a number.

Lezinski: Always start with the business first, not the technology. Ask what business value you want to create or what problem you are trying to solve. Then look to technology to enable that value or solve that problem. Too often, businesses choose their technology solutions first, without letting the business drive the decision.

Sharma: Audit your team’s actual time before you evaluate any vendor — not what you think they do, but what they actually do. For example, one of our customers actually took a stopwatch to calculate how long it takes his credit team from application to credit decision — not the perceived latency, but the actual latency from the moment an application hits the inbox. This type of data point tells you exactly where automation creates real return, and it protects you from buying a solution to a problem that isn’t actually your problem. Start with identification of your own challenge, not the tool.

Tolland: Start by naming the business problems in front of you. The harder part is deciding which one is the priority right now, and that shouldn’t be a one-off judgment call. Build a simple rating system and score every potential priority against the same criteria — for example, reach, impact, cost and whether you can actually execute it. Whatever scores highest using your system is what gets the resources and the attention. The rest waits, even if it’s still worth doing eventually. Having one consistent system, instead of deciding case by case, is what turns technology strategy into something repeatable instead of reactive.

Taken together, these four perspectives sketch an industry at a real inflection point, not just another technology cycle. The lenders pulling ahead are not necessarily the ones with the most sophisticated tools. They are the ones who started with a business problem, measured whether their fix moved a number and built the discipline around data, around risk monitoring and around talent development to scale what worked.

What is notable is how little of that discipline is about AI itself. Naming a metric before adopting a tool, pushing verification upstream instead of treating it as a downstream check, matching the level of scrutiny to the size of the risk, deciding deliberately how junior staff will still learn the business: None of it requires a single new algorithm. It requires leadership willing to make the harder calls technology cannot make for them.

As AI capability continues to compound, the gap between lenders treating it as a strategic investment and those treating it as a checkbox will likely only grow. The panelists’ shared advice cuts through a lot of the noise: Start with the pain point, own your data, measure what matters, and build the judgment, human and automated, to act on what you find.