Most AI startups chase breadth. They build a capable, general-purpose product and sell it across as many industries as are willing to buy it. Ari Malik, CEO of Salient, took the opposite approach. He picked one vertical, auto lending, and went deep enough that he eventually moved his company’s headquarters to Los Angeles to be closer to a single customer, Westlake Financial.
Two years later, Salient is live inside roughly half of the top 20 auto lenders in the country. By Malik’s account, the company has never lost one. In today’s episode, he walks us through how a decision that “wasn’t really ideological” turned into a full-stack servicing layer now handling collections, total loss, compliance, and loss recovery for some of the largest lenders in the U.S.
The origin story is less a strategic plan than lived experience. Malik came to the problem through the finance side of a prior company, Tesla, where he got a close look at how tangled loan servicing had become. “The workflows in this hyper-regulated part of the economy were so tangled with compliance obligations, state-specific regulations, legacy systems that weren’t API-friendly or real-time,” he said, “that we sort of felt like you could build something generic and horizontal, and actually have it work.”
Malik’s conviction has only hardened as horizontal AI tools have tried to move into consumer lending and run headlong into compliance violations. For Malik, the read was simple: a market this large and this regulated was ripe for going deep.
Watch the episode
Listen to the full podcast
Subscribe: Apple Podcasts I SoundCloud I Spotify
Read the whole transcript (for TS Pro subscribers)
Depth over breadth, by necessity
Salient’s vertical bet rests on two observations: the regulatory complexity of auto lending makes horizontal tools brittle, while the market itself is enormous. Roughly $800 billion in new U.S. auto debt is issued every year, backed by an army of human agents making calls, sending letters, and negotiating payments. “It was a structural problem that was sitting at the largest asset classes, right, and across the largest lenders in the U.S.,” Malik said.
The incumbent alternative, in his telling, was servicing handled through BPOs with little visibility and significant compliance exposure. That left room for “an excellent software solution” to compete on trust as much as efficiency. Malik is candid that there was no grand plan behind any of this – just a landscape that pointed toward one large, underserved problem rather than many small ones.
That thinking also drove Salient’s decision to relocate to Los Angeles and be in closer proximity to Westlake Financial. Malik calls it “probably the most important thing we’ve ever done as a company.” Working alongside Westlake’s compliance, servicing, and total-loss teams surfaced nuances that would have been invisible from a distance. “What we think is extraordinarily intuitive may not be for the customer,” he said.
AI maturity is landing unevenly across lenders
Malik gently pushed back on the assumption that consumer lending is still stuck at the “is this real?” stage of AI adoption. Two years ago, he said, that doubt was genuine. Salient reportedly ran a demo at the time, in which an AI-generated voice of “Steve Jobs” called lenders, and the reaction was disbelief that the call wasn’t prerecorded. That skepticism has largely faded, but unevenly.
Bulge-bracket banks, somewhat counterintuitively, have moved fastest. Malik described enterprise-grade lenders as “extraordinary fast adopters of deploying AI not only in a consumer-facing way but also at the back end on the workflow side, where they can accrue more value.” Mid-market lenders and fintechs, by contrast, have been more hesitant, which he attributes to having less exposure to the cost and regulatory pressures pushing larger institutions to act.
The pitch has changed accordingly. Where the early conversation was about risk (what happens if the AI says something wrong), the argument now, backed by data from the largest deployments, is that AI can be safer and more compliant than a human workforce, with measurable improvements in roll rates. This reframing, Malik said, “hasn’t really percolated down fully yet to the smaller fintech.”
A product stack built by customer pull
Salient started with Taylor, a voice agent handling collections, servicing, and total-loss calls. The rest of the stack, Malik said, came from lenders pulling the company deeper into their workflows rather than from an internal roadmap. Calls that appeared resolved on the surface kept spawning downstream work: a credit dispute or chargeback that got routed to a human queue and sat there for two weeks. “We were like, this is taking two weeks. Why is this taking so long? We should resolve this with the customer,” Malik said. So Salient built that piece too.
Marshall, the company’s compliance and QA product, emerged the same way. As Salient’s consumer-facing tools scaled inside regulated institutions, lenders started asking how to prove the AI itself was compliant: audit trails, model risk documentation, and system-of-record checks. Rather than lenders hiring more staff to police the new technology, Malik said the response was to “create this product called Marshall, and it will automate compliance monitoring, QA monitoring, checking the system of record, cleaning it up, and outputting almost an audit report that can go to regulators.”
From there, the suite expanded into total loss, credit-card chargebacks, and charge-offs, each one a response to a specific pain point rather than a feature added for its own sake.
Selling risk to banks, efficiency to everyone else
The sales conversation, Malik said, changes entirely depending on who’s buying. Mid-market and non-bank lenders respond to efficiency: handle time, operating expenses, recovery dollars, and scalability. Bulge-bracket banks are a different animal entirely, involving a steering committee, a compliance team, a vendor risk team, IT security, and legal review. “I had a lot more hair before we started navigating through bulge-bracket banks,” he joked.
With that audience, the real conversation is about risk reduction, model-risk documentation under frameworks like SR 11-7, and satisfying regulators such as the OCC.
Notably, Salient doesn’t have a dedicated sales team. Malik said essentially every customer has come through referrals, and he stays personally involved in every large enterprise relationship, checking in with customers weekly or monthly. He describes the signal that a lender has truly bought in not as a single moment but as a change in tone: the question stops being whether Salient can do something and becomes when. “That kind of shows that they’ve internalized this daily as part of their almost key operating model, rather than a pilot that’s constantly being run,” he said.
The clearest evidence of that model working is Westlake, which set Salient a target of $15 million in annual savings. Deploying the technology across collections, total loss, and compliance didn’t just cut costs. Average handle time dropped, compliance violations dropped, and findings dropped alongside them. Malik said Salient’s mapping of Westlake’s customer interactions found that roughly 80% of use cases could be handled entirely by AI, freeing human agents for cases that actually require judgment.
The pitch, as he frames it, moved “from just being this weird, fun toy or experiment of generative AI… into a service quality story, into a risk mitigation story.”
What the industry is still underestimating
Malik sees the biggest opportunity in making loss recovery predictive and individualized. Historically, treatment has been bucketed by simple delinquency rules rather than who a customer actually is. His example: two customers each miss two payments in the last 60 days, one with a poor payment history and the other who has never missed a payment. Under most current systems, they’re treated identically. Malik wants recovery decisions to reflect those differences, keeping good customers with the brand through a difficult stretch rather than pushing them out over one short-lived data point.
Looking ahead, Salient plans to launch three new products by the end of 2026, with Malik less involved in the day-to-day than he was in the company’s earlier years. “We’d really like to do it without me being super in the weeds on day one, but trusting my team to do more of it,” he said.
It’s a fitting next step for a founder who has spent two years arguing that the way to serve a fragmented, regulated industry well is to get close enough to understand it completely, then build precisely for what you find.
The post Why Salient bet everything on auto lending and never looked back appeared first on Tearsheet.