Forward Deployed: Why the AI Era Belongs to Services

Everyone is racing to deploy AI. Patricio Pazmiño, Chief Product Officer at Kin Analytics, thinks most of them are solving the wrong problem.

In this episode of The Hive by Kin, Patricio draws on nearly eight years of experience applying AI in equipment finance, starting as a data scientist building credit models and now leading product, to make a case that cuts against the current wave of AI enthusiasm. Building technology has never been the hardest part. As AI makes software faster and cheaper to build, the real differentiator is shifting to something that takes years to earn: deep business expertise, close customer collaboration, and the ability to navigate business complexity.

The Real Barrier Is Business Complexity 

Some lenders still operate on legacy systems that can't even connect to APIs. In those cases, modernizing the technology stack is necessary. But those situations are the exception rather than the rule. More often, technology is already capable of supporting the business. The real barrier sits behind it: disconnected systems, fragmented data, and operational knowledge that lives only in people's heads. As Patricio explains, replacing the technology doesn't solve those underlying challenges.

Part of the problem is that clients often arrive with a solution already in mind. "Most of the time, people who need to solve a problem don't even understand what problem they're trying to solve," he notes. When the AI boom hit, many came asking for AI for its own sake, while the real work was upstream: cleaning data, fixing the stack, and understanding the underlying problem before introducing new technology.

AI Hasn't Changed the Game. It's Exposed What Was Always True

The latest wave of AI hasn't created a new challenge. It has exposed an existing one. As building software becomes faster, easier, and cheaper, the challenge shifts away from development and toward understanding how a business actually works.

"Technology is not going to navigate itself through your business complexity," Patricio says. On one side sits a business's accumulated complexity; on the other, the future state AI could unlock. That gap is exactly what AI can't cross on its own, and it's why companies like Kin exist.

As building software becomes more accessible, the competitive advantage shifts from writing code to understanding the business. That's where business expertise creates value.

Automating for Speed vs. Automating for Risk

Underwriting is where this distinction becomes tangible. Lenders, Patricio stresses, are in the business of managing risk — "that's not going to change no matter what technology comes into place." So automation should serve risk, not just speed. Freeing underwriters from data validation so they can focus on real risk analysis is a win. Automating for its own sake is dangerous: "Maybe you're going to be just operating faster but at a higher risk" — moving wrong deals from inbox to books faster.

A risk-first approach treats every automated step as an input to smarter decisions. Extract the data from application documents with AI instead of analysts grabbing only the minimum fields, and you can store it as a new input for your risk model. The problem across the industry is disconnection: ad hoc tools automate one step but never sync back, so the data is lost and downstream decisions never improve.

What Forward Deployment Actually Looks Like 

"Forward deployed" is how Kin has worked for years, long before the term became popular. The principle is simple: business complexity is nearly impossible to understand from a distance. If you interview a user, build what they described, and you often discover it wasn't what they actually needed, triggering months of back-and-forth. The alternative is to work alongside the business, become the user, and build with that context. "Instead of seeing value in six months, you see value in a month," Patricio says. 

The Client Story: Driving Adoption from 30% to 85%+ 

One client engagement illustrates why. A client struggled to move deals from the inbox to the system so their credit team could analyze them. Kin started with a light version of forward deployment: discovery sessions, weekly meetings, and a solid MVP that received great feedback. But adoption didn't match. "Users tend to give you feedback that is more positive than what they are actually receiving," Patricio notes.

So Kin went further: two team members were literally onboarded into the client's submissions team, with an email, system access, and real deals to process. "Instead of uncovering the process from some sort of distance, we were actually doing the work," he says — "and things just completely changed." Iterating twice a week, adoption climbed from around 30% in the first month to 85%+ within a couple of months, and users shifted from being pushed to use the tool to asking to onboard more of their team.

Why Adoption Is Where Projects Fail 

Adoption is where most projects fail, and Patricio sees one main cause: users can't adopt a tool when they feel left out of the process. When someone "just comes one day with 'hey, now you have to use this,'" it fails. When users are involved from the first discovery session, they engage. That's why implementation is a milestone, not the finish line. Value is only created when people actually change the way they work. "We don't want to be vendors, we want to be partners." 

The End of the Vendor Era

SaaS defined the last two decades by making software easier to buy, faster to deploy, and simpler to scale. That model worked well for standardized workflows, but equipment finance has never been a standardized business. Every lender has its own processes, risk appetite, and operational complexity.

As AI continues to reduce the cost and effort of building software, technology alone is no longer a competitive advantage. The advantage shifts to something much harder to replicate: deep industry expertise and the ability to understand how a business actually works.

He points to Salesforce's evolving platform strategy as a signal of where the industry is heading. The long-term value isn't just the interface. It's the infrastructure, the datasets, the integrations, and the APIs built over years that allow businesses to build solutions tailored to the way they actually operate.

The conversation ultimately comes back to the same idea: don't start with the technology. Start with the business problem.

Because lenders aren't in the business of building software. They're in the business of managing risk. The organizations that succeed with AI won't simply be the ones adopting new technology. They'll be the ones combining AI with operational knowledge, close customer collaboration, and solutions designed around the way their business actually works. In Patricio's view, that's what will define the AI era: not software alone, but the expertise and partnerships that turn technology into business outcomes.

Frequently Asked Questions

What does "forward deployed" mean in AI services?
It means embedding directly inside a client's business, doing the actual workflow alongside their team and building with that firsthand context rather than from interviews at a distance. It compresses time-to-value from months to weeks by enabling rapid iteration with full context.

Why does business expertise matter more than software in the AI era?
Because AI has made building faster and cheaper, the technology itself is easy to replicate. The lasting advantages are deep vertical expertise, closeness to customers, trust, and the ability to navigate business complexity — all of which take years to earn.

What is the difference between automating for speed and automating for risk?
Automating for speed just processes applications faster. Automating for risk treats each step as an input that improves credit decisions, like extracting and storing application data to strengthen the risk model. Speeding up a flawed decision only produces wrong answers faster.

Why is AI adoption difficult in equipment finance?
AI adoption in equipment finance is challenging because the biggest obstacles usually aren't the models themselves. Disconnected systems, fragmented data, manual workflows, and industry-specific processes make implementation far more complex than simply deploying new technology. Successful adoption requires combining AI with deep operational knowledge and close collaboration between technology teams and business users. 

‍/Like this post?

You’ll love our newsletter.