The value of Odessa in a world of AI coding agents

If you’ve spent any time with the latest AI coding tools, you’ll understand why the question is being asked more openly now: do you truly need to buy specialized software anymore? Tools like Claude Code and Cursor have changed what’s possible.

Developers can move faster, non-technical team members can contribute more directly, and software that once required large specialist teams can now be approached with smaller ones. The distance between “we need this” and “this exists” has genuinely shortened.

It’s a fair question to ask – and one worth taking seriously.

Writing code has never been the hard part

Building asset finance software has never been challenging because of the code, but because of the complexity of the domain.

Think about what a modern lease or loan origination workflow contains: credit decisioning logic, regulatory compliance across multiple jurisdictions, tax and accounting treatments that vary by asset class, integrations with bureau data, document generation, e-signature flows, funding coordination, and the list goes on. Before a single line of code is written, someone has to understand all of that. They must know the right questions to ask, the critical edge cases, what “correct” looks like – and be able to empathize with the person whose job it is to approve that funding.

An AI coding agent can turn a set of requirements into working software. What it cannot do is define those requirements in a way that truly reflects the nuances of the business. That still requires someone who has spent years in this industry, watching how the details compound, seeing which shortcuts create problems down the line, and understanding what the business is really asking for even when the brief is incomplete.

That expertise takes time to accumulate. At Odessa, we’ve been serving exclusively in asset finance for over 27 years. That’s not a number we cite for its own sake. It represents a specific kind of institutional knowledge – about what the industry needs, how it’s evolved, and where it’s heading – that can’t be reverse-engineered from a few prompts.

What a diverse client base teaches you

There’s another dimension to this that’s easy to underestimate. We work with some of the world’s leading asset finance companies globally, all running on a single unified platform. Not a collection of acquired products stitched together, but one platform, built with a consistent architecture and set of design principles.

That matters for a specific reason: when you serve that range of clients on one platform, you learn things you can’t learn any other way. You see how different organizations solve the same problem. You discover what works across market conditions, geographies, and portfolio types. You develop an instinct for what constitutes genuine best practice versus what just happened to work in one context.

Coding agents don’t have that. They can generate sophisticated logic based on their training data. They can’t draw on the experience of having built, deployed, and iterated on the same core platform across hundreds of client environments over nearly three decades.

Why the build-or-buy question isn't as simple as it looks

Some organizations respond to the cost and complexity of a software platform by taking a different route: hiring a digital agency or development team to build something bespoke. With today’s AI tools, those teams can move faster than they used to, and the case for going custom has become easier to make on paper.

But faster at what, exactly? If what you’re building is a standard internal tool or a customer-facing application without deep regulatory constraints, this approach can make a lot of sense. Asset finance is a different matter. The domain knowledge required to build software that handles the full lifecycle of a lease or loan portfolio correctly – not just technically, but compliantly, operationally, and at scale – is specialized in ways that take years to internalize.

There’s also a side of AI-assisted development that doesn’t get discussed enough: what happens after the code is written. AI coding tools can generate large volumes of code quickly, but that code still needs to be tested, validated, and maintained. Without deep domain knowledge guiding what’s being built and why, teams often find themselves with a codebase that works well enough in isolation but becomes increasingly difficult and costly to maintain as requirements evolve. The speed of generation can mask the true cost of ownership – and in a regulated industry like asset finance, that cost has a habit of surfacing at the worst possible time.

In practice, asset finance has a way of revealing complexity that wasn’t visible at the start. Workflows that seemed simple turn out to have regulatory dependencies. The further you get from the original build team, the harder it becomes to course-correct.

When judgment is wrong – when a workflow mishandles a regulatory requirement, or an edge case slips through that someone with experience would have caught – the consequences are real. Contracts are affected. Customers are affected. Perhaps regulators are involved. The question of who owns the fix, and how quickly they can resolve it, becomes urgent in a way that no prototype ever prepared for. What self-build often buys you is speed today – and complexity tomorrow.

If the primary skill a software provider brings is the ability to write code, then you may as well ask an agent to do it. The difference is whether they bring something beyond the code: the judgment about what the code should do, the experience to anticipate what will go wrong, and the accountability to stand behind it when it does.

The role of people is changing – and that's a good thing

AI is changing the composition of the teams that build and maintain software. It’s reducing the number of steps between a business user who understands a problem and a developer who can solve it. That’s genuinely valuable. It means the person closest to the problem – the one who really understands why it matters and what it needs to do – can be more directly involved in shaping the solution. Less translation, fewer handoffs, better outcomes.

What this shift really does is change which skills matter most. When writing code is no longer the bottleneck, the bottleneck becomes understanding: understanding the client’s business, asking the right questions, knowing when a technically correct answer is still the wrong answer for this particular situation. Those are deeply human skills, and they’re becoming more central to good software delivery, not less.

Skilled engineers remain at the heart of what we do. What’s changing is what’s expected of them. The best engineers in this space were never just people who could write code. They were people who understood the domain well enough to push back on a brief, spot a design flaw before it became a production problem, and make judgment calls that no tool can make for them. AI amplifies that capability; it doesn’t replace it.

The people who truly understand how asset finance works will matter more as AI handles more of the routine work. They are the ones who can sit with a client, challenge a preconception, and translate what they hear into something the platform can deliver. The value has always been in the judgment.

Built to the standards that matter

An AI coding agent can help you build software. It cannot take responsibility for how that software is operated, secured, or audited. In financial services, that distinction matters enormously.

Running a platform that meets the operational and security standards financial institutions are held to requires deliberate investment – in infrastructure, in processes, and in the people who maintain them. Odessa Cloud and our Enterprise Customer Support (ECS) offering are built around exactly that. Our SOC 2 and ISO 27001 certifications aren’t incidental. They reflect years of work to ensure that what we build, we can also run – reliably, securely, and to a standard our clients can point to with confidence.

Partnership isn't a feature

None of this is abstract for us. Implementing a core platform isn’t a project with an end date. It’s an ongoing relationship that evolves as our clients’ businesses evolve, as regulations change, and as new capabilities become available.

You can’t get that from an AI agent. What makes those relationships work is the human dimension: the accumulated understanding of a client’s portfolio, their team dynamics, their strategic priorities, and often the specific pressures they’re navigating at any given time. That context is what allows us to give advice that’s useful, rather than advice that’s technically correct but disconnected from what’s really going on.

Local insight, global depth

The final piece is one that’s genuinely difficult to replicate through build-your-own approaches: the combination of local market knowledge and the depth that comes from being a global organization.

Asset finance looks different in North America from how it does in Europe or Asia-Pacific. Regulatory frameworks differ. Market conventions differ. Customer expectations differ. At the same time, the underlying platform complexity is the same.

What Odessa offers is teams who understand the local context – who know the regulatory environment, who speak the language of the market, who have worked through the specifics of that region – backed by a platform that has been tested and refined across global deployments.

The real question worth asking

So, back to where we started. Can you build your own asset finance software with today’s AI tools? More easily than before, yes. The question isn’t whether it’s technically possible. The question is whether the result will be as good over time, as working with an organization that has spent over 27 years specializing in building just that.

We’ve built something that’s impossible to replicate: a proven track record, a platform that reflects real-world complexity, and a set of relationships grounded in genuine care about whether our clients succeed. If you strip it back to the business case, what you’re really choosing when you choose a platform over a bespoke build is something more fundamental than software.

It’s shared accountability – a partner who carries institutional memory that doesn’t walk out the door when a key person leaves, who has seen your category of problem before, and who has a vested interest in getting it right. Bespoke systems are often only as resilient as the team that built them. When that team changes, the knowledge tends to go with them.

AI will keep making everyone more productive – including us. But it won’t replace the judgment, the experience, or the accountability that comes from having skin in the game for nearly three decades.

We’d be glad to connect if you’d like to talk about this further.