Vivek Jetley, president and head of Insurance, Healthcare and Life Sciences at EXL, a global data and AI company serving Fortune 500 firms.
Welcome to the era of AI overcorrection. First, investors thought AI would eat the software industry. Now, the plot has shifted, and investors worry they may be facing a reality check as some valuations have started to tilt into bubble territory. Both are missing the bigger point.
The specialized software solutions driving the world’s financial services, insurance and healthcare functions are too embedded and complex to be replaced by off-the-shelf large language models.
But when highly trained specialists who understand the inner workings of those complex workflows start applying AI in the process, the old way of doing things will most certainly evolve, making it possible for businesses to start reimagining the way they operate.
AI Transformation: Not A Zero-Sum Game
That is the critical detail that often gets overlooked in most mainstream reports about the future of the software industry.
Yes, AI is fundamentally changing the way businesses think about software spend. And, yes, many businesses are experiencing challenges integrating AI into legacy functions and complex workflows.
The fact that both are true is exactly the point. AI’s success is not a binary, either-or scenario in which one industry group will win at the expense of the others.
As AI becomes more deeply entrenched in the inner workings of business processes, it is exposing the need for a new type of collaboration between businesses, software providers and frontier AI labs.
Addressing AI’s Change Management Challenge
We’re seeing this phenomenon play out today with the huge gap that’s emerged between enterprises struggling with AI and those that have managed to integrate AI across core workflows.
BCG reported last year that 50% of companies are still stuck in early-stage AI adoption. My firm’s recent survey of 322 C-suite and other senior decision-makers in the banking and finance, insurance, retail, utilities, life sciences and healthcare payer industries found that just 10% of companies have deployed AI past the pilot stage across nearly all core business functions.
Digging deeper into what’s keeping businesses from broader AI integration, it becomes clear that the problem is often business context, not technology. Getting to a level of large-scale integration of any new technology requires more than just throwing the new tech into an existing workflow.
Countless systems and processes have been built up over decades of refinement that need to be addressed before the status quo can be overhauled. As someone who leads AI integration work for insurance, healthcare and financial services clients, I’ve seen this repeatedly when working with senior leaders on enterprise AI transformation.
This is why incumbent SaaS technology players are still so important. There is a reason the specialized software developed for complex business operations plays such a central role in the modern economy. It has been refined and pressure-tested for high-stakes work where there is zero margin for error.
The idea that an off-the-shelf AI model could suddenly replace those systems of record is not grounded in the in-the-trenches reality of how these software systems work.
A New Approach To Collaborative Innovation
Business leaders evaluating new AI capabilities are acutely aware of the difference between a highly specialized piece of business software and a generalist AI model. They are not looking to replace software that has become a core part of their operations, but to find ways to integrate AI into those operations.
Accordingly, some of the most exciting innovations happening inside large organizations today are playing out deep in the weeds of operational workflows.
For example, I recently worked on an AI integration project with a major life insurance and annuity provider to enhance their platform. The client didn’t want to replace their platform but to evaluate risk, transfer books of business and modernize processes more efficiently without abandoning decades of infrastructure.
Embedding AI into existing life insurance platforms is helping firms streamline new business acquisition, risk transfer and compliance. In financial services, leaders are migrating legacy codebases to more modern Python configurations, which support more agile order management and compliance workflows.
These projects are not about replacing current tech infrastructure, but about improving it over time through constant nips and tucks to processes that were not as efficient as they could be.
This does not mean SaaS-based businesses can afford to sit back and relax while AI-driven innovation happens around them. There is a very real risk for legacy software models that cannot be upgraded, integrated across disparate data sets and enhanced to support AI-enabled workflows.
For businesses relying on third-party software providers for countless key functions, it’s essential to make sure AI investments being made at the enterprise level can be supported by technology vendors.
No One Will ‘Vibe-Code’ Their Way To An Enterprise Solution
Companies that try to replace all of their software with homemade AI agents, or bring in forward-deployed engineers from a frontier AI model developer expecting instant, enterprise-wide deployment, often run into more friction than they anticipate.
When high-stakes operational workflows are subject to strict regulation and detailed scrutiny, the existing guardrails and protocols cannot simply be replaced by vibe-coded hacks or paved over with new AI models.
The change happening inside these companies is far more nuanced. Experts in these complex workflows are implementing AI in pockets, carefully choreographing incremental improvements, not replacing decades of innovation with DIY projects.
For example, in the life insurance transformation project mentioned earlier, one of the first steps was to use AI to help migrate the codebase from COBOL to Python, which is a much more modern and flexible architecture. While this may sound like a back-office technicality, it is a critical piece of unlocking an AI-enabled future, because Python supports the real-time inference and API-driven architectures that modern AI systems require.
A New Way of Thinking About SaaS Business Models
The legacy software industry won’t be replaced, but AI-driven business efficiency gains will force business models to evolve, requiring new pricing, new schedules and collaboration between vendors, clients, frontier AI developers and AI services firms increasingly connecting them.
Getting enterprise AI right is rapidly becoming an exercise in choreography, assembling data, technology and expertise to build new workflows. All players still have a role, but some must learn new moves before full integration is realized.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
