Humano Seguros

Juan Carlos Restrepo

Business Integration Catalyst

Juan Carlos Restrepo is Vice President of Technology at Humano Seguros, a Dominican insurance company offering health, life, auto, and property insurance. He has over 25 years of experience in technology leadership and digital transformation across the insurance and manufacturing industries.

When AI Development Outpaces Business Decisions

I never imagined that developing software too quickly would become a problem for our IT organization. For years, we worked to improve our teams’ productivity, eventually adopting agentic development teams capable of delivering features to production in just two days instead of a two-week sprint. We hadn’t anticipated that our developers would end up waiting for business decisions.

A few months earlier, we had started incorporating AI into application development. Initially, our developers used AI-generated code suggestions while maintaining their traditional practices. Only 18 percent to 22 percent of the implemented code came from those suggestions, and development speed improved by approximately 10 percent.

We knew we could do better. We began experimenting with agentic development teams, where developers moved from writing most of the code to supervising AI agents, providing context, giving instructions, and reviewing results.

The difference was significant. Features that previously required a two-week sprint to reach production could now be delivered in just two days, provided business specialists were immediately available to perform functional testing and validate the solutions.

We decided to apply this approach to a real project: a system supporting one of our core health insurance authorization processes. Its complexity and resource requirements had kept it in our backlog for quite some time. We started with Scrum, as usual. But within hours, we were running out of user stories.

When we reviewed productivity metrics, we found developers spending considerable time waiting for answers to business questions. Business specialists told us they hadn’t expected those questions so early. Under our previous development model, many of those decisions would have come up months later. For the first time, IT was waiting for the business. We had changed the speed of software development without changing how quickly the organization could make decisions.

Turning Faster Development into Greater IT Capacity

That experience forced us to reconsider how we worked. In the teams adopting this model, we moved away from traditional sprints toward continuous delivery every two or three days. Daily stand-ups became technical discussions about architecture, cybersecurity, and interoperability.

We also started changing our team structures. Teams that previously needed four developers began evolving into smaller units where one developer, supported by AI agents, could handle much of the work.

  What we are learning is that agentic development is not only changing how we write code. It is making us reconsider how we organize IT, develop our people, and work with the business.  

But the most important change wasn’t having fewer developers per team. It was what we could do with the capacity we freed up. Some developers took on projects we previously outsourced, while others focused on reducing technical debt that had accumulated over the years.

Instead of measuring AI’s impact only by how much faster we could develop software, we began using that additional capacity to address needs we previously couldn’t handle with our own resources.

Establishing the Guardrails for AI-Generated Software

This also made us realize that our developers’ skills needed to evolve. Knowing a programming language or framework was no longer enough. We needed professionals who could understand business processes, discuss business rules, and work directly with users to define solutions.

We also underestimated the effort required to define the architecture, design, and cybersecurity guidelines that would allow AI agents to build applications consistently and in line with our standards.

Although we can now deliver features to production in just two days, that doesn’t mean we’ve eliminated quality controls. We introduced AI agents to support QA, while our quality specialists focus on automating functional tests and validating results, supported by specialized code analysis tools. Business users must also participate actively in functional testing and solution acceptance.

Speed cannot become an excuse to compromise quality.

Bringing Business Decision-Making Up to Speed

This change is also affecting our conversations with business leaders. We no longer discuss only priorities, budgets, and delivery dates. We now need their teams to be much more involved in defining processes, clarifying business rules, making decisions, and performing functional testing.

That remains one of our biggest challenges. Business specialists already have operational responsibilities that consume nearly all their time. How do we make them available for a development process that now moves much faster?

We are still working on the answer.

Making Organizational Decision-Making Keep Pace with AI

If I were to share a few lessons with other CIOs, I would start by developing business knowledge, communication, and collaboration skills within technical teams, while preparing the organization for a much greater demand on business specialists.

I also wouldn’t wait until everything is perfect before getting started. AI has considerably reduced the cost of refactoring and adjusting solutions, although this requires clear standards and careful management of token consumption.

What we are learning is that agentic development is not only changing how we write code. It is making us reconsider how we organize IT, develop our people, and work with the business.

For decades, companies have organized business involvement around the speed of their software development teams. AI is changing that speed, but decision-making structures have largely remained the same.

The challenge for CIOs is no longer just developing software faster, whether through internal teams or technology partners. It is working with other business leaders to build an organization where the ability to make decisions, validate solutions, and take responsibility keeps pace with our ability to deliver them.