No APIs, No AI: Organizing Software Engineering for Today's AI Reality

**Karen Stokes-Lockhart** (0:04)
Welcome to Gartner ThinkCast, I'm Karen Stokes-Lockhart. Today, we're focusing on software engineering and how it must evolve in a world driven by generative AI. You'll hear from a few of our software engineering experts, Manjunath Bhat, Akis Sklavounakis and Shameen Pillai. They'll explore the team structures that help scale GENAI to repeatable delivery, the platform patterns that cut cognitive load and help drive value, and the API or Application Programming Interface strategies that AI implementation can't function without. First up, here's Gartner distinguished VP analyst, Manjunath Bhat, to explain the difficulties around scaling GENAI and how adopting the right team topologies can help.

**Manjunath Bhat** (0:51)
Scaling generative AI application delivery is incredibly hard based on what Gartner surveys have revealed. Identifying the right use cases, honing in on the right use cases makes it easier to demonstrate business value. Needing security and governance requirements at scale is also very difficult when you have multiple applications and use cases. You're looking at choosing the appropriate design patterns in architecture, whether you should use open versus closed models, or you should use self-hosted versus third-party APIs directly.
The perils of an uncoordinated approach to implementing generative AI, I would count as three things. First is low phenops awareness that is related to token costs, GPU costs, influencing costs. Second is security and compliance violations. This could lead to potential reputational damage. Third is a lot of inconsistent practices leading to duplicated effort and poor quality, which ultimately hinders your ability to scale innovation across the organization. Gartner recommends adopting a team topologies approach as an organizing principle to structure your engineering themes. The team topologies approach recommends four different kinds of themes. The first is product themes, which in the team topologies parlance is called stream aligned themes. These are your regular application themes building customer focused functionality. The second is enabling themes. The idea of an enabling theme is to bring together specialized expertise. Think of these as internal consultancies. In many ways, enabling themes are the teams that help drive new capabilities within the organization and Genetive AI happens to be one of the new capabilities that you will be rolling out within the organization. Additionally, what you could also think of is creating subsystem teams. These in the team topologies construct are called complicated subsystem teams because they reduce complications and they also reduce cognitive overload for your teams. The complication increases exponentially when you embed Genetive AI applications, move towards extending, and finally, when you're trying to build your own models. Once you have identified enabling teams, product teams, and the subsystem teams, lastly, you have platform teams. Platform teams help ensure consistency and also provide a curated experience for the rest of your teams. So put all of these together and what you find is a very healthy recipe for scaling Genetive AI capabilities across all applications and use cases. A great example of a leading organization that has implemented this construct is Verizon. They have two kinds of teams. One which they call Genetive AI Center of Excellence, which maps to enabling teams within the team topologies construct. And within these enabling teams, Verizon has assembled experts from not just IT, but also risks, legal and finance teams. So all of this makes it easier to disseminate expertise, spread knowledge and spread best practices. They have a very interesting pattern when it comes to platform teams. So their platform teams provide a curated experience. But think of this as when you have to implement a design pattern, like generating new content, that design pattern actually triggers a workflow that goes all the way from choosing the right tools, providing reusable components plus also templatized workflow. So all of this together means that your development teams don't have to think for themselves. So everything is provided as a recipe. To embark on this journey of scaling generative AI applications, you should take a three-pronged approach. First, inventorying your current applications and use cases. Second, identifying scaling issues and competency gaps that may exist. And lastly, institutionalizing best practices across the organization so your applications don't just end up as prototypes, but actually go to production. By implementing these strategies, you can ensure that both your internal employees are productive and creative, plus your external customers get the most out of your generative AI investments.

**Karen Stokes-Lockhart** (5:21)
AI is everywhere. But what does it mean for your business? Gartner is the world authority on AI, with more than 200,000 client conversations and more than 6,000 written insights on AI in 2025 alone. Leaders across the C-suite, just like you, are partnering with Gartner to turn AI ambition into impact. Go to gartner.com/ai to learn more.
Next, Gartner Senior Director Analyst Akis Sklavounakis walks through how great people aren't enough if the surrounding work environment slows them down. And why platform engineering clears the path for high-velocity development and delivery. Here's Akis.