AI Agent Design:  Build Teams, not Bots

IDC’s research indicates that by 2028, 60% of the G2000 will have adopted an agent development life cycle to scale agentic AI across their enterprises. Organizations will build custom agents (52%) and use prebuilt agents (48%). 

However, to unlock the value of agentic AI, enterprises must rethink the role of AI agents. They need to move beyond narrow, purely technical automation scenarios and instead adopt a human-centered mindset where AI agents become fully empowered, collaborative colleagues and knowledge workers, forming digital teams that work alongside and in unison with their human colleagues. 

This requires considering both the technical and non-technical aspects of AI agent design. It means choosing an agent-building platform that can not only address technical aspects of scalability, consistency, and governance but also help forge productive, integrated working relationships among agents and their human colleagues.

Good colleagues are collaborative, knowledgeable, empowered, transparent, and trustworthy. Organizations should think about these qualities when designing teams that include AI agents. 

But this consideration begs the question: What type of platform can help an organization build or customize reliable AI agents that can scale, act autonomously, and serve as valued digital colleagues to an existing team?

Seamless tools encourage seamless teams

Core principles of effective team design include reducing friction and enabling cooperation. When the tools organizations use to build agents are streamlined, unified, and intuitive, AI agents are more likely to promote cohesion. For example, a streamlined unified platform that features low-code and no-code tools for building agents can help non-technical, critical domain experts fully and directly engage in designing AI agents, and a unified tooling environment usually leads to smoother deployments.

When a platform offers a comprehensive suite of prebuilt AI agents, all designed on the same integrated platform with the same tools, it helps teams to work with various agents across workflows, without having to question which tools, frameworks, and capabilities were used. While agentic systems will be inherently heterogeneous, with mixes of models and custom and prebuilt agents from various vendors, some level of standardization across the AI stack will likely promote more cohesion, higher adoption, and better human–AI agent interactions.

Collaboration and workflow

Much of the discussion around agentic AI centers on automation, but collaboration is an equally important aspect. Organizations need to ensure that all colleagues on a team (AI and humans) can collaborate in a robust workflow. Workflows must close the human loop, ensuring that prompt engineers, human evaluators, AI engineers, and others can collaborate with each other and their AI agent colleagues. The goal should be to bring the best of automated workflow, agent autonomy, and human action together to achieve business outcomes. Bringing teams together, both agents and human workers, requires change management. In fact, 30% of organizations worldwide report that resistance to process changes required by AI is a top challenge in realizing its benefits, according to IDC's Future Enterprise Resiliency & Spending Survey, Wave 7 (September 2025).

Teams need access to corporate knowledge and internal systems to be effective. A critical feature for any agentic platform is advanced knowledge integration. Those platforms that enable rapid preparation of unstructured documents and feature integrated RAG empower AI agents to serve as fully informed knowledge workers and enterprise domain experts within a team. Additionally, platforms must enable AI agents with permissioned access to key internal systems and sources across the enterprise so they can distill and share knowledge with team members or be empowered to act autonomously on acquired knowledge. 

Choice matters: models and infrastructure

Like their human colleagues, AI agents benefit most from an optimal balance of consistency and choice. They can perform optimally when designers can choose the right model for the job and the right inference infrastructure. When matched with a unified platform, model choice allows teams to achieve the optimal mix of choice, control, and consistency. It ensures that AI agents will behave and perform consistently, even across different domains of varying complexity, while also leveraging the most capable models and optimal infrastructure. 

Trust is critical to cohesive teams

Teams break down when members cannot trust their colleagues to complete assigned tasks, comply with established policies, and follow corporate governance and security rules. And when a team member gets it wrong, all that is required is transparency and accountability. To be trusted by colleagues, AI agents must demonstrate the same level of trustworthiness. AI agents built on platforms that enable full agent audit trails, observability of agent behavior, and built-in governance tools engender trust. Given the heightened concerns around AI safety, agents built on trusted unified platforms, secure infrastructure, and that offer end-to-end management of the agent life cycle are more likely to win the trust of colleagues. The majority of organizations are focused on exactly these performance metrics, with 54% evaluating agents for reliability and 52% measuring the accuracy of task completion. The impact of AI agent performance on the satisfaction of human workers is being measured by more than 40% of organizations, according to IDC's Future Enterprise Resiliency & Spending Survey, Wave 7, (September 2025).

Build teams on unified AI platforms

For AI agents to fulfill their role as digital colleagues, enterprises need to build them using unified platforms with integrated tools and flexible model and infrastructure choice. Such platforms can provide the right balance of consistency and choice that enables AI agents to become fully integrated, trusted digital colleagues. Automation is a technical undertaking. Building teams is a profoundly human endeavor. The right mindset can truly unlock the potential of agentic colleagues. 

Message from the sponsor 

Vertesia is a SaaS platform built to accelerate how organizations develop and deploy custom generative AI (GenAI) applications, agents, and services. Designed for speed, scale, and simplicity, Vertesia combines enterprise-grade infrastructure with a low-code environment to reduce complexity, lower costs, and shorten time to value. With its API-first architecture, the platform integrates easily with existing systems, streamlines intelligent content preparation, and offers powerful tools for building and managing autonomous agents. 

Vertesia helps businesses move beyond GenAI experimentation—turning AI initiatives into strategic, scalable capabilities. 

To learn more about Vertesia's autonomous agents, please visit: vertesiahq.com.