The new AI builder: Skills for the agentic AI era

Coding remains at the heart of AI development. Agentic AI, however, is changing what it means to be an AI builder.

An AI agent can work through a task, decide what needs to happen next, use different tools and act towards a defined outcome. Building such an agent means thinking about much more than the underlying solution. Builders need to understand the problem, the environment in which the system will operate and the decisions it may need to make along the way.

Start with the problem, not the technology

Building an effective AI agent starts with a clear understanding of the problem it is intended to solve. Understanding how a business process works, where time is lost, and which decisions require intervention can be just as important as choosing the right technical approach.


Product thinking comes into play here. A builder needs to ask whether an agent is actually improving a workflow and making someone's work easier. An impressive technical solution has limited value if it does not fit the way people work.

Systems thinking matters too. Enterprise agents rarely operate alone. They need to connect with databases, applications, Application Programming Interfaces (APIs), documents and other systems. Designing the architecture around these interactions becomes an important part of building something that can function reliably.

Autonomy needs direction

Agentic AI brings another question into focus: how much should an AI system be allowed to do on its own?

An agent managing a routine workflow may be able to act independently. A decision involving sensitive information or significant business consequences may require human review. Building for these situations requires judgment, effective human-agent interaction design and clear oversight mechanisms. This includes determining when an agent can act independently, when it should seek guidance and how its actions can be reviewed.

The ability to assess agentic systems is equally important. An agent needs to perform consistently across different scenarios, not simply produce an impressive result in a controlled sandbox environment. Evaluation therefore becomes part of the building process, with teams having to consider accuracy, consistency and the outcomes their system is designed to deliver.

The skills behind the next generation of AI

These demands are expanding the skill set required of AI professionals. Technical expertise remains essential, but it increasingly needs to be complemented by product and systems thinking, an understanding of business processes and user needs, and the ability to translate AI capabilities into practical business outcomes.

The ET AI Hackathon: Agentic Edition, presented by Accenture, is open to AI and technology professionals with three or more years of experience, including AI/ML engineers, software and data engineers, architects, and consultants. Participants will work on enterprise-focused challenges and build agentic solutions that can reason, plan and act while accounting for the role of human direction.

The competition also offers a ₹2 lakh prize pool, with the grand champion receiving ₹1 lakh, the first runner-up receiving ₹60,000 and the second runner-up ₹40,000. Participants will showcase their solutions to Accenture leaders and industry experts. Select eligible participants will also be invited to interview for career opportunities with Accenture.

For AI professionals, the challenge is no longer limited to building a solution that works. The next step is building systems that understand a task, work through it effectively and know when a person needs to step in.

The ET AI Hackathon: Agentic Edition offers a platform to put those skills to the test.

Registration closes by September 20. Register Now!