An AI agent is a software entity that can perceive information, reason about it, and take actions, often independently, to achieve a defined goal.
Unlike traditional automation scripts, which follow fixed rules and sequences, AI agents can adapt to dynamic situations, use multiple tools, and handle non‑deterministic processes where every run might look different.
They might:
- Pull in data from multiple systems
- Call APIs
- Trigger workflows
- Decide how to complete a task without strict pre-programming
This flexibility makes them powerful, but also requires careful design.
Many SaaS vendors now ship proprietary agents designed for their own ecosystems. But there’s also a wide and growing landscape for organisations to build their own, from writing custom code to leveraging orchestration platforms or automation frameworks with built‑in agent creation capabilities.
It’s well understood that data access and integration are integral to making an agent useful. But there’s a leap from simply exchanging data in a predetermined way to enabling an AI agent to operate autonomously in a non-deterministic fashion.
Why we need the AI Agent Designer
I believe there is a growing need for a new type of role in the design and implementation of agentic orchestration, someone who acts as an architect and a designer. This role requires a solid understanding of the technical building blocks of an AI agent, even if only at a high level, combined with the ability to grasp business processes and requirements and translate them for different audiences.
In my view, an Agent Designer doesn’t need to be an expert in the intricate details of machine learning, vector databases, or large language models, but they do need to understand what to expect from these technologies, as well as how and when to use them. Based on this perspective, I have outlined what I believe to be the key responsibilities and tasks for an Agent Designer.
What an AI Designer has to consider
As an Agent Designer, the responsibility is to define and outline an AI agent that can automate part, or several parts, of a business process. At a high level, this involves determining the agent’s purpose and ultimate goal, identifying the tools and data it will require to achieve that goal, deciding the appropriate level of autonomy it should have, and establishing how users and systems will interact with it. By combining an understanding of business requirements with the technical possibilities of AI agents, the Agent Designer ensures that the solution is purposeful and capable of delivering measurable value.
I believe one of the first, and perhaps most important, tasks for an Agent Designer is to determine when it is appropriate to use an AI Agent and when it is not. There are many situations where it might be tempting to implement an agent because it appears to add value to a process or step. However, a closer examination often reveals that the actual benefit is minimal or even non‑existent. The ability to critically assess whether building an AI agent is worth the time, effort, and resources is, therefore, one of the Agent Designer’s most essential skills.
Skills and knowledge areas
While the depth of expertise varies by use case, I believe an effective Agent Designer needs to know:
- AI literacy: An Agent Designer should have a solid conceptual understanding of large language models (LLMs), AI reasoning, retrieval‑augmented generation (RAG) solutions, and vector databases. While deep technical expertise in these components isn’t required, it is essential to understand what each is used for, how it works at a high level, and what to consider when incorporating it into a solution. This knowledge allows the Agent Designer to select the right AI capabilities for the intended business outcome.
- Prompt and context engineering: Designing effective AI Agents requires the ability to craft and refine prompts, system instructions, and contextual information that guide the AI’s behaviour. This includes structuring role definitions, conversation flows, and decision boundaries so that the agent produces accurate, relevant, and consistent results. It also involves testing and adapting prompts for different use cases, handling exceptions, and ensuring they remain aligned with business goals and compliance requirements.
- Integrations: The tools that an AI Agent uses often take the form of integrations, enabling it to create, update, delete, or retrieve data from connected systems, applications, and services. An Agent Designer benefits greatly from understanding system integration concepts and API architecture at a high level. With this knowledge, they can design an agent’s toolset effectively, reusing existing integrations where possible and defining new ones when required to meet the agent’s objectives.
- Process design: AI Agents are particularly valuable in workflows that involve non‑deterministic steps – situations where decisions and outcomes can change from case to case. An Agent Designer needs to be able to analyse and model processes, identify where AI‑driven automation could make the most impact, and ensure those improvements are aligned with broader business objectives. A strong foundation in process design helps ensure the agent’s role fits seamlessly into the overall workflow.
- Governance and compliance: An Agent Designer should understand data security, privacy regulations like GDPR, and the ethical principles that guide responsible AI use. This means ensuring the agent complies with legal requirements and upholds fairness, transparency, and accountability to protect the organisation and its users.
Why the Agent Designer matters
As organisations move from trying AI to operationalising AI, I think there is a growing need to ensure agents:
- Serve a clearly defined business purpose
- Are equipped with the right tools and integrations
- Can be measured, optimised, and improved over time
- Operate securely, ethically, and in compliance with regulations