Anything You Tell The Agent Twice
In previous articles, I've explored why successful agentic development depends on clear specifications and effective communication. The quality of an agent's output is directly influenced by the quality of the instructions and context it receives.
As organisations begin to adopt AI-powered ways of working, a new question emerges: how do we stop repeating ourselves?
Many people start their journey with AI agents by repeatedly providing the same instructions. They explain where information can be found, define preferred ways of working, or remind the agent how particular tasks should be approached. This works, but only for a while.
Eventually, patterns emerge. The same guidance is provided repeatedly. The same documents are referenced. The same processes are followed.
That repetition is often a sign that knowledge should be captured rather than continually re-explained.
The difference between knowledge and skills
One of the most important concepts in agentic development is understanding the difference between what an agent needs to know and what it needs to do.
Some information remains broadly relevant regardless of the task being performed. This could include project structures, key documentation locations, governance requirements, naming conventions, or organisational context.
Most modern agentic platforms provide mechanisms for storing and automatically loading this type of information. Whether it is a CLAUDE.md, copilot-instructions.md, AGENTS.md, or another form of persistent instruction file, the principle remains the same.
Anything you find yourself telling the agent twice is a candidate for onboarding knowledge.
This information forms the foundation from which agents operate, much like the onboarding materials provided to new employees joining a project or team.
However, knowledge alone is not enough. Agents also need repeatable behaviours.
Turning repetition into capability
Some tasks follow a predictable pattern.
You ask the same questions each week. You gather information from the same systems. You present findings in the same format. Over time, these activities become less about knowledge and more about process.
This is where skills become important.
A skill captures a repeatable way of working. It provides the agent with guidance on how to approach a specific task, where to retrieve information, and how the output should be structured.
The concept should feel familiar.
When onboarding a colleague, you explain both the context of the project and the steps required to complete recurring activities. Skills serve a similar purpose for agents.
Rather than repeatedly instructing an agent how to perform the same task, the process can be captured once and reused whenever it is needed.
From SharePoint site to project expert
This became particularly relevant during a project involving a SharePoint environment that had evolved over many years.
Like many long-running projects, information existed across multiple locations. Documentation, status reports, financial data, source code, and historical decisions were all available, but finding the right information often depended on knowing where to look.
The challenge was not that the information was unavailable. The challenge was that the knowledge required to navigate it effectively existed largely in people's heads.
Rather than restructuring the site, I focused on helping the agent understand it.
By combining onboarding knowledge with task-specific skills, I was able to guide the agent towards key project artefacts and define common activities such as reviewing status reports, analysing financial performance, and answering project-related questions.
The result was a significant shift in how information was accessed.
Instead of searching through folders and documents, I could focus on the questions I was trying to answer.
The agent became more than a retrieval tool. It became a guide capable of navigating the project landscape on my behalf.
Why this matters for Workday customers
As organisations explore agentic capabilities within the Workday ecosystem, the conversation often focuses on models, platforms, and emerging technologies.
These topics are important, but they are only part of the story.
The greatest opportunity may lie elsewhere.
The organisations that gain the most value from AI agents will be those that successfully capture institutional knowledge and transform repeatable processes into reusable capabilities.
- Creating reusable skills around testing, integrations and configuration activities.
- Capturing governance requirements and development standards.
- Defining repeatable operational and project-management processes.
- Improving consistency across teams and delivery practices.
- Enabling new team members and agents to become productive more quickly.
The objective is not simply to automate work. It is to make knowledge more accessible and expertise more scalable.
Onboard the agent
When organisations think about AI adoption, they often focus on choosing the right tool.
In my experience, the bigger opportunity is learning how to onboard the agent effectively.
The same principles that help new employees succeed also help agents succeed. Provide context. Share expectations. Capture repeatable processes. Create clear guidance.
Most importantly, pay attention to repetition.
Every time you find yourself giving the same instruction, answering the same question, or following the same process, there is an opportunity to convert that knowledge into something reusable.
Because successful agentic development is not just about building better agents. It's about creating better ways for those agents to learn, operate and deliver value.
In the next article, we'll explore what happens when agents move beyond static knowledge and gain access to live systems, real-time information and enterprise tools through MCP.
See you in third gear.
Want to explore what this could mean for your organisation? Book your AI Navigator session with Kainos to identify where agents, automation and smarter ways of working can create practical value across your Workday landscape.