You've been given a sports car. You're driving it in first gear.
In the series opener of this series, I described AI as a force multiplier. The question isn't whether these tools can create value. It's how much value you'll be able to unlock from them. One of the most important ways to increase that multiplier is through better specifications.
A friend recently asked for help building a document ingestion pipeline for his website. He was already using AI across his business, but he was struggling to make progress with this particular challenge.
The breakthrough wasn't a different model. It was a different way of working.
After describing what he wanted to achieve, we built the solution in a matter of hours using coding agents. Today, those same approaches are helping him automate processes across his business and accelerate how work gets done.
The difference came down to understanding the tools around the model.
Many organisations have access to incredibly powerful AI capabilities today. The challenge isn't capability. It's knowing how to unlock it. That's where the sports car analogy comes from. Too many teams are sitting behind the wheel of a high-performance machine and only using a fraction of what it can do.
Beyond the chat window
Many people experience AI through a chat interface. They ask a question, receive a response, and move on to the next task.
Agentic development takes things further.
Instead of working only through conversation, agents can interact with files, call tools, execute commands, inspect results, and continue working towards an objective. The model provides the intelligence. The surrounding tooling provides the environment where work can happen.
In my experience, the biggest barrier is not technical capability. It's learning how to work with these tools effectively.
In fact, organisations rarely hit the limits of the model first. They hit the limits of the instructions they're giving it.
Why specifications matter
As projects become more complex, short prompts often leave too much open to interpretation.
A specification provides the context, requirements, assumptions, and desired outcomes needed to guide an agent towards the right result. The more clearly you describe what success looks like, the more likely you are to get a useful outcome.
Every decision that isn't documented becomes a decision the model has to make on your behalf.
For simple tasks, that may not matter. For business processes, applications, and enterprise workflows, it often does.
One principle I regularly share with teams is simple:
If a requirement isn't defined, the model will make an assumption.
That assumption may be reasonable. It may not. The more clarity you provide, the more predictable the outcome becomes.
Defining what good looks like
The most effective specifications do more than describe the solution. They define how success will be measured.
For example, if a requirement is to create a responsive application, what does responsive actually mean? How is it tested? How do you know when it's complete?
Clear acceptance criteria help both people and agents understand when an objective has been achieved. They also create a foundation for validation and governance as solutions move into production.
Good specifications don't restrict creativity. They create alignment.
What this means for Workday teams
This shift is particularly relevant for Workday developers, consultants, and HRIS teams. With Workday's recently announced Developer Agent and the continued evolution of agentic tooling across the ecosystem, organisations have an opportunity to accelerate development, simplify complex workflows, and unlock new ways of working.
The skills that matter are changing too. Success increasingly depends on defining requirements clearly, providing the right context, and understanding how agents can be guided towards the desired outcome.
This is why I believe one of the most valuable skills of the next decade will be specification writing. Whether you're a developer, consultant, product owner, project manager or HR leader, the ability to clearly define intent will become increasingly important.
The future isn't everyone becoming a developer. It's everyone becoming a better specification writer.
Start with a simple experiment
If you've only ever interacted with AI through one line prompts, try approaching your next task differently.
Take a challenge you would normally describe in a sentence. Then, write a detailed version that captures the requirements, constraints, objectives, assumptions, and expected outcome.
Run both. Compare the outputs.
In my experience, the difference between the two results is often far greater than people expect. The organisations that gain the most value from AI won't necessarily have access to better models. They'll have people who know how to communicate intent more clearly, capture requirements more effectively and guide agents towards better outcomes.
The tools are already here. The opportunity now is learning how to use them effectively.
And that's the journey we'll continue exploring throughout this series.
The 8-word diagram is an unmade decision wearing a flowchart. One request, three levels of detail: 2, then 4, then 11 of 11 decisions on the page. See the three-gear report →
Better specifications lead to better outcomes.
The same principle applies to your AI strategy. Our AI Navigator sessions help Workday customers cut through the noise, prioritise the right opportunities and build a practical path towards AI-enabled transformation. Book today to start shaping your AI roadmap with Kainos experts.