I Stopped Writing Better Prompts. I Started Onboarding AI Like a New Teammate.

When Copilot became widely available, I did what most people did.

I experimented with prompts, watched videos about “perfect prompts,” and tried different prompt engineering techniques. Some helped, but most only slightly improved the results.

Then I noticed a pattern.

When a new Business Analyst joins my team, I don’t expect them to deliver meaningful work from a single instruction. I don’t assign a task and hope for the best.

I onboard them.

We cover the business, customers, structure, stakeholders, priorities, and success criteria. We clarify acronyms and explain why certain processes exist. By the end, they’re not just executing tasks—they’re making informed decisions because they understand the environment.

I realized I had never done this with AI.

I was expecting it to understand my world from a few sentences.

So I changed my approach.

Instead of focusing on better prompts, I started onboarding AI.

I explained my role, the types of projects I work on, how my organization is structured, the teams I interact with, the decisions I support, the terminology we use, and the metrics that matter.

Then I asked AI to turn that into a living reference document I could refine over time.

That single step improved my results more than any prompt template I had tried.

Not because I taught AI anything complex, but because I gave it the same context I would give a new teammate.

Since then, I rarely just “ask a question.”

I provide context first.

For example, I once received an email that felt politically charged rather than purely technical. My instinct was to ask AI to analyze it and help draft a response.

Instead, I added context first.

I explained that the two teams involved had conflicting incentives—one optimized for speed, the other for risk reduction. Neither reported to the other, and both believed they were acting in the business’s best interest.

Only then did I share the email.

The response changed completely.

Instead of focusing on wording, AI identified the underlying organizational tension, explained each team’s likely perspective, surfaced hidden assumptions, and suggested a response that avoided escalation.

The email didn’t change.

The context did.

That’s the same principle Business Analysts rely on every day.

We don’t jump straight to solutions because we know the first answer is rarely the full story. We ask questions, explore processes, and uncover what isn’t explicitly said.

Yet with AI, we often skip that step.

We request outputs—emails, summaries, recommendations—without sharing the context that actually shapes the answer.

Over time, I’ve stopped thinking in terms of prompts altogether.

I think in terms of onboarding.

Does AI understand the problem space?

Does it understand the organization and stakeholders?

Does it know the constraints and priorities?

If not, better prompts won’t fix the gap.

Prompt engineering has value, but I believe most people are optimizing the wrong layer. They’re refining instructions when they should be expanding context.

As Business Analysts, context is our core strength. We collect it, structure it, and use it to improve decisions.

That’s why working with AI feels natural once you treat it less like a search engine and more like a colleague still learning the business.

You don’t need a perfect prompt.

Embrace proper onboarding; this simple step has empowered me to harness AI more effectively, so I can Start with Data and End with Value.