Fixed price Agile: fix epics, flex requirements
31 August 2026
A Backlog Tells You What Might Be Built. Business Context Tells You Why.
12 September 2026Reading time: 15 min.
The BA as AI Translator: Steer It or It Will Steer You
There is a lot of discussion about whether AI will replace business analysts. In practice, I see something different. Most BAs I meet are not avoiding AI. They are experimenting with it, learning where it helps, and trying to work out how it fits into their daily work.
That raises a more interesting question than whether AI will replace the BA:
Who is steering whom?
AI can help us explore ideas, analyze information, challenge requirements, prepare workshops, identify gaps, and generate alternatives. But it can also influence how we frame a problem, which questions we ask, and which options we consider. The BA therefore has an important new responsibility: acting as a translator between business reality and AI.
Translation works in two directions
We often think about prompting as giving instructions to AI. But for a BA, the interaction is more than that. There are two directions:
Business reality → AI
The BA translates the situation into something AI can work with. That may include:
- stakeholder needs and perspectives
- business rules
- terminology
- constraints
- assumptions
- organizational context
- quality criteria
- known uncertainties
AI does not automatically know any of this.
But there is another direction that is just as important.
AI → Business reality
When AI produces an analysis, a set of requirements, a stakeholder map, acceptance criteria, or possible solutions, somebody needs to translate that output back into the real situation.
- Does this make sense here?
- Which assumptions has AI made?
- What is missing?
- What needs stakeholder validation?
- Which suggestions are useful, and which should simply be rejected?
That is not prompt engineering. It is business analysis.
AI can steer you without making the decision
There is a more subtle risk in working with AI. AI does not have to make a decision for you to influence your decision. Imagine asking:
What are the main risks of introducing this new customer portal?
AI produces eight plausible risks. The list looks professional and comprehensive, so the natural next step is to discuss those eight risks. But why those eight? Perhaps the biggest risk in your organization is something AI could not possibly know: an upcoming regulatory change, an overloaded operational team, a difficult dependency on a supplier, or a stakeholder who strongly opposes the change.
The AI-generated list has now created an anchor for the analysis. The same thing can happen when AI proposes stakeholders, requirements, process steps, solution options, or acceptance criteria. Because the output is fluent and structured, it can easily become the starting frame for our thinking. And once the frame exists, we tend to work within it. That is what I mean by letting AI steer.
The BA needs to control the analytical loop
Using AI well is therefore not simply: Prompt → Output
It is a loop: Understand → Ask → Examine → Challenge → Validate → Refine
Before using AI, the BA considers what is already known and what is uncertain. Then AI can help explore the problem: suggest questions, identify possible gaps, generate alternatives, or challenge an existing analysis. But its output should become input to further analysis, not the conclusion.
A generated stakeholder list becomes something to challenge. A proposed requirement becomes something to examine. A set of risks becomes a hypothesis to validate. Suggested acceptance criteria become a starting point for discussion with the right people. The BA remains responsible for deciding what deserves attention and what happens next.
This is where traditional BA skills become even more important
Interestingly, working effectively with AI does not make traditional business analysis skills less relevant. It makes many of them more important. Just to give you some examples:
- Elicitation helps us discover information AI cannot know.
- Critical thinking helps us challenge plausible-looking output.
- Stakeholder analysis tells us whose perspective is missing.
- Modeling helps us understand relationships and consequences rather than relying only on generated text.
- Domain knowledge helps us recognize when something does not fit the real situation.
- Validation helps us distinguish a convincing answer from a correct and useful one.
And perhaps most importantly, asking good questions determines where the analysis goes next. These are not skills AI removes from the BA role. They are the skills that allow the BA to use AI without becoming overly dependent on it.
Start by giving AI a role, not the steering wheel
A useful way to introduce AI into your work is to give it a specific analytical role. Instead of:
Analyze these requirements.
Try: Review these requirements and identify possible ambiguities. Explain why each one could be interpreted differently. Do not rewrite them yet.
Or: Here is my stakeholder analysis. Challenge it. Which perspectives might I have overlooked, and what assumptions are you making when suggesting them?
Or: Here are three solution options identified by the team. Identify questions we should investigate before comparing them. Do not recommend an option.
Notice the difference. AI is supporting a step in the analysis rather than performing “the analysis” on your behalf. You determine the purpose. You determine the boundaries. And you decide what to do with the result.
The BA role is evolving, not disappearing
AI can generate content faster than any BA. That is not where we should try to compete. Our value lies in understanding the business situation well enough to direct that capability: knowing what information matters, what question needs to be asked, what assumptions need challenging, whose perspective is missing, and when an apparently good answer does not fit reality.
In that sense, the BA increasingly becomes a translator between business reality and AI. But translation alone is not enough.
We also need to remain in control of the conversation.
Because if we do not consciously steer AI, its answers can very easily start steering our analysis.
