We are starting a major AI implementation and I’m trying to define the BA's responsibilities. How does a Business Analyst ensure that AI requirements are feasible and aligned with business goals? I’m specifically interested in how we manage data quality requirements and stakeholder expectations when the AI models are "black boxes" that don't always provide predictable outputs.
3 answers
In the age of AI, the BA role shifts from gathering static requirements to managing dynamic data expectations. You must act as the bridge between the data scientists and the business units. A key responsibility is defining "Success Metrics" for the AI—it’s not just about accuracy, but about business impact. At iCertGlobal, we recommend BAs focus on "Data Lineage" and "Explainability." You need to ensure the business stakeholders understand why a model makes a decision, even if they don't understand the underlying math. This requires you to facilitate workshops that define clear ethical boundaries and data governance policies before the first line of code is ever written.
Are you using a specific framework like CRISP-DM to guide your analysis, or are you adapting your existing Agile user story format for these AI features?
You should look into the IIBA’s Guide to Business Data Analytics. It provides a great framework for BAs moving into these more technical, data-heavy roles.
I agree with Steven. The IIBA certification path for Data Analytics is becoming essential for BAs who want to remain relevant in the current job market.
Mark, we are currently trying to adapt our Agile user stories, but it’s proving difficult. We’ve started adding a "Data Requirements" section to every story to specify what features the model needs for training. This ensures that the dev team doesn't start building until we’ve confirmed the data is actually available and clean. It’s a bit of a shift for the team, but it has already saved us from two "dead-end" sprints where the data we needed simply didn't exist in our production environment.