When someone says, “we should use AI,” how do you unpack what’s really being asked?

I agree with you, many times people are just pressured to use AI, but it is necessary to get the requirements clear first.

In response to How to Unpack when someone says, “We Should Use AI”

Start with determining the Purpose. Provide a clear, non‑technical framework to evaluate AI proposals, align stakeholders, and reduce delivery, value, and governance risk.

1. Problem Clarity (Start Here)

☐ What **specific problem** are we trying to solve?

☐ Is this driven by **missed expectations**, **broken or stalled processes**, or **risky decisions**?

☐ What outcome is currently not achievable with existing approaches?

2. Is This Actually an AI Use Case?

☐ Does the work require **classification, prediction, interpretation, or intelligent routing**?

☐ Are we asking the system to **make sense of data**, not just move it?

☐ Would rules‑based automation be insufficient?

3. AI Pattern & Mental Model

☐ Which **AI pattern** does this resemble (e.g., classification, forecasting, personalization)?

☐ What are the expected **inputs and outputs**?

☐ What **mental model** will the team use to reason about this work without technical depth?

4. Data Readiness

☐ Do we have the **right data**, at the right quality, to support this?

☐ Is data preparation, monitoring, and ongoing evaluation understood?

☐ Are there hidden **human‑in‑the‑loop** dependencies?

5. Delivery & Process Impact

☐ How will AI **change the workflow**, not just the toolset?

☐ Will it reduce manual routing, interpretation, or bottlenecks—or introduce new ones?

☐ Are we improving delivery, or layering AI onto a broken process?

6. Value Definition

☐ What **measurable value** is expected (speed, accuracy, personalization, responsiveness)?

☐ Are expectations realistic given the problem and data?

☐ How will value be assessed over time?

7. Risk, Governance, and Oversight

☐ What **ethical, privacy, or compliance risks** are introduced?

☐ Where are **governance guardrails** and review cycles required?

☐ Are we avoiding vendor‑led decisions that don’t match our workflows?

8. Stakeholder Alignment

☐ Do business and delivery teams share a **common language** for what “AI” means here?

☐ Are assumptions documented and understood early?

☐ Is ambiguity reduced before committing to delivery?

The Executive Framing Statement would be:

"Before approving an AI initiative, we confirm the problem, validate that this is truly an AI use case, identify the AI pattern, assess data readiness, define value, and establish governance to manage risk responsibly.”

The term “AI” has become a buzzword that often masks very different business needs. To clarify the intent, I usually focus on these three pillars:

1. The indicators for distinguishing the work
Generative AI (Content/Creativity): If the person is looking to draft quick reports or generate images, they are seeking efficiency in creation.
Predictive AI (Analysis/Data): If they ask, “When will the project be finished?”, they’re looking to reduce uncertainty based on historical data.
Process Automation: Sometimes they say AI, but what they really need is a workflow that eliminates repetitive manual tasks.

2. What Can Go Wrong?
When everything gets mixed up, the biggest risk is a misalignment of expectations. If the client expects a “magic solution” that makes complex ethical decisions and we deliver a basic chatbot, the project is doomed to fail from the start. Additionally, data cleaning costs are often underestimated if the type of AI to be used isn’t defined.

3. My personal experience
I’ve been in meetings where an executive asked for AI to “be more innovative,” while the technical team understood they had to implement complex machine learning models. What helped me was asking, “What specific problem are we trying to solve that’s currently taking up the most of our time?” That grounds the conversation immediately.

When someone says “we should use AI,” they are usually expressing a desire for improvement—but not a clear requirement for sure.
Job on hand is to turn that statement into specific, measurable, feasible objectives by exploring:
• The problem
• The goal
• The data
• The users
• The workflow
• The risks
• The expected outcome / value