The AI Readiness Audit: Five Questions to Answer Before You Build

Every week, another executive team decides it's time to "do AI." Budget gets allocated, a vendor gets selected, a pilot gets launched — and eighteen months later there's a demo that impressed everyone once and a production system that doesn't exist.

In our experience, the difference between companies that ship AI and companies that stall is rarely the technology. It's the groundwork. Before you build anything, you should be able to answer five questions honestly.

1. What decision or workflow are we actually improving?

"We want a chatbot" is not a use case. "Our underwriters spend four hours per application re-keying data from PDFs" is. The best AI initiatives start from a measurable pain point — a decision made too slowly, a workflow with too many manual touches, knowledge locked in documents nobody can search.

If you can't name the metric you expect to move, you're not ready to pick a tool, let alone build one.

2. Where does the data live, and who can touch it?

AI systems are downstream of your data reality. If the information a model needs is scattered across a legacy ERP, three spreadsheets, and someone's inbox, that's not a blocker — but it is the actual first project. Data access, quality, and ownership questions resolved early save months later.

3. Who owns this after launch?

AI systems are not fire-and-forget. Models drift, prompts need tuning, edge cases accumulate. A solution without a clear internal owner — someone accountable for its performance after the consultants leave — becomes shelfware within two quarters.

4. What does 'safe enough' mean for us?

The right risk posture for an internal research assistant is very different from one for a customer-facing system in a regulated industry. Defining your guardrails up front — what data can be used, what actions can be automated, where a human must stay in the loop — turns governance from a late-stage blocker into a design input.

5. Are we buying, building, or assembling?

Some problems are solved by configuring an off-the-shelf product. Others demand custom engineering on your proprietary data and workflows — that's where durable competitive advantage lives. Most real answers are an assembly: foundation models and cloud platforms underneath, custom integration and domain logic on top.

Answer these five honestly and the path forward usually becomes obvious. If you'd like a structured version of this exercise, that's exactly what our strategy sessions are for.