Questions Before Platforms: Building the Data Foundation AI Actually Needs | Episode #0036 by Start Small, Think Big

Watch the full episode on YouTube, listen on Spotify, or read and follow the Start Small, Think Big newsletter.

Summary

Sid Atkinson, co-founder and CEO of Applied Curiosity and co-host of The Data Culture Podcast, joins Brad Groux to explain why AI readiness starts with better questions, consistent data practices, and a clear business purpose—not another platform.

Brad and Sid discuss lightweight governance for smaller organizations, how to preserve institutional knowledge with context and effective dates, why leaders need to close the gap between the documented process and the work people actually do, and where human accountability belongs in agentic systems.

Slow is smooth. Smooth is fast.

Start With Questions, Not Platforms

Every organization has more questions than time, money, and attention. A useful data strategy does not pretend to answer everything at once. It creates a disciplined way to decide which business questions matter now, what evidence would change a decision, and what capability must exist to answer them reliably.

Smaller companies do not need enterprise bureaucracy. They do need repeatable accounting, access controls, naming conventions, retention decisions, ownership, and enough rigor to keep a manageable mistake from becoming an existential one.

Capture the Why Before It Walks Out the Door

An SOP can document steps. It rarely captures all the judgment behind them: why a technician listens for a specific sound, why a project manager calls one customer before changing a schedule, or why an estimator distrusts a measurement in one situation.

Record the work. Ask experienced people to explain what they notice and why they make each decision. Then give that knowledge dates, versions, and owners. An AI system will not automatically know that a 20-year-old procedure was replaced last quarter.

Close the Perception Gap

Leaders make decisions using their understanding of how the company operates. The people closest to the work know how it operates today. Those are not always the same thing.

Ask the people doing the work what problem they are solving, how the work supports the mission, and where the written process diverges from reality. Listening is not a soft alternative to operational rigor. It is how leaders find the ground truth required to make the system better.

The Model Is Not the Mission

Frameworks such as CRISP-DM, DAMA-DMBOK, and CMMI can reveal capabilities an organization needs. They are not the business outcome.

The model is not the purpose. The model is to help you have a capability so you can do the things you want to do.

Backups, logging, access control, definitions, and ownership create the fitness to pursue an outcome safely. Use the model to identify the useful capabilities, then keep the work pointed at customers and purpose.

Automate Discovery, Not the Relationship

One practical automation almost any organization can attempt is finding commitments buried across meetings, messages, notes, and email. Use AI to surface what you promised, who needs a response, and what deadline you accepted. Then let the person follow through.

Automate the discovery and the research, but don’t automate your interactions.

Know Your Nouns

If teams do not share definitions for customer, market, segment, channel, and product, dashboards conflict, automations drift, and AI scales the ambiguity. Those definitions are not an IT side project. They are leadership’s job.

Start with the questions. Define the business language. Then choose the technology.

Key Takeaways

Start with the business questions that matter, then identify the data and capabilities required to answer them.
Small organizations need lightweight rigor, not enterprise bureaucracy.
Capture the reasoning behind experienced employees’ decisions, not only the written steps.
Tag, version, or retire outdated knowledge so AI does not treat every document as current truth.
Use governance frameworks to identify capabilities; do not mistake the framework for the mission.
Automate the discovery of commitments while keeping human relationships human.
Define the shared nouns of the business before attempting advanced analytics or AI.

Resources Discussed, in Episode Order

Listen and Follow

Find the podcast, video, newsletter, and every listening platform at linktr.ee/digitalmeld.