From Messy Data to Meaningful AI: The Two-Step That Starts the Journey
“My leadership says we must start using AI, but our data is such a mess I don’t even know where to start.”
The friction is real, and this comes up in almost every conversation I have with CS leaders.
True. Your data is messy. Executives want world-class AI, yesterday. To CS leaders, this feels like being stuck between a rock and a hard place. After all, nobody wants to be the person that just released a bot that scaled bad customer data across millions of customer profiles (please, please don’t be that person).
And here’s the crisp, actionable truth: AI isn’t waiting for perfection. It thrives on intentional progress.
That’s why I call it the Two-Step.
I also call it the Two-Step because I lived in Louisiana for 14 years, and really wish I could dance the real two-step as well as we can do this one together, but I digress. \
Let’s dive in.
Step One: Map the AI Opportunity as It Exists Today
You don’t need a pristine data warehouse to start seeing value. Just look at your friction points: meeting prep chaos, renewal lag, disjointed handoffs.
Zero in on the friction until you have a narrowly defined situation with concrete, accurate information. Not the entire customer record, or every engagement you’ve ever had with this book of business. Look at narrow instances where you have concrete information.
Real examples of narrowly defined opportunities:
Keyword Risk Alerts: AI scans only the latest customer email or call transcript for phrases like “budget cut,” “renewal,” or “looking at alternatives,” and pings the CSM with an alert.
QBR Slide Drafting: AI auto-generates just three slides (“Usage Highlights, Top Issues, Next Steps”) directly from product usage data—no narrative or full deck creation.
Support Case Digest: AI produces a one-paragraph summary of the most recent closed support case, highlighting “Problem, Resolution, and Next Risk” for the CSM.
Sales Engagement Nudge: AI notifies the CSM only when Sales reaches out to their customer, including a two-sentence summary of the outreach and a suggested next action.
Renewal Prep Coach: AI guides the CSM with a simple renewal prep checklist (questions to ask, objections to expect, messaging tips), without requiring any customer or product data.
That’s Step One: spot where AI can help now.
Step Two: Activate a Narrowly Scoped Pilot
Once you’ve mapped a friction point, the next move is to design a pilot that proves value fast, without relying on perfect data.
Take the Renewal Prep Coach example.
This company has mostly junior CSMs. They don’t need a predictive churn model, they need practice. Instead of waiting on customer data that’s too messy to use, the AI agent creates a custom roleplay for renewal conversations based on a few form-field inputs such as:
Customer size (Enterprise, Mid-Market, SMB)
Renewal amount (small, medium, large)
Primary objection (budget, product fit, competitor, internal change)
The AI then generates a short simulation: a customer persona, likely objections, and suggested talking points. The CSM can “practice” the conversation in a safe space before the real call.
This pilot doesn’t touch the messy CRM. It doesn’t pretend to solve all of renewals. It targets one concrete friction point (junior CSMs lacking confidence in renewals) and provides immediate, measurable value.
That’s the essence of Step Two: move from spotting opportunity to running an intentional pilot in a narrowly defined slice of the workflow.
Managing Up While Managing Reality
This approach gives you narrative not paralysis:
Show leadership real AI impact—today.
Buy time to clean data and refine processes behind the scenes.
Shift from “waiting for perfect” to “launching intent-led progress.”
The Real Opportunity
No CS team has perfect data or processes. That’s not the blocker, it’s the context.
The opportunity? Start smart:
Map what’s real.
Activate what works.
Ideal isn’t necessary. Intention is.
