
Episode 50: Thinking of AI? Think Data First.
We sat down with an SVP of Revenue Strategy who shares how to build a clean data foundation, define your ICP, and scale AI the right way.
In this episode of the RevOpsAF podcast, co-host Camela Thompson is joined by Sonia Chawla, SVP of Global Revenue Strategy and Operations at Climate Impact partners and a veteran of the RevOps space, with 15 years of experience helping companies - from $4M startups to $2B enterprises - build and scale revenue operations. Sonia brings a sharp analytical lens shaped by her background in engineering, FP&A, and systems thinking. Together, they unpack one of the hottest (and most misunderstood) topics in the space right now: AI. While generative AI promises a new era of automation and efficiency, Sonia makes a compelling case that if your data foundation is broken, no amount of AI can fix it.
This conversation is a must-listen for any RevOps professional wondering how to responsibly introduce AI into their GTM stack - and where to begin when the hype feels overwhelming.
Sonia opens the episode with a critical reminder: many RevOps teams are pressured to automate before they understand the process itself. Her litmus test is straightforward but powerful - “If you can’t sketch the workflow on paper, don’t automate it.”
Too often, leaders see automation or AI as a silver bullet to solve people or process issues. In reality, this just leads to deeper inefficiencies. Automation at scale amplifies broken logic. It doesn't fix it. Sonia emphasizes the importance of first modeling your workflows manually, gaining stakeholder buy-in, and stress-testing the edge cases before turning to tech solutions.
“If you’re not automating the right thing, you’re just creating work that will be undone later.” – Sonia Chawla
👉 Want to read more about laying the groundwork before scale? Check out Beyond KPIs: The Art of Storytelling With Data for how narrative can guide insights.
Many companies are jumping into AI tools without addressing the fundamental question: What business problem are we trying to solve? Sonia and Camela both point out that teams often reach for AI to cover up weak messaging or unproven GTM strategies - particularly in cold outbound.
For example, if your cold email campaigns aren’t resonating with prospects, throwing generative AI at the problem only allows you to send bad messaging at scale. The root cause - unclear ICPs, poor segmentation, or weak value props—remains unaddressed.
“If your messaging isn’t resonating already, why are you trying to scale it with AI?” – Camela Thompson
In early-stage startups that haven’t found product-market fit, this problem is even more pronounced. Sonia argues that companies at this stage should avoid AI entirely and instead lean into highly manual, customer-driven discovery work: conducting interviews, documenting use cases, and deeply understanding buyer psychology.
The episode takes a practical turn as Sonia outlines how to approach Ideal Customer Profile (ICP) work across different maturity stages.
For companies with <100 customers, Sonia recommends building an internal data enrichment process by hand. Start with the basics:
This qualitative + quantitative approach yields far richer insight than AI summarization or pre-trained models. Once the ICP is validated manually, it becomes a powerful base for scalable segmentation and predictive modeling.
At the other end of the spectrum, enterprise companies with thousands of customers and deep telemetry data can justify investment in data providers and predictive tools. But even here, Sonia urges caution - conduct small sample tests, run regression analysis, and cross-check data accuracy before building pipeline or campaign strategy on top of a black box.
“Every nugget from a customer conversation at this stage is gold. You don’t want it abstracted through an LLM.” – Sonia Chawla
When organizations start experimenting with AI, they often overlook the importance of internal systems. But messy CRMs, broken lead-routing logic, or poorly mapped account hierarchies can sabotage AI effectiveness just as easily as bad training data.
Sonia walks through some of the most common pitfalls:
Without proper system architecture and documentation, AI initiatives will surface incomplete, misleading, or outright false insights.
This is especially critical in companies that sell into large global organizations. In one example Sonia shares, an enterprise had different legal entities across regions, and marketing automation platforms were incorrectly collapsing records due to matching email domains. The RevOps team had to spend months disentangling system logic just to establish a reliable customer record.
Need help structuring the often messy data of parent-child relationships? Read How to Structure Parent‑Child Account Relationships in Your CRM.
Sonia also highlights the importance of identifying the primary contact for each opportunity and tracking key engagement signals. Why? Because it’s not just about winning the initial deal - it’s about preparing for renewals, upsells, and churn prevention.
Tracking product usage, meeting frequency, and email responsiveness gives RevOps teams predictive power around retention and expansion. But none of that matters if you can’t confidently link those signals back to a named decision-maker.
And this is where Sonia gets tactical:
If your company isn't using a conversation intelligence platform yet, Sonia says you’re leaving massive value on the table. Whether it's Gong, Chorus, or lower-cost alternatives like Fathom or Fireflies, call recording enables downstream automation, competitor intel, churn signal detection, and sales enablement - all in one.
These platforms allow RevOps teams to:
“The cost of recording calls is so low now. There’s no excuse not to do it.” – Sonia Chawla
Camela adds that for marketing teams, these tools also provide a goldmine of messaging data. How are prospects describing their pain? What competitors are they mentioning? What analogies or metaphors resonate?
RevOps pros often struggle to get leadership buy-in for “data cleanup” work. Sonia’s advice: reframe it. Don’t pitch a data quality initiative. Pitch a revenue acceleration lever.
Instead of saying: “We need to clean up our contact mappings,” say:
→ “This will allow us to identify churn risks earlier and create renewal alerts before we lose customers.”
Instead of: “We’re fixing our lead routing logic,” say:
→ “This helps AEs respond 3x faster and improves conversion by 20%.”
“Don’t pitch it as a data project. Pitch it as a time-saver, a revenue lever, or a way to get more accurate performance insights.” – Sonia Chawla
And if you're wondering how AI intersects with something like marketing analytics, don't miss GenAI + Marketing Analytics: What RevOps Needs to Know.
Sonia’s message is clear: AI is not the first tool you buy - it’s the last. You start with defined workflows, validated ICPs, accurate internal data, and stakeholder education. Only then can AI drive meaningful automation or insight.
It’s not that AI isn’t valuable - it is. But it's only as good as the underlying data, logic, and systems it's built upon. If you're just chasing shiny tools, you'll end up doing rework later. Smart RevOps leaders focus first on infrastructure - and then scale what’s working.
Looking for more great content? Check out our blog, join our community and subscribe to our YouTube Channel for more insights.
Our average member has more than 5 years of RevOps experience. That means you’ll have real-time access to seasoned professionals. All we ask is that you’re generous with your knowledge in return.