Why I'm Betting My Career on AI in GTM Ops
Right now, we're witnessing the complete transformation in the way we work.
Six months ago, a data cleanup project that should have taken us weeks only took 47 minutes. That's when I knew I was betting my entire career on AI in GTM Ops.
If you’ve talked to me you will know, I am incredibly bullish on AI – especially as it relates to RevOps and GTM Ops.
More often than not, when I am working with a client, we are dealing with:
Processes that require relatively complex decision-making.
Rules that are difficult to maintain & adhere to.
Reliance on unstructured data.
As an example, there’s an organization that I started working with that fundamentally did not have the same definition of a customer across multiple systems. For the sales team, it was any Account with a Closed Won opportunity, Finance defined it as when the first invoice was paid, and Customer Onboarding considered a customer as someone who has started their implementation.
This lack of consistency across different business units with something as simple as the definition of a ‘customer’ creates massive confusion — not only from a business intelligence perspective but also, business processes and scalability.
Now, once we all came together and agreed upon an org-wide definition of customer we began the clean up across systems.
Typically this would involve multiple CSV exports, endless XLOOKUP formulas, and my laptop fan screaming like it's trying to escape. Even after days of work, we'd still find edge cases we missed, requiring another round of manual fixes.
Introducing AI
1. Show the AI Model What Success Looks Like
When we get less than ideal results from AI in a scenario like data clean up, it’s typically because we let the model decide what success looks like which may not be what we are expecting. The simple solution, give it an example and show your expected output
2. Tell the AI How to Reason
There were multiple different date values and formats (ie. YYYY-MM-DD vs. MM-DD-YYYY). When AI is faced with this decision, you need to tell the AI how to reason.
If multiple different values are found across the data set — Use the value found in ‘salesforce-export.csv.’
Ensure the date format is ‘YYYY-MM-DD’
The more context you can provide the model, the better the output will be.
3. Anonymize Data
One of the most talked about issues in with AI is data privacy. Regardless of what these providers are saying, it is currently the best practice to anonymize any personal data. In my case, I omitted actual company names in place of a random ID.
4. Let the AI model do its Magic
The result? Less than an hour later, we had a perfectly mapped dataset identifying a number of customer record conflicts, complete with resolution recommendations for each discrepancy. A project that typically would have consumed several weeks and multiple stakeholders was completed with a single person and an AI model.
Punchline: Work HAS Changed
That's the moment I realized we're not just looking at a helpful tool — We're witnessing the complete transformation in the way we work and interact with go-to-market data.