How Mistral AI built a global GTM engine with Clay

They populated Salesforce with over 25,000 enriched accounts, creating a foundational TAM upon which they built geographic segmentation, equitable sales territories, and an initial account prioritization model. 

This TAM mapping effort gave the team the planning data they needed to inform their global GTM strategy. For example, how “big” are the Nordics for Mistral AI? What is the quality level of accounts in that region? How many reps are needed to support them?

Alice Liu, RevOps Manager, began the account mapping exercise by using Find Companies in Clay to source accounts that met a minimum employee size threshold. 

‍“I was surprised at the breadth and depth of the accounts that were available globally. Over the course of just 2 weeks with Clay, we had 25,000 accounts that formed a robust starter TAM. Normally this would have taken 2 months.” - Joel Davidson, VP of Revenue Operations

She and her team deduped, enriched, and standardized their company list in Clay, using different Clay tables for each region and industry to accommodate different cleaning processes. Next, they reconciled it with companies that were already in their Salesforce instance, either updating existing account records or creating new ones. Joel says, “Clay’s ability to curate and sculpt company sourcing data is good, especially for finding niche companies and public entities.”

Mistral AI uses over 10 firmographic providers in Clay, where formulas and AI features standardize the data before syncing it with Salesforce. This increases global data coverage and improves accuracy at a lower cost than signing multiple large annual contracts. 

‍“One of the biggest wins with Clay is that you avoid vendor lock-in and have optionality around the data providers that will get you the best information for your global markets.” — Joel Davidson

If a data point is missing or inaccurate, Mistral AI fills in the gap with a Claygent prompt, powered by Mistral AI’s own models. For example, the RevOps team relies on a company’s estimated revenue for segmentation, routing, and scoring. When data providers are missing revenue numbers – e.g., for small European companies or young digital native startups – Mistral uses a Claygent prompt to search the web and combine up to five data points to offer an estimated revenue range. 

“I was used to only having one data provider. You love it or you hate it, and if they don’t have a data point, then they just don’t have it.  One of the first things I learned with Clay was that I can try and seek this information out. I can run a Claygent prompt or bring in another provider, and sync it all with our CRM. And that was new and interesting.” — Joel Davidson

With AI in Clay, Mistral runs custom research at scale and brings richer attributes into their scoring model than traditional firmographics alone.

To build these attributes, Mistral’s RevOps team asks sales reps which account-research questions matter most. They then find proxy data points and turn them into “Mistral AI Metrics” in Clay. 

One example estimates an organization’s investment in AI compared to its peers: Claygent, using Mistral AI models, searches news mentions of AI investment and looks at the ratio of open AI jobs per $1B in revenue. They also use AI paired with technographic data from HG Insights to estimate a company’s GPU capacity and on-prem footprint. This information helps reps take the right sales approach for each account.

“We did an analysis of on-prem trends regionalized across the world in 3 days. This market research analysis would have taken my team 3 months to do without Clay. It was enlightening to see how speedy you could get to a V1 answer we could further refine.” — Joel Davidson

‍

Signals like these make up 25% of Mistral AI’s account scoring logic today; the other 75% is classic firmographics. Joel hopes to add more AI-driven scoring attributes to flip that ratio. “Clay is the coolest tool I’ve used in terms of using AI to get niche data points,” he says. “We have a feel for what good accounts look like, so it’s about continuing to sculpt it and make the attributes real for teammates who are onboarding.”

This intelligence powers an automated ICP dashboard in Salesforce that helps the team refine their ICP and understand the size and makeup of their TAM.

Mistral AI’s sales team uses product usage as a signal to help prioritize which accounts to reach out to. To do this, Alice created a workflow that uses Clay as an orchestration layer that connects Salesforce to Mistral AI’s invoicing software and customer usage data in BigQuery. 

Mistral AI’s sales reps can see product consumption data in Salesforce, which highlights self-serve signups that are ramping up organic usage and ready for a sales conversation. Reps can also see which customers are about to hit their usage threshold so they can reach out to expand the account, as well as track rolling 3- and 6-month averages for customer expansion and contraction.

"Clay is powering 95% of our consumption workflow," says Joel. “We have signup telemetry flowing to the sales team: hey, there’s smoke here, there might be fire. We’ve started to see some huge ramp-ups from companies we really should know are ramping up. Now, reps can take action a little bit sooner.”

“Clay is a great ETL platform. At past jobs, I spent $500k on data integration consultants and platforms, and I spend zero money on this with Clay. Within 6 weeks the system was live and working, which is pretty incredible.”— Joel Davidson