How ContractSafe uses Dust and Fathom to deploy AI Agents across teams

ContractSafe was founded in 2015 and provides businesses with a refreshingly easy, affordable AI-powered contract lifecycle management (CLM) software built for the way real teams work. From request to renewal, ContractSafe brings the full contract workflow into one place without the complexity or cost of heavier enterprise systems. Consistently recognized for ease of use, fast implementation, and being easiest to do business with, it's built for businesses that want contract management their whole team will actually use. 

When Alexa Carr joined ContractSafe in 2025 as a go-to-market engineer, she saw a chance to push the company's AI ambitions further. For an AI-native product like ContractSafe, investing in a frontier AI stack made sense, and she was ready to lead the charge.

Her mandate was initially narrow: find ways AI could help the go-to-market team. But she quickly understood the opportunity was much larger than a few sales automations. The real question was how to build something that could reach every team in the company, not just revenue. 

The data was there. ContractSafe runs on conversations: sales demos, customer kickoffs, annual business reviews, renewal calls. All of it recorded in Fathom, ContractSafe’s meeting notetaker. All of it sitting largely untapped beyond the notes each rep walked away with. The same was true across their CRM, their customer success platform, their product usage data. 

The challenge extended beyond technicalities, it was organizational as well. How do you get an entire company, spanning sales, customer success, product, and engineering, to actually change how it works? And how do you do it without a team? 

From the start, Alexa knew ContractSafe needed AI agents. The question was how to actually get the whole company using them. 

"I knew right away that this department was going to need to build agents. You can do it in Claude Code, of course. But how are we going to share it? We just really didn't see a way to socialize it with Claude Code." 

The engineers kept using Claude Code for what it does best. Dust handled everything else: a shared environment where agents & skills could be built once and deployed across the whole company, where anyone could jump into a workflow, and where the AI team could actually stay in the loop. 

"The fact that any user can @mention me right inside a Dust conversation has been world-changing," Alexa said. 

On top of that, Dust integrated natively with Fathom, turning every recorded conversation into live, queryable data for their agents. 

From there, Alexa started building. And almost immediately, the Fathom transcripts that had been sitting untouched became the raw material for some of the team's most valuable workflows. 

ContractSafe uses two approaches to get Fathom data into Dust. For agents that need to react to a single recent call, they use the native Fathom MCP connector. For agents that need to reason across large volumes of conversations over time, they route Fathom transcripts into Dust folders via an N8N workflow. That two-track setup powers nearly everything that follows. 

When a demo call ends, a Dust agent reads the Fathom transcript and automatically populates BANT(their sales methodology of budget, authority, need, and timeline) fields in HubSpot. No manual entry, no follow-up reminder.

Those fields then feed directly into a proposal generator agent. It knows the company, the likely plan, the number of contracts. It builds the proposal. Once this pipeline is fully live, the entire post-demo workflow will run with minimal human intervention. 

After a CSM completes a kickoff call or an annual business review, a Dust agent pulls the Fathom transcript, scores it against a defined rubric, and sends the rep instant feedback via Slack. No waiting, no manager bottleneck. 

At the end of the week, the CSM manager gets a summary: stack rankings, who is performing well, who needs support, and where the coaching gaps are. This feeds directly into one-on-ones the following week. The same agent is being extended to cover AEs for sales calls. 

ContractSafe's CSMs own gross revenue retention (GRR) as a core KPI. Predicting churn used to mean pulling data from multiple systems manually. Now, a Dust agent does it automatically. 

Every week, CSMs get alerts flagging high-risk accounts: usage drops, open Jira tickets, renewal dates, ARR at stake. Everything in one view. 

"When is their renewal date? Is it in two weeks? How many Jira tickets have they submitted? Is it related to their drop in usage?" 

This kind of analysis simply did not exist before. "This was just work that wasn't getting done. It's a net new thing for us." 

Using Fathom folders loaded with historical recordings, a Dust agent scans large volumes of calls to surface competitive patterns: which competitor names come up, in which contexts, how often. Sales leaders can now access this intelligence on a regular cadence. Before, it required manual review of individual calls. Now it runs continuously across the full archive. 

When a CSM has an annual business review on the calendar, an agent detects it, pulls all relevant account data, builds a talk track, and generates the presentation. Automatically. 

"That is at least an hour per meeting." 

Five agents running is one thing. Getting everyone at the company to actually build and own them is another. 

This quarter, ContractSafe is running a company-wide agent building contest. Every employee builds at least one agent. The goal is that everyone, regardless of role, gets hands-on experience with what it feels like to design and deploy an AI workflow. 

Early submissions are already generating ideas the AI team hadn't considered. Many of them are built around Fathom data. 

"A lot of the use cases that our users have come up with for the contest, a lot of them have to do with Fathom and what they're getting out of their calls." 

The results are still early, but the direction is clear because some of the time savings are already calculable. CSMs are saving at least an hour per annual business review. Once the end-to-end sales pipeline is complete, AEs are expected to save one to two hours per deal. With a full calendar of customer reviews and demos, those hours compound fast. 

For the churn monitoring work, the ROI calculation is still in the works. The weekly alerts launched just weeks before this story was written. But the shift is already real: CSMs now have a single view of account risk that didn't exist before. The work is no longer reactive. 

ContractSafe is building a Dust-powered agent registry to track the impact of each workflow more systematically as the stack grows. 

Alexa is direct about something that gets lost in most AI conversations: AI has not reduced the workload at ContractSafe. It has changed what kind of work gets done. 

"It's not that we're working less now that we're AI-native. It's that we're working so much bigger. We're getting to know our business in a different way and moving faster. That is really what Dust has been able to do for us." 

Perhaps the most concrete proof of that shift may be Alexa's own job title. She joined ContractSafe as a go-to-market engineer. Six months later, she is Head of AI Operations and Automation, with a mandate spanning every team in the company. The job morphed into something with greater impact than ContractSafe originally thought was possible. 

That is what Dust was built for: multiplayer AI that compounds knowledge across an entire organization. With the Fathom MCP, every conversation becomes part of that intelligence. Every

conversation your team has had is still sitting there, waiting to be useful. The only question is who decides to make it count.