PandaDoc

The challenge: breaking free from the ticket-taking cycle

At PandaDoc, a SaaS platform that streamlines the entire lifecycle of business documents, the data team found themselves stuck in a familiar loop.

Roughly 80% of analysts’ time was spent fielding repetitive requests: basic metrics, ad hoc queries, and endless “what happened?” questions. Stakeholders waited on answers — or worse, made decisions without accurate data. Strategic, high-impact work sat waiting on the sidelines.

When business users began experimenting with AI chatbots and arriving at inconsistent numbers, Hannah Burak, PandaDoc’s Data Product Manager, knew they needed a better path forward. She began searching for an AI-powered self-serve analytics tool that her team could trust and scale across the organization.

Why Hex Threads outperformed standalone AI solutions

After testing out a handful of standalone AI-for-BI chatbots and in-house solutions, Hex's conversational interface, Threads, was the clear winner. Why?

- It was more trustworthy for business users: Their homegrown solution connected dbt documentation through Select Star, but Claude couldn't leverage their semantic layer, making it essentially unusable for business users. "The semantic layer is really the key for us in getting better performance," Hannah explains. By connecting Cube's semantic models directly into Hex via Semantic Sync, analysts defined business logic and metrics in one place, which means less time untangling SQL discrepancies and more time delivering answers.

- It was a faster integration and setup: Hex solves this by integrating directly with their semantic layer in Cube and existing workflows, with setup taking minutes compared to hours with other tools.

The solution: accelerating analysts and empowering business users

Delivering relevant answers for the data team

Speed was essential. The team needed to cut data request delivery times for a variety of ad hoc questions. Threads helps by surfacing results that are actually relevant to the question being asked. “The hardest challenge with any LLM on top of a data warehouse is finding the right data and that's what Threads does well,” Hannah explains.

This means getting faster answers at a fraction of the wait time. In one example, Hannah used Threads to get click-through rate data for a pricing page button when their web analyst was on leave. Threads was able to deliver an answer that the data team could verify in five minutes instead of the 20 it would have taken manually.

Reducing self-serve ad hoc burden with the finance team

PandaDoc also rolled out Threads to 20 business users in finance. The goal was to help the finance team get quick answers to routine "what happened" questions on their own, in order to free analysts to focus on more strategic work.

With Threads' natural-language interface, finance users could ask data questions in plain English and get SQL-backed answers. PandaDoc configured it to prioritize semantic models and fall back only to endorsed warehouse tables. For straightforward descriptive questions, Threads delivered accurate results. For more complex asks, it gave finance users enough context to either refine their question or know when to loop in an analyst.

"Threads doesn't just give you an answer — it walks you through the logic like a colleague would. It feels like working with an analyst who knows how to explain their thinking every step of the way.”

Easy for the data team to audit

Finally, Threads made it easy for Hannah and her team to validate answers that were generated with the agent. Their stakeholders can send them a thread for review, and then the data team can jump into the notebook to see exactly how the sausage was made: from queries written to semantic models referenced to charts generated.

"We love that because it's inside Hex, when we're called on to validate results, it's really easy to grab whatever is in there, run it ourselves, inspect the SQL, and confirm it's using the right tables," Hannah said,

The impact: Quick answers that are relevant

Threads is helping Hannah’s team take the first meaningful steps toward freeing analysts from reactive, “what happened” questions and empowering business users to explore data themselves. While the data team still spends much of its time on foundational analytics, Threads is already making their work faster, more relevant, and easier to audit.

Early results have shown that Threads can deliver useful and contextually relevant answers. This balance between speed and confidence is especially valuable for product and business teams making quick decisions, where timely insights matter as much as precision.

By syncing semantic models into Hex, Hannah’s team ensures that Threads draws from the right data sources — turning what was once a challenge of accuracy into a foundation for trust. Though the journey toward full-scale self-service across the organization is still unfolding, these early results show clear promise: a future where analysts spend less time on routine requests and more time driving strategic, impactful work.