Here's something few people talk about when it comes to deploying agents: most of the time they lack the context they need to actually be useful.

Your agents are likely connected to your knowledge base or even your CRM, but when they have no idea what your users are actually doing: where they experience frustration, where they drop-off, or where they continually hit pay walls, you risk losing opportunities or even pushing them to churn. 

And that’s exactly why we partnered with Fin. Because we saw a better path forward; one where support agents have context on who your users are and how they behave in your product. With our native MCP integration, Pendo gives Fin real-time visibility into what users are actually doing inside your product and the ability to act on it proactively, before a user ever files a ticket.

Why "AI-powered support" has mostly missed the mark, until now

The promise of AI in customer success has always focused on three things: 

  • Catch problems earlier 
  • Personalize at scale
  • Let your team focus on high-value work

The reality has been an only slightly smarter ticketing system, and the reason is simple. 

Support AI operates on what users say, not what they do. And users are notoriously bad at articulating the real problem. When they say, "this feature doesn't work." What they mean is: "I've tried to complete this workflow five times, I don't understand the UI, and I'm about to give up and cancel."

Product behavioral data tells you the second thing. But until now, there was no clean path to get those behavioral signals into your support agent in real time.

Pendo captures that behavioral layer: frustration signals, workflow completions, feature adoption rates, NPS, sentiment, and predictive churn scores built on actual usage patterns. What we've built with Intercom is the first native connection between that intelligence and an AI agent that can actually act on it.

What this looks like in practice

Three scenarios that represent the biggest cost centers for CS and support teams today:

A user is stuck before they know they're stuck. A user is mid-workflow and starts rage-clicking. They haven't filed a ticket. They haven't typed anything into chat. But Pendo has already flagged frustration signals;dead clicks, u-turns, time-on-step anomalies. Fin picks up that signal and surfaces contextual help in the moment: the right documentation, a guided walkthrough, or a direct offer to connect with support. The ticket that would have been filed in 20 minutes never gets created in the first place.


An at-risk account starts disengaging before it shows up in your QBR deck. Usage on a key account drops 35% over three weeks. Users stop logging in, or are only acting a fraction of the time they used to be. However, no one has said anything. In the old world, a CSM would catch this in a monthly review, send an email, and hope for the best. With Pendo and Fin, that account is proactively flagged as at risk, and Fin then triggers a personalized re-engagement sequence; a check-in, a "here's what you might be missing," a prompt to schedule time with the team. Proactive action, weeks earlier than any human would have caught it.



A power user hits the ceiling right when the conversation should happen. A user has exported 47 files this month. Their plan limit is 50. They've also spent time in the last two weeks exploring features that sit behind a paywall. Pendo surfaces both signals. Fin initiates a timely, non-pushy conversation about what's available at the next tier. Instead of relying on the traditional broadstroke email blast, now you have contextual messaging in-product, the moment where a user is the most likely to engage with it.

The data behind the decision

We didn't need to brainstorm ways to bring our products together. We built this because the data that was already there made the problem impossible to ignore.

When we looked across Pendo accounts that churned versus accounts that renewed and expanded, the behavioral signals were consistent and predictable:

  • Declining feature breadth
  • Drop-off at specific workflow steps
  • Reduced session frequency
  • Rising frustration signals
  • Declining sentiment and low NPS scores

The data was there the whole time. Now there's an automated system to connect that intelligence to your activation layer.

The teams we built this for

Connecting customer support and upsell motions to product behavioral data has been genuinely hard to do, until now. By partnering with Fin, we've unlocked new use cases for real-world workflows:

Customer Success teams spend too much time on reactive firefighting and not enough time on strategic accounts. The integration handles the high-volume, time-sensitive interventions automatically so CSMs can focus on the relationships that actually need a human.

Support leaders trying to improve CSAT and deflection rates without adding headcount. Contextual, proactive support, before the ticket, is the highest-leverage deflection strategy available.

Product and Growth teams that already know where users are struggling (because Pendo tells them) but have lacked a direct line to intervene in real time. Now they do.

Getting started

The integration is now in beta and available to test for customers on both platforms.

If you're interested in getting started, fill out the form to let us know and we'll be in touch.

The Pendo × Fin integration is built on Pendo's MCP (Model Context Protocol), which connects Pendo's behavioral intelligence directly to AI tools and agents. Learn more about Pendo MCP.