🎉 Decagon raises $131M series C at a $1.5B valuation
Read our post

Blog

Product

Watchtower: Always-on QA for every conversation

June 26, 2025

Written by Bihan Jiang

AI agents are redefining customer experience, handling thousands of conversations at once, resolving support issues faster, and freeing up human agents for higher-value work. But at this scale, a critical question emerges: How do you know what your customers are actually saying across all of your AI and human agents?

For CX leaders, this lack of visibility means potential risks go unnoticed, while valuable insights get lost in the noise. The current quality assurance (QA) playbook of “vibe checking” a few conversations or running reactive audits simply doesn’t scale.

That’s why we built Watchtower: an always-on QA system that gives you visibility into every conversation, without the manual burden. Whether it’s catching a compliance red flag or identifying recurring customer issues, Watchtower reviews every conversation—whether with an AI or human agent—against your customizable criteria to turn a sea of raw transcripts into actionable insights and business opportunities.

The visibility gap in AI-driven CX

Most QA processes for CX rely on spot-checking a few conversations flagged by customers or triggered by basic keyword matches. But when agents are handling thousands of conversations every day, that approach breaks down fast.

Traditional QA methods fall short for a few key reasons:

  • Reactive and incomplete: Spot-checking a small fraction of conversations means you’re always behind. Issues only surface after customers are impacted, or not at all. And in regulated industries, it’s untenable. These companies need every conversation reviewed to stay compliant and avoid costly blind spots.
  • Operationally heavy: Manual QA eats up time and headcount. Teams waste hours sifting through conversations to find the few that matter, rather than working on more strategic projects. As volume grows, this approach becomes increasingly expensive and unsustainable.
  • Shallow and rigid: Static keyword matching might catch a mention of “credit card,” but it won’t tell you if it’s an actual compliance risk or just a harmless reference. Without contextual understanding, teams lose time chasing false positives and miss the nuances that matter.

Introducing Watchtower: Your always-on QA system

Watchtower gives you complete, real-time visibility into every conversation, without the manual overhead of traditional QA.

Whether you're monitoring for compliance risks, negative sentiment, or upsell signals, Watchtower reviews every interaction against your custom criteria and surfaces what matters.

Here’s how Watchtower solves the visibility problem:

  • Always-on and comprehensive: Instead of reviewing a small sample of conversations after the fact, Watchtower analyzes every single conversation in real time. Whether it’s a compliance mandate, a tone issue, or an upsell signal, nothing slips through the cracks.
  • Effortlessly scalable: Watchtower eliminates the need to manually sift through transcripts. It automatically flags conversations based on custom criteria and categorizes issues for fast triage, so teams get the insights they need without the typical operational burden.
  • Context-aware: Unlike static keyword alerts, Watchtower understands why a conversation is relevant. It considers context, tone, and behavior to distinguish between harmless references and real issues, providing more accurate and granular analysis of customer interactions.

How Watchtower works in practice

Watchtower is easy to set up and fully self-serve. Within minutes, teams can define what they want to monitor and start uncovering insights, with no engineering support required. Under the hood, it offers powerful capabilities that make conversational oversight scalable and simple.

Natural language flags

Define what to monitor using everyday language—not rigid rules or static keyword lists. Whether you're looking for compliance risks or signs of customer frustration, just describe what to track, and Watchtower will flag the right conversations using contextual understanding.

Flexible time windows

Some teams need ongoing visibility into every conversation, while others want to investigate specific time periods or past incidents. Watchtower supports both. Simply set a desired time range, and it handles the analysis.

Precision filters

Zero in on the conversations that matter. Watchtower lets you filter by CSAT, support channel, user metadata, whether a conversation was resolved or escalated, and more. This keeps your team focused on high-impact conversations and avoids them drowning in irrelevant data.

Custom categories

Organize flagged conversations into subcategories for deeper insight. A “compliance” flag might include payment handling or data privacy. A “sentiment” flag might break into frustration or confusion. Categories help you uncover granular patterns and prioritize issues accordingly.

Dashboards and drilldowns

Move seamlessly between the big picture and the fine details. Watchtower’s dashboard highlights trends across flags and categories so you can spot issues, track progress, and understand where to focus. If you want to dig deeper, you can jump straight into the conversations that were flagged.

From compliance to competitive advantage

AI agents have unlocked a new level of quality and scale in customer experience, but today’s QA approaches have failed to keep up. Without knowing what your customers are saying, you can’t ensure agents meet company standards, spot issues early, or act on emerging opportunities.

Watchtower gives you that visibility. It automatically reviews every conversation against your custom criteria, flags what matters, and helps you uncover insights that improve performance and reduce risk—without adding operational overhead.

Many teams start by using Watchtower to catch negative outcomes like compliance violations, low CSAT, or escalations. Over time, they expand its use to uncover product insights, common feature requests, and signals tied to buying intent. With the expanded scope, your support org becomes a source of truth for how to improve your business across product, ops, and more.

If you're deploying AI agents to redefine your customer experience, let’s talk! Get a demo to see how Watchtower and the rest of Decagon’s conversational AI platform can help.

Blog

Product

Watchtower: Always-on QA for every conversation

A deep dive into our always-on QA system that reviews every conversation against your customizable criteria, without the manual burden.

AI agents are redefining customer experience, handling thousands of conversations at once, resolving support issues faster, and freeing up human agents for higher-value work. But at this scale, a critical question emerges: How do you know what your customers are actually saying across all of your AI and human agents?

For CX leaders, this lack of visibility means potential risks go unnoticed, while valuable insights get lost in the noise. The current quality assurance (QA) playbook of “vibe checking” a few conversations or running reactive audits simply doesn’t scale.

That’s why we built Watchtower: an always-on QA system that gives you visibility into every conversation, without the manual burden. Whether it’s catching a compliance red flag or identifying recurring customer issues, Watchtower reviews every conversation—whether with an AI or human agent—against your customizable criteria to turn a sea of raw transcripts into actionable insights and business opportunities.

The visibility gap in AI-driven CX

Most QA processes for CX rely on spot-checking a few conversations flagged by customers or triggered by basic keyword matches. But when agents are handling thousands of conversations every day, that approach breaks down fast.

Traditional QA methods fall short for a few key reasons:

  • Reactive and incomplete: Spot-checking a small fraction of conversations means you’re always behind. Issues only surface after customers are impacted, or not at all. And in regulated industries, it’s untenable. These companies need every conversation reviewed to stay compliant and avoid costly blind spots.
  • Operationally heavy: Manual QA eats up time and headcount. Teams waste hours sifting through conversations to find the few that matter, rather than working on more strategic projects. As volume grows, this approach becomes increasingly expensive and unsustainable.
  • Shallow and rigid: Static keyword matching might catch a mention of “credit card,” but it won’t tell you if it’s an actual compliance risk or just a harmless reference. Without contextual understanding, teams lose time chasing false positives and miss the nuances that matter.

Introducing Watchtower: Your always-on QA system

Watchtower gives you complete, real-time visibility into every conversation, without the manual overhead of traditional QA.

Whether you're monitoring for compliance risks, negative sentiment, or upsell signals, Watchtower reviews every interaction against your custom criteria and surfaces what matters.

Here’s how Watchtower solves the visibility problem:

  • Always-on and comprehensive: Instead of reviewing a small sample of conversations after the fact, Watchtower analyzes every single conversation in real time. Whether it’s a compliance mandate, a tone issue, or an upsell signal, nothing slips through the cracks.
  • Effortlessly scalable: Watchtower eliminates the need to manually sift through transcripts. It automatically flags conversations based on custom criteria and categorizes issues for fast triage, so teams get the insights they need without the typical operational burden.
  • Context-aware: Unlike static keyword alerts, Watchtower understands why a conversation is relevant. It considers context, tone, and behavior to distinguish between harmless references and real issues, providing more accurate and granular analysis of customer interactions.

How Watchtower works in practice

Watchtower is easy to set up and fully self-serve. Within minutes, teams can define what they want to monitor and start uncovering insights, with no engineering support required. Under the hood, it offers powerful capabilities that make conversational oversight scalable and simple.

Natural language flags

Define what to monitor using everyday language—not rigid rules or static keyword lists. Whether you're looking for compliance risks or signs of customer frustration, just describe what to track, and Watchtower will flag the right conversations using contextual understanding.

Flexible time windows

Some teams need ongoing visibility into every conversation, while others want to investigate specific time periods or past incidents. Watchtower supports both. Simply set a desired time range, and it handles the analysis.

Precision filters

Zero in on the conversations that matter. Watchtower lets you filter by CSAT, support channel, user metadata, whether a conversation was resolved or escalated, and more. This keeps your team focused on high-impact conversations and avoids them drowning in irrelevant data.

Custom categories

Organize flagged conversations into subcategories for deeper insight. A “compliance” flag might include payment handling or data privacy. A “sentiment” flag might break into frustration or confusion. Categories help you uncover granular patterns and prioritize issues accordingly.

Dashboards and drilldowns

Move seamlessly between the big picture and the fine details. Watchtower’s dashboard highlights trends across flags and categories so you can spot issues, track progress, and understand where to focus. If you want to dig deeper, you can jump straight into the conversations that were flagged.

From compliance to competitive advantage

AI agents have unlocked a new level of quality and scale in customer experience, but today’s QA approaches have failed to keep up. Without knowing what your customers are saying, you can’t ensure agents meet company standards, spot issues early, or act on emerging opportunities.

Watchtower gives you that visibility. It automatically reviews every conversation against your custom criteria, flags what matters, and helps you uncover insights that improve performance and reduce risk—without adding operational overhead.

Many teams start by using Watchtower to catch negative outcomes like compliance violations, low CSAT, or escalations. Over time, they expand its use to uncover product insights, common feature requests, and signals tied to buying intent. With the expanded scope, your support org becomes a source of truth for how to improve your business across product, ops, and more.

If you're deploying AI agents to redefine your customer experience, let’s talk! Get a demo to see how Watchtower and the rest of Decagon’s conversational AI platform can help.

Download the

Blog

Product

AI agents for concierge customer experience

Get a demo