Delivering Insights with Exceptional Accuracy

Observe.aiObserve.ai

Company:

Industry:

Technology

About Observe.ai:

Observe.ai is the leading AI platform for contact centers, serving leading customers across verticals with their comprehensive conversation intelligence platform. Their platform processes millions of customer calls and chats, extracting insights from both conversation transcripts and structured metadata

Accuracy 95%

"We wanted to make insights accessible to everyone at interactive speeds and uncompromising accuracy standards. After evaluating alternatives like Snowflake Cortex and other leading vendors offering NL to SQL translators, Iris stood out for the level of accuracy and quality of the user experience."

- Vache Moroyan, Chief Product Officer at Observe.AI

Challenge

Expert Analyst Quality At Interactive Speeds 

As contact center managers increasingly demand near instant insights for operational decisions across unpredictable scenarios, Observe.AI decided to build a service called AskObserve - a conversational intelligence platform that enables unconstrained exploration of contact center transcripts and data at uncompromising levels of accuracy. AskObserve is a strategic initiative for Observe’s Chief Product Officer, Vache Moroyan and is led by Sarika Kumari, a lead Product Manager in the team.  To deliver the solution to some of the largest companies in their respective industries, Observe.AI was seeking to address the systemic challenges of traditional BI based solutions outlined below

  • Analyst dependence slows business impact: Given the high accuracy requirement, important analyses currently need to be run by data analysts who understand the data models . However, given the small size of these teams, it creates a bottleneck  "By the time you get these insights,  it's like a week or so, making them practically not useful," Sarika explains
  • Scarcity of analyst bandwidth makes insights inaccessible for operational decisions: Given the demand on analysts, access to them was limited to specialist teams, not the operations managers who needed to make daily decisions. However, running a Contact Center efficiently requires continuous adaptation and optimization‍
  • Pre-created dashboards do not offer necessary degrees of freedom in exploration: While dashboards are available in plenty to seemingly fill the gap, they fail when a problem needs exploration that does not fit the confines of a dashboard. As Sarika Kumari points out, "They'll share a report, then you will again send back a mail saying, hey, these are my follow up questions"

The business consequences of these inefficiencies were noticeable. For example,  common tasks such as Agent coaching required synthesizing data from multiple dashboards, taking 20-30 minutes per agent just to understand performance patterns. What was needed is a product that meets the accuracy bar of an expert analyst without any of the bottlenecks of a human centric operation.

The technical challenges that needed to be addressed to realize such a system were significant.  Building a conversational analytics product that  could interpret problems stated as arbitrary natural language questions using business jargon, figure out a solution plan and generate accurate SQL queries is really difficult.  Sarika recognized that generic AI tools wouldn't solve this structured data challenge:  "Converting natural language questions into accurate SQL queries isn't something you can solve by simply using standard LLMs with your data—it requires specialized expertise and an intense focus on the experience."

‍

Solution

Embed Iris as an Accurate and Intelligent Analyst

Observe.AI integrated Iris to power structured data analytics within their AskObserve platform. Iris interprets questions that use customer specific terms and acronyms,  generates accurate SQL and synthesizes insights from the analyses. These questions relate to contact center metrics such as Average Handle Time, CSAT scores, agent performance metrics sliced by numerous dimensions. This enables contact center teams to ask business questions in plain English and receive accurate, data-backed insights without requiring SQL knowledge or analyst intervention. 

The decision process to move forward with Iris was rigorous. Vache pointed out that “After evaluating alternatives like Snowflake Cortex and other leading vendors offering NL to SQL translators, Iris stood out for the level of accuracy and quality of the user experience for business teams” . The product’s business user experience, coupled with the innovative approach to  ground the product with nuances of context about the data were clearly differentiated strengths of the product. Additionally, the Iris team’s approach of partnering closely to delight the end user of AskObserve aligned with Observe’s vision. 

The engagement with the Iris team has been strategic and the richness of the partnership goes beyond a traditional vendor relationship. "It's not just kind of a vendor engagement, it's more of a strategic partnership and it feels like one team," Sarika explains. The Iris team has been a keen listener and collaborator,  and has evolved the product to ensure that the users of AskObserve are delighted with the experience.

Results

Ask Observe with Iris is Live and Serving Brands You Know

  • It's working. Insights that were previously inaccessible in the time scales where they were useful are now readily available in under a minute. The sheer volume and variability of contact center data made comprehensive analysis practically impossible to staff and deliver accurate answers for - adding more dashboards wasn't an option. 
  • The transformation is strategic. Contact centers can now operate with intelligence that was previously locked behind analyst workflows, enabling data-driven decisions at the speed of operations rather than the speed of reporting. This isn't just about faster analysis—it's about fundamentally changing how contact centers operate.
  • It's scaling beautifully. The platform launched with enterprise customers first. The natural language interface enables any persona—from operations to QA to leadership—to get accurate, actionable insights without requiring any technical training.