​Are You Realizing The Power Of Your Archived Customer Data?

Stu Sjouwerman is co-founder and CEO of ReadingMinds, a pioneering AI-moderated interview platform for conducting sentiment analysis.

getty

"There’s gold in them thar hills!" is an apocryphal cry generally tied to the old frontier gold rush and sometimes attributed to Mark Twain’s 1892 novel The American Claimant. That gold rush is long over, but there’s a new kind of gold ready to be mined today: customer data.

Virtually every organization today is sitting on an archive of rich customer data in the form of conversations, whether they realize it or not. Recorded sales calls. Quarterly customer reviews. Demos and onboarding sessions. Zoom calls. The list could go on and on.

David Haber, a general partner at Andreessen Horowitz, describes this as “a living context layer”— unstructured voice data lying ready to become structured and poised to provide context for both employees and AI agents.

Why Data Lacking Structure Lacks Value

While recorded data are plentiful today, the vast majority of that data is unstructured. The continually growing motherlode of customer conversations lies dormant and is no more decision-ready today than it was before technology allowed its capture and storage.

The problem is that while audio transcripts document what was said, they readily lack the richness and nuance of tone, intonation and intent, which would make it more valuable to marketers and others.

When Words Survive While Context Disappears

Reading that a customer said: “okay,” doesn’t indicate whether they were resigned or excited. When a prospect remarks, “that’s interesting,” it could indicate interest or dismissal.

Organizations are rapidly losing the value of actionable data that could help drive revenue decisions. But they have the opportunity to capture that context and put it to use by incorporating an expression context layer.

The Expression Context Layer

You’ve already got the data in the form of recordings. Structuring that data by adding time-stamped evidence to show exactly where a customer expressed hesitation, indicated conviction or urgency or displayed genuine interest or disengagement reflects the gold in your data.

We refer to this as an "emotional system of record," a tangible indication conveyed through a structured analysis to show how responses were expressed, and allowing every finding to be traced back to a quote, timestamp, signal label or intensity score.

Using expression labels (sad, angry, confrontational, neutral, cheerful and enthusiastic) and scoring them to indicate intensity over time can provide longitudinal intelligence. Note that these labels describe how responses are expressed in conversation, not actually what individuals privately feel.

The system can track expression patterns at the individual account level (with appropriate governance). In aggregate, across segments, trends such as increasing hesitation or anger related to a specific topic, or a shift in enthusiasm following a product release, can be revealed.

Today, it’s not just humans who use these inputs to guide decisions. AI agents are also consuming data directly. Agents using only flat transcripts miss the expression context that would make decisions more accurate. That missing layer leads to actions being taken on customer conversations based only on what was said and not how it was delivered.

When an agent cites expression evidence before triggering an escalation or drafting a renewal outreach, the action is traceable. Without it, the action is a guess with consequences that compound at agent speed.

Of course, along with the promise of positive potential from this technology, legitimate concerns emerge and must be addressed.

Respecting Boundaries

The ability to precisely analyze expression signals can be powerful for driving business decisions, but must never become a tool to serve as “lie detectors,” to diagnose individuals’ mental states or to monitor employees.

These boundaries are important to understand both from a brand perspective and to ensure compliance with existing and emerging laws and regulations.

For instance, the European Union AI Act specifically prohibits AI systems that infer emotions in workplace and educational settings. More broadly, organizations using expression analysis should treat governance as a design requirement, not as an afterthought.

It’s important for organizations to develop and communicate policies that include consent, source authorization, retention controls, redaction, access policies and audit trails.​

Chances are good that you already have masses of prospect and customer input data that could be mined for the gold that lies within. Converting flat transcripts into searchable expression evidence offers access to trend intelligence that can power AI agent interactions into providing accurate, relevant and real-time trend analysis useful for business decision-making.

Your transcripts can tell you what was said. Conducting a sentiment analysis and exploiting an emotional system of record can yield further insights into how those words were expressed and what that potentially signals about the customer experience.

You already have an archive. The question is, when will you start mining for gold?

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?