The Narrative Intelligence Gap
Why Leaders Still Don’t Know What Their People Know
The most valuable intelligence inside any company isn’t in your dashboards. It’s in what your people would tell you if you asked the right question, the right way, at the right time. Closing the distance between what your dashboards show and what your people actually know is the most important leadership problem of the next decade, and the one your current stack can’t solve.
By Chuck Cotter, CEO and Founder, Savo 14-minute read
Overview
The Narrative Intelligence Gap is the distance between the vast amount of experience and knowledge the people inside your organization have, and what your systems can capture and show. It exists primarily because dashboards, HCM platforms, CRM platforms, surveys, and passive call-recording tools were built to store what already happened. None were built to actively elicit the judgment, context, and lived experience of every person in your company, in their words, in their stories. Layering generative AI on top of existing structured data systems doesn't help to close the gap, it actually widens it, by producing fluent-sounding summaries of incomplete input. Closing the gap requires a new capability: a measurement-grade system that conducts intelligent conversations and anchors every conclusion to evidence. That capability is what we at Savo call Narrative Intelligence, and it is the category we are building.
A question I keep getting from CEOs
A few weeks ago I was on the phone with a public-company CRO. Twenty years in the seat, a smart operator, he runs a team most of us would envy. He said something I have now heard, almost word-for-word, from a dozen leaders this quarter.
“Chuck, I have more data than I have ever had in my life. CRM with eighty fields per opportunity. Call recordings transcribed and themed. Quarterly engagement survey, pulse survey, exec 360, brand tracker. An AI tool that summarizes everything for me on Monday morning.”
He paused.
“And I still cannot answer the questions my board asks. What do my people actually know that I don’t?”
I have spent more than thirty years on the inside of enterprise software — at SAP, Informatica, Workday, and now Savo. The pattern he was describing is not new. What is new is how visible this tacit knowledge gap has become, now that every executive has a layer of generative AI sitting on top of the same incomplete inputs and producing the same plausible, polished, partly-wrong summaries — only faster.
That distance, between what your people know and what your systems can show you, is what I call the Narrative Intelligence Gap. It is the single largest unmonetized asset on the modern enterprise balance sheet, and it is the reason most “AI transformation” initiatives are quietly disappointing the people who funded them.
Here is what the gap is, why it has gotten worse rather than better, what closing it requires, and why it’s critical to place human intelligence at the center of the AI transformation we are experiencing in real time.
What is the Narrative Intelligence Gap?
The Narrative Intelligence Gap is the measurable distance between two kinds of knowledge inside your organization. On one side: the judgment, context, and lived experience that lives in the minds of your people — sales reps, customer success managers, leaders and managers, engineers, customers, candidates. On the other: the information your decision systems can actually capture and surface for a leader to act on.
Roughly eighty to ninety percent of organizational knowledge never enters a system in a usable form. It lives inside conversations, in instincts, in objections a rep heard on a Tuesday and never logged, in a comment from an exit interview that everyone nodded at and no one captured. It lives in what people would say if you asked them the right question.
This isn’t a reporting problem, a workflow problem, or a survey design problem. It is an intelligence access problem, and the tools we have used for two decades to address it were built for a different job.
Why every system you already own misses it
The reason the gap exists is not that your tools are bad. It is that your tools were built for different jobs.
A CRM was built to administer the pipeline. A survey was built to collect responses to questions you already wrote. Neither were built to tell you where the intelligence gaps are, what you might be missing, what’s inside your people’s heads but can’t show up in drop down fields. An HR system was built to administer people, not understand them and is filled with bias (reviews, comp, succession). Listening platforms were built to transcribe and summarize what happened with inference engines. Every one of these is excellent at its original job. None was built to elicit and measure what is in your people’s heads. Layer generative AI on top of these disjointed single purpose systems and you inherit the original constraint: AI can only analyze what was captured. It cannot contemplate the conversations it never had.
Here is how this shows up in the four places it costs the most.
Your CRM gives you pipeline performance art, not pipeline truth
CRMs aren’t capturing most of what your reps actually know about a deal. They’re capturing what your reps are willing to type into a field after a long day, knowing their manager will read it Friday.
The result is filtered, rear-view reporting. Reps log what their managers want to see, not what they’re sensing about the competitor, the buying committee, or the territory. Pipeline data becomes performance art. By the time front-line insight reaches the C-suite, it has been sanitized through three layers and is dated information. Churn signals show up too late to course-correct. Forecasts miss for reasons nobody can quite explain.
Doug Robinson, who recently retired as President of Workday, put it as plainly as I have heard it.
“Before scale, a startup intimately knows its customers, its revenue data, and how to get the sale done. As companies grow, that clarity gets replaced by outdated CRM data, cascading calls, and blurred accountability. The biggest challenge of scaling is preserving the rigor and insight that made the company successful to begin with.”
That clarity is an intelligence access problem, disguised as a CRM problem.
Your engagement survey is asking last quarter’s questions about this morning’s tensions
Surveys are shallow by design. They ask a fixed set of questions and produce a fixed set of scores. They tell you engagement dropped twelve points. They never tell you why.
And yet, it is precisely ‘the why’ that a Chief People Officer needs, in order to act with impact. Without ‘the why’, the survey is a smoke alarm that goes off after the kitchen has already burned. People leaders end up managing culture with a limited view into what’s actually happening and no line of sight into team-level risk. They are running on hollow metrics and missed warning signs, and they know it.
People leaders need two things their current tools cannot give them: the why underneath what is happening, and the ability to turn that why into action employees can see, closing the loop and building trust before the quarter ends.
Your call-recording tool listens, but it never asks
Tools like Gong and Fireflies are good at what they do, which is to record, transcribe, and create themes based on conversations that have already happened. That is useful. But, recordings and themes are not “intelligence”. Ron Johnson, a Savo advisor and a world-class revenue operations executive, said this to me last year and I have not stopped thinking about it.
“Every other tool in the revenue intelligence space passively listens to conversations and attempts to extract meaning after the fact. They analyze wherever the conversation happens to go. Today’s tools have no control or direction over asking the right questions at the right time in the right sequence to create actionable intelligence.”
Passive listening is a constraint, not a capability. If the right question that would have unlocked the deal was never asked, no amount of post-hoc meeting transcription will recover it.
Your “AI insight” tool is summarizing the wrong thing
I'm sure you've done this. I know I have. As an attempt to get better insights, leaders will pipe the input from one of the systems above — survey responses, call transcripts, open-text comments — into a large language model with a prompt that says “tell me what’s important.”
What comes back looks impressive. Fluent. Themed. Bullet points. Executives nod.
But it is, at best, a more articulate version of weak input. And at worst, a confident hallucination dressed in your brand colors. Current AI insight platforms carry three structural problems that no amount of prompt engineering will fix:
- Bias and distortion. AI interpretation often correlates with speaker gender, race, and education, corrupting comparisons across diverse groups.
- Fragility and inconsistency. Re-run the same prompt next week and the conclusions move. The system is reflecting the prompt, not reality.
- No evidence chain. Today’s tools rarely show their work. Confident conclusions get built on thin or absent data, and apparent trends are often measurement artifacts.
A cleaner summary of weak input is still weak input. Faster themes from a poorly designed conversation just help an organization become confidently wrong, at scale. It reminds me of the classic navigator’s problem: a 1-degree error at the start of a 2,000-mile journey puts you roughly 35 miles from where you are meant to be. The error feels invisible at takeoff but it sure hurts when you show up at the wrong location and you've already burned the fuel.
The biggest failure isn’t in the analysis. It’s in the elicitation.
This is the part the AI industry mostly does not want to talk about, so I will.
The first and most consequential failure in conversational intelligence or listening intelligence categories does not happen during analysis. It happens upstream, during the conversation itself. If you do not design the conversation to capture the right signal in the first place, no amount of downstream AI cleanup will recover what the conversation never reached.
A measurement system you can trust defines the constructs it claims to measure — engagement, trust, competency, sentiment, intent, friction — before the first question is asked. It knows what counts as evidence against each construct. It adapts in real time to where the signal is, not where the script said it should be. Most platforms do none of this.
A second failure compounds the first. Most systems use a single AI instance to both run the conversation and judge the quality of what it produced. That is the AI equivalent of a student grading their own homework. It is not a measurement system. It is a false confidence machine.
A real measurement system is built differently: conversation delivery handled by one specialized agent, evidence sufficiency assessed in real time by a separate Signal Monitor, scoring and abstention and drift monitoring handled by specialized downstream components, every insight traceable to the respondent language that supports it. In essence, you need an agentic team that is purpose built to measure without bias or unchecked influence.
Finally, the most important capability in a system like that is the ability to flag when the evidence is thin, i.e. Weak Signal. A system that always produces a score, regardless of evidence, will mislead you most in the cases where honest acknowledgment would have been most useful. In measurement, abstention is a feature. Any vendor who cannot tell you how their system measures is selling you summarization and themes, not trusted insights.
The Business cost of the Narrative Intelligence Gap
Leaders who do not close their Narrative Intelligence Gap pay for it in a compounding way every week, month, quarter, and year. They cannot see it on the income statement because legacy administrative platforms are blind to the missing information gap:
- Misdirected initiatives and investments. Incomplete information paints a skewed picture. The cascading effects show up two quarters later as missed numbers nobody can quite explain.
- Tool bloat. When the gap exists, organizations tend to buy another reporting or automation tool. The result is more dashboards, more integrations, and the same blind spot at the center. None capture Narrative Intelligence.
- Missed forecasts. Leadership commits to numbers built on filtered, biased, rear-view data, then has to explain the variance.
- High-cost consulting backfill. When the gap shows up in a board meeting, the answer is often a six or seven-figure engagement to go ask the questions the system should have asked in the first place.
- Culture erosion. Employees stop answering surveys honestly when nothing changes and survey fatigue sets in. Trust drops. Turnover rises. The signal you most needed was the one your tool was never built to hear.
- Losing to a faster, better-informed competitor. Market shifts arrive as faint signals long before they arrive as revenue.
The Narrative Intelligence Gap exposure is not theoretical. It is in your last forecast variance, your last regrettable turnover report, your last surprise customer churn report, and your last engagement survey that told you nothing new.
What closing the Narrative Intelligence Gap requires
The answer is not “more AI.”
Closing the gap requires a new capability layer – a system designed to actively elicit, measure, and interpret the expansive human knowledge and insights that exist inside your organization, your customers, and your partners.
New capabilities your current stack simply cannot do:
- Actively elicit vs passively record. Asking the right question, at the right time, and adapting in real time to where the signal actually is.
- Produce evidence-anchored insights, not summaries or themes. Every conclusion and recommendation must be traceable back to a specific phrase. If you can’t see the chain, you can’t trust the conclusion.
- Know when to abstain. When the evidence is thin, the system has to say so. A confident answer to an under-supported question is the most dangerous thing an insight platform can give you.
- Enable action to deliver closed loop signal, to insight, to action, to outcomes.
I’d add one practical diagnostic for your next vendor conversation. Ask any vendor whether every conclusion can be traced back to a specific phase, and if it knows if it has adequate evidence to make a claim. Those two questions separate measurement from articulate summarization and give you everything you need to know about what you are getting.
How Savo closes the gap
I’ll keep this section short, because I would rather show you what I’m talking about on a call instead of making you read too much more.
Savo is a Narrative Intelligence Platform. We replace surveys, focus groups, telephone-tag information sourcing, and most consulting engagements with AI-guided conversations that harness what your people think, feel, and experience and transform those narratives into decision-grade insights..
We do this with Savo Signal Events™ which is a custom-built AI-facilitated conversation, that is scientifically grounded, with constructs and behavioral anchors defined before the first question is asked. Our AI Interview Guide facilitates the conversation and intelligently adapts in real time. A separate Signal Monitor watches the conversation as it happens and steers toward depth where signal is thin. Every insight delivered ties back to the specific respondent language that supports it.
Instead of twenty interviews over six weeks from a consulting firm, Savo runs five hundred (or thousands) conversations in parallel and delivers decision-grade intelligence in hours. Instead of telling you engagement dropped twelve points, Savo tells you engagement dropped because the new manager eliminated the team standup that was the primary source of psychological safety for the junior engineers, with the evidence underneath, pulled directly from the conversations.
We do this for research, revenue, people, customers, candidates, frontline workers — anywhere a leader has to act on what a person knows. Anywhere the cost of being confidently wrong is too high. Anywhere where human insight must be surfaced, elevated and prioritized as a system of intelligence.
That's everywhere now. Which is why we built Savo.
The opportunity ahead
The leaders who move first on this won’t be the ones who already have the most data, they will be the leaders who realize that they don’t have the most important data, the data that exists only in the heads of their people. They will be the ones who recognize, faster than their peers, that their problem is an intelligence access problem.
The enterprise platform that matters in this category isn’t defined by how much language it can process. It’s defined by how effectively it can turn human knowledge into decision-ready intelligence leaders can stake real decisions on.
Every organization I’ve ever worked with has thousands of untapped narratives sitting just beneath the surface. The leaders that learn to surface them, structure them, measure them, and act on them are going to be very hard to compete with over the next decade.
Stop guessing. Start knowing.
Onward,
Chuck
Frequently asked questions about the Narrative Intelligence Gap
The Narrative Intelligence Gap is the distance between what people inside an organization actually know — their judgment, context, lived experience, objections, and instincts — and what the organization’s systems can capture and surface for leaders to act on. Roughly eighty to ninety percent of organizational knowledge lives in conversations and never enters a system in a usable form. The gap is between that knowledge and the leadership decisions it should be informing.
Because generative AI inherits whatever input it is given. Layering AI on top of a CRM, survey, or passive call-recording tool produces faster, more fluent summaries of incomplete input. A cleaner summary of weak input is still weak input — only now it carries the visual authority of a finished insight. Speed without signal is exposure, not advantage.
Summarization condenses what was said into themes for a reader. Measurement scores define constructs against behavioral anchors specified before the conversation, traces every conclusion back to specific source text, and reports whether the evidence was actually sufficient. Summarization tells you what was said. Measurement tells you what it means and how confident you should be in that meaning.
CRMs were built to administer pipeline transactions, not to elicit what reps actually know. They capture filtered, rear-view data shaped by team politics and reporting incentives. By the time front-line insight reaches the C-suite, it has been sanitized and is months old. Leadership ends up committing to forecasts built on curated narratives rather than real-time truth.
Surveys are shallow by design. They ask a fixed set of questions and produce a fixed set of scores. They surface that something is wrong, but rarely the why underneath it. People leaders need the why to act, and they need it before the quarter is over. A static instrument cannot give them that, no matter how well the dashboard is designed.
Savo Signal Event™ is an AI-conducted conversation designed to measure specific constructs — engagement, trust, competency, sentiment, intent, friction, and others — against behavioral anchors defined in advance. Signal Events adapt in real time to where the signal actually is, capture every claim as a traceable Evidence Unit, and abstain when the evidence is insufficient. The result is decision-grade intelligence with a full evidence chain underneath every insight.
Six questions to ask: How is the business question defined before the conversation begins? What is being measured against what behavioral anchors? How does the system know when evidence is sufficient? Can every conclusion be traced back to specific respondent language? How is comparability protected across reruns? Does the system abstain when evidence is insufficient, or does it improvise a score? If a vendor can’t answer those clearly, what they are selling is articulate summarization, not measurement.
Gong, Fireflies, and similar tools passively transcribe and theme conversations that have already happened. They do that well. They do not design the conversation, control the question sequence, measure against defined constructs, or abstain when evidence is thin. Narrative Intelligence is the active, measurement-grade version of what passive call-recording tools approximate.
