- You are probably hearing about a fifth of the truth. Surveys only capture what a customer chose to give you. The QBR transcript, the support thread that escalated, the email nobody analyzed, the G2 review left at midnight: that is the 80 to 90% your team never reads. Signal Intelligence finally reads it.
- An email, a QBR, and a support ticket can all become a clean NPS response. The team demoed all three live, dropping raw text into Gaige and getting back a score, a summary, drivers, and a follow-up ready to send.
- Every signal ties to an account and a dollar figure. Fewer than 10% of businesses successfully connect CX to revenue. Signal Intelligence was built so every piece of text links to an account, its revenue, and a next action.
- Accuracy took six months to get right. The system trained on millions of real NPS comments and scores, then calibrates per customer and segment to stay honest over time.
- This is not the end of surveys. It is the end of the silent account. When you are already sitting on thousands of conversations, you may not always need to send a survey to know the score.
Where Signal Intelligence Came From?
To understand why this launch matters, it helps to rewind to 2019, the year CustomerGauge coined the term Account Experience. The idea was to take account management and marry it to customer experience, built for the way B2B actually works. Not one happy contact and a single score, but multiple stakeholders inside one account, real revenue riding on the relationship, and the need to report across divisions, locations, and segments. As Adam put it, it is account management married with customer experience, and it is something B2C simply does not have to think about.
That foundation has been close to 20 years in the making. Millions of data points, a stack of benchmarks, and what the company calls a success playbook for turning all of it into return on investment. The serious AI work kicked off with the Gaige launch in September 2025. This webinar was the next chapter: agents joining the lineup, and the arrival of an ecosystem built to listen, understand, prioritize, and act at account scale.
What Are the CustomerGauge AI Agents?
There are currently around six agents. Here is the quick tour.
- Analytics Agent. A full LLM living inside CustomerGauge, so you can ask questions about your program in plain English and get graphs and answers back.
- Loop Agent. It drafts the responses that help you actually close the loop.
- Summary Agent. Hands you the insights you need before you walk into a QBR.
- Interview Agent. A natural-voice bot that sits on top of a normal survey, talks to your customer like a person, and gives you both the transcript and the audio afterward.
- Signal Intelligence Agents. The newest arrival, and the reason everyone showed up.
Signal Intelligence takes that transcript, or any other unstructured text you have lying around, and turns it into something you can track and act on: themes, tags, churn and renewal signals, and a real NPS score tied to the account, the revenue, and a trigger.
What Are the Two Kinds of Feedback?
The clearest idea in the whole session was also the simplest. There are two kinds of feedback, and most B2B programs are wired to hear only one of them.
Explicit Feedback: is the feedback your customer chose to give you. The survey response. The NPS score and the comment underneath it. The interview you scheduled. Anything that started because you asked a question first. Adam was careful here: this is not going anywhere, because an explicit ask gets you an explicit answer, and there will always be a place for that. But be honest about the size of it. Explicit feedback is maybe 10 to 20% of the truth sitting in your business.
Signal: is everything your customer tells you anyway, whether you asked or not. It is in the emails, the tickets, the reviews left in public. It is the other 80 to 90%, the part almost every program walks straight past. Dark matter, as the team likes to call it.
"Imagine what it's going to look like when you can put all these together..."
How Did CustomerGauge Make Signal Intelligence Accurate?
Two principles shaped the build, and they pulled in slightly different directions.
The first was action and monetization, and it came first for a reason. Anything Signal Intelligence produced had to slot cleanly into Account Experience. The text had to attach to accounts, link to revenue, and run through the existing close-the-loop framework. Skip that, and you have built a sentiment toy, something that feels clever in a demo and changes nothing on Monday. This is where that sobering number landed, fewer than 10% of businesses successfully tie CX to revenue. CustomerGauge did not want to add another tool to the pile that ignores the money.
The second principle was accuracy, and on this one there was no wiggle room. Customer comments are gloriously messy. There is the random noise of mood and fatigue and the accidental misclick. There is systematic bias, the cultural habit in some markets of never, ever giving a 10. And there are mixed signals, because one score almost never tells the whole story. Throw false positives and false negatives at people and you do real damage. As Adam warned, you lose the trust of your own team, and once that goes, the whole program goes with it.
The fix came down to two things: scale and calibration. Because the system trained on millions of real comments and scores, it starts accurate rather than guessing. Then it keeps learning from each customer and each segment. Japan came up directly here, where NPS routinely runs lower than in the West, exactly the kind of skew the calibration exists to handle. And it reaches well past NPS too, pulling intention to buy, innovation ideas, themes, and more out of the same text.
Signal Intelligence in Action
The first the team dropped in a real email thread between a product person and a frustrated customer, stretched across several days. Trishaala pasted it into Gaige and asked it to read the signal as NPS. Seconds later, back came a score of 5, a tidy summary, and the drivers behind it. Then it kept going, pulling the account manager, account name, and job title straight from the thread, spinning up a working response engagement, and color-coding the tags. In 50 seconds the case was assigned, an alert had fired, and Gaige had written a follow-up ready to send. The point Camilla kept circling back to: this layers on top of your explicit feedback, it ties straight back to the account, and carries the dollar value with it.
The second part of the demo went bigger. Instead of one record, Trishaala bulk-uploaded around 20 QBR transcripts, the kind you export from Gong or Teams, each one 25 minutes to an hour long. Because you are handing this to an agent that "does not have fatigue," they stretched the question model well past a basic NPS comment to pull renewal intent, churn signals, and any mention of ROI. One transcript came back as a 7, with a realistic mix of tags and a read on how likely the renewal was. Camilla connected it straight to the money: promoters are two to three times more likely to renew, grow, or buy more, so finding one buried in a QBR is not trivial, it is pipeline.
The third part of the demo pushed furthest, into support, where the volume is brutal and each ticket is a back-and-forth before anything gets resolved. Here, the signal comes straight out of the conversation that already happened, no survey required. Running almost 31 questions, Gaige scored each engagement, on things like agent empathy and responsiveness, surfaced root causes, and flagged where the sentiment shifted mid-conversation.
Why Use NPS for Signal Intelligence?
It would have been easy to skip NPS and just do generic sentiment. The team chose not to, and Trishaala gave three reasons that are hard to argue with. NPS is a leading indicator, so you can tie it back to revenue and put a number on it. It benchmarks against the rest of your feedback, so it never sits off on its own. And it is action-driven, which plugs it straight into the close-the-loop machinery. Use CSAT or CES too if you like, the platform will happily synthesize all of it, but NPS stays the connective tissue.
Where AI Fits, and Where Humans Still Do
CustomerGauge has been chasing this exact problem for years, through earlier swings like Account Vitals and first-generation text analytics. The vision was always there. The technology, frankly, was not.
That is the whole reason this lands now. The signals were never missing. They were scattered across every conversation a company has with its customers, just out of reach. What changed is that you can finally gather them from the places that used to be too hard or flat-out impossible, and drop them into one view next to revenue and actions.
But notice what Signal Intelligence is actually for. It does not replace the human in the relationship, it adds value to it. The agent does the listening, and analyzing. The human still does the part that matters most, the actual conversation with the customer. And none of it floats free of your control. You decide what Gaige goes looking for, you can change those instructions whenever you want, and it follows the same data protections you already signed up for.
Is This the End of Surveys?
You can probably guess the question that came up, so let's answer it plainly. No, this is not the end of surveys. Surveys are still the explicit signal, the direct ask that earns a direct answer, and that is genuinely useful. What might change is the volume. When you are already holding a mountain of conversations, some of those survey sends start to feel optional.
Signal Intelligence is about getting more insight out of more of your customers, more of your book of business, and turning that insight into something you actually do. Your customers have been telling you their NPS for years.
The only real question is whether your program was ever built to hear it.
Get Started With CustomerGauge
CustomerGauge is the B2B experience platform built for revenue-connected NPS, account-level feedback, and close-the-loop programs that actually move retention. If you want to see what Signal Intelligence would surface inside your own program, the team is happy to show you.
About the Author
