For 17 years, Tom Cox has been obsessed with a deceptively simple idea: the best human salespeople have already solved the problem of how to sell effectively. The challenge has always been how to capture that expertise and scale it. With Humara, an AI sales agent platform now winning recognition for excellence, Cox and his team have built something that genuinely mirrors how top performing salespeople build rapport, discover customer needs, handle objections, and close deals in one of the most complex product categories imaginable: telecommunications.
Why Telecom Is the Perfect Proving Ground
Telecommunications products have become extraordinarily difficult to navigate. Providers want to be your source for everything, including mobile, broadband, TV, and increasingly your smartwatch and connected home devices. The industry has pushed aggressively toward converged offerings and multi-year contracts, which creates genuine confusion for customers trying to compare options.
The simplicity that once defined telecom decisions has essentially vanished. Cox points out that pretty much everything is unlimited now, whether that means data, minutes, or home internet speeds that far exceed what most households actually need. When every plan has unlimited in the name, customers struggle to understand the real differences between unlimited max, unlimited pro, and unlimited plus. The actual distinctions now involve throttling policies, perks, benefits, and other factors that require careful explanation.
According to J.D. Power research, customer satisfaction with telecom providers drops significantly when customers feel confused about their options or surprised by their bills (https://www.jdpower.com/business/telecommunications). This friction creates massive opportunity for solutions that can actually guide customers through the complexity. Humara has found its sweet spot precisely because telecom presents such a difficult challenge, reasoning that depth of capability comes from specialization rather than trying to play everywhere.
Building AI That Sells Like Elite Humans
The training approach behind Humara reveals why the platform performs so differently from typical chatbots. Cox describes four major sources of training data that create what he calls sales intelligence. First, Humara has accumulated millions of sales interactions from 15 years of guided sales products deployed at global telecom companies. Humara’s previous generation of products used machine learning and deterministic flows, but addressed the same fundamental problem.
The second source proved even more revealing. Humara hired a team of top percentile telecom salespeople from both the UK and US markets and had them simulate thousands of conversations with end users who believed they were talking to AI. This inversion mattered enormously because it showed how elite salespeople succeed even when customers give short answers and skip pleasantries. The agents had to humanize conversations and build rapport without any help from the other side.
The third layer came from working with consumer psychologists to analyze the psychological behaviors present in those transcripts. This helped Humara understand which behaviors equated to certain techniques like anchoring, active listening, or mirroring. This enabled the AI to understand how and when to deploy which techniques.
The SPIN Framework and Beyond
One methodology that emerged from analyzing top performers was SPIN selling, which stands for Situation, Problem, Impact, and Need. The best salespeople consistently started by understanding a customer's current situation, then identified the specific problem they were trying to solve and the impact it was having on their life. Only after establishing that foundation would they discuss needs and recommend products. This sequencing matters because it creates a foundation for personalized recommendations rather than generic product pitches.
The framework extends well beyond a single technique. Humara breaks the sales conversation into distinct stages including rapport building, discovery, product positioning, objection handling, and closing. Each stage has its own specialist capabilities and techniques. The objection handling loop proves particularly interesting because good human salespeople always try to identify what is stopping someone from buying today and then systematically address each concern until no reasons remain not to purchase.
The platform adds two additional layers beyond the core sales framework. Intent detection recognizes that two customers looking at the same bundle might care about completely different things, perhaps one prioritizes TV quality while another focuses entirely on price. Persona detection adapts to whether someone wants detailed comparisons or just a quick answer. The combination creates experiences that match both what customers want and how they want to receive information.
Results That Challenge Assumptions
The performance metrics coming from Humara deployments have genuinely surprised even the team building it. Conversion rates reach up to 23 percent, and customers add 24 percent more products on average compared to standard website flows. Cox attributes these results to what he calls breadth and depth. The breadth comes from identifying and resolving friction at any stage of the buy flow, from initial research through credit check and installation scheduling. Telecom purchases involve high confidence thresholds because customers sign contracts for multiple years and truly do not want to get the decision wrong.
The depth comes from specialization in the art of persuasion. Because Humara focuses exclusively on sales within the telecom vertical, it can build capabilities that generalist platforms cannot match. The platform currently resolves about 92 percent of sales conversations without requiring human handoff.
Perhaps most surprising has been how customers interact with the AI. Cox expected users to behave differently than they would with humans, giving terse answers and skipping pleasantries. Instead, customers write long paragraphs explaining exactly what they want, include please and thank you throughout, and even ask whether the AI will get commission if they come back to purchase. The human instinct to be polite persists even when people know they are talking to a machine.
Learning That Flows Both Ways
An interesting loop has developed between Humara and the human sales teams at its customers. Telecom companies initially want their best salespeople interviewed and incorporated into the platform so that their specific approaches get reflected in the AI. Since going live, Humara has gathered real-world data from thousands of users, which it uses to continuously improve its performance through reinforcement learning. As Humara improves its capabilities,those companies then ask how they can take the learnings back and train their human agents to follow the same techniques.
The platform also generates actionable insights for product and marketing teams. Humara can now report the top five reasons customers abandon purchases at specific points in the buy flow, whether that involves credit check hesitation, confusion about installation, or concerns about contract terms. These insights enable companies to address friction in how they position products and present information, potentially eliminating the need for intervention in the first place.
Cox notes that this creates a somewhat paradoxical goal where the insights should eventually do Humara out of a job by helping companies remove friction before it occurs. But the reality of converging telecom products suggests complexity will continue increasing faster than companies can simplify their offerings.
Enjoying insights from industry leaders? Subscribe to The Winners' Circle podcast on your favorite podcast player and never miss an episode. Listen and subscribe at bintelligence.com/podcast.