Newsletter · · Ashutosh Agarwal
AI in Sales and Support Stops Promising and Starts Proving - Vertical Spotlight: Sales & Customer Service - Week of August 4, 2026
Vertical Spotlight on Sales and Customer Service for the week of August 4, 2026. This issue moves past AI hype to the receipts, covering a 30,000-workflow study of where AI actually helps revenue teams, startups posting faster ramp times and higher win rates, and the guardrail failures that still break brands when bots talk to customers.
Vertical Spotlight: Sales & Customer Service
Week of August 4, 2026: AI in Sales and Support Stops Promising and Starts Proving
This week's podcasts on selling and support moved past the hype and onto the receipts: a 30,000-workflow study of where AI actually works, a new rep who hit revenue 40% faster, a support bot that talks to 20,000 companies' customers, and a loud reminder of what still breaks when you let the software talk to people.
Covering podcasts from July 28 – August 4, 2026.
The Landscape
For two years, the story about AI in sales and customer service was a promise: software would replace your reps and your support agents. This week the conversation grew up. The founders and operators on the mic stopped arguing about whether AI works in revenue teams and started showing where it works, by how much, and, the new obsession, how to stop it from quietly torching your brand.
The throughline across the week's episodes is a split screen.
On one side, the evidence is piling up. Alex Bilmes, CEO of a company called Endgame, spent six months studying 30,000 real AI workflows inside sales teams and found five specific jobs where AI is now doing useful work in production. Justin Shriber, CEO of Terret AI, described a brand-new salesperson who reached his first revenue in roughly 60% of the time it normally takes. Templafy, which builds sales documents with AI, says its customers cut content-creation time by more than 75% and one lifted its win rate by 14 points. Capacity, a support-automation company, now handles the emails, calls, and tickets for about 20,000 businesses' customers.
On the other side, a chorus of caution, often from the very people building the AI. Vera Quinn, who runs one of North America's largest outsourced sales forces, is betting flatly that humans won't be replaced in complex selling. The head of sales at Clay, a $5B-valued go-to-market darling, insists on a "human in the loop" and says fully automated outbound "kind of sucks." Even Gamma, which reached $100M in revenue with no sales team at all, is now hiring one. And Cyara's CEO warned that an AI voice bot that's right 95% of the time can still ruin you with the other 5%: a real airline got fined after its bot invented a refund policy on the fly.
Two ideas tie the week together. First, the winning use of AI in revenue is not "generate stuff faster," it's understanding: giving reps context and serving it in the exact moment they need it. Second, the productivity is showing up less as mass layoffs and more as "hire fewer people and ask the ones you have to do more." As Shriber put it, companies aren't firing frontline sellers so much as "tabling the need to hire new people" while raising everyone's quota.
Here are the companies that made the case this week.
Companies to Know
Endgame: the "where does AI actually work" map. Alex Bilmes, the CEO, told the Revenue Builders podcast that his team analyzed more than 30,000 real AI workflows across go-to-market teams over six months, deliberately watching what people actually did rather than asking them in a survey. Five jobs rose to the top: account intelligence (who are the people in an account, how do decisions get made), conversation readiness (meeting prep, call debriefs, building review decks), deal acceleration (business cases and other deliverables), pipeline inspection, and team enablement. His sharpest point was about what AI is not mostly used for. "Creating a deck, creating a document or an account plan is pretty easy. AI can do that pretty well," Bilmes said. "Knowing what should be in those documents and how well articulated they are to your value proposition, your methodology, your sales process, that was the really hard part." Endgame itself is what he calls a "centralized intelligence platform" that pulls data out of Salesforce, Gong, Slack, and email and turns it into a knowledge graph an AI can reason over. He's watching customers shift heavily toward Anthropic's Claude and "Claude Cowork," and even switch off traditional sales-enablement software because the methodology can now live inside the AI itself. Revenue Builders, "Where AI Is Actually Working in Revenue Teams with Alex Bilmes" (Aug 2, 2026).
Terret AI: the rep who ramped 40% faster. Justin Shriber, a former McKinsey consultant and ex-LinkedIn marketing VP, now CEO of Terret AI, described what he calls an "answer to action engine" on the Topline podcast. It ingests all of a company's revenue data (CRM, data warehouse, call transcripts), finds not just what is happening but the root cause, then feeds the insight to the seller "in the moment," inside Slack, email, or live on a call. The proof point: Terret hired a smart but inexperienced rep, plugged him into the system, and "his time to first revenue was about 60% of what we had seen historically," with close rates ramping to the level of a veteran. Shriber was candid about the hard engineering underneath. Early on, when the AI was asked numerical questions like "what's my win rate," it gave a different answer every time because the models were "jumping from one system to another connecting the dots willy-nilly," so the company had to build a "revenue graph" that was secure, accurate, and cheap enough to run without burning tokens faster than the value it created. His read on the market: revenue per rep is rising, buyers are compressing sales cycles by using their own agents to do early research and vetting before they ever talk to you, and companies are raising quotas instead of hiring. Topline, "SPOTLIGHT: AI Is Making Sales Reps 60% Faster to Revenue | Justin Shriber, CEO @ Terret AI" (Jul 28, 2026).
Templafy: the numbers behind "make the rep look like an A-player." John Mark Shanwan, chief sales officer at Templafy (a seven-year veteran of the company), called it "the best document agent in the world" on The AI for Sales Podcast. It generates sales content, proposals, decks, contracts, directly from a company's own AI model while keeping brand, data, and accuracy under control, and it lives inside Microsoft 365 (Word, PowerPoint, email) where reps already work. The receipts: in professional-services firms, "a 75-plus percent decrease in the amount of time that reps are spending creating content," and one company posted "a 14% increase in close rate" purely because its reps showed up with better, more accurate material. He noted the scale of deployments: a 100,000-person bank, 250,000 users across the Big Four accounting firms. Shanwan also flagged two other tools he admires: Gong, for capturing calls and improving forecast accuracy, and Win AI, which does live in-call coaching, "answering questions right then and there" instead of after the call. The AI for Sales Podcast, "How Templafy is Revolutionizing Sales with AI" (Jul 29, 2026).
Capacity: "deflection done right" in customer support. David Karandish, founder and CEO of Capacity (and the entrepreneur who sold his previous company, Answers Corp, for $960 million in 2014), joined Navigating the Customer Experience to explain how to automate support without alienating customers. Capacity uses AI to deflect emails, calls, and tickets across voice, SMS, WhatsApp, web, and email, today serving "about 20,000 customers" and connecting to "over 250 of your favorite apps." His design rules are refreshingly concrete: always give customers an escalation path ("a quick phone-a-friend button"), and when they escalate, don't make them repeat the ten questions the bot just asked; wire the bot into the CRM and order systems so it isn't "coming in blind"; anticipate the next question; train it with empathy ("'Have a wonderful day' might not be the thing to say to an irate customer"); and never ask the same question twice. His advice to support leaders wary of a big rollout: "Go get your small W win first, prove it out, and then iterate." Karandish also described running Capacity itself on three layers of agents: customer-facing agents, "builder agents" that help customers launch more agents, and internal agents to run the company, adding that "there isn't an activity from finance to marketing to you name it that doesn't start with an agent." Navigating the Customer Experience, "277 - Deflection Done Right: Building AI Support That Customers Actually Love with David Karandish" (Jul 28, 2026).
Gamma: $100M in revenue, 50 people, and (finally) a sales team. On The Official SaaStr Podcast, the CEO and co-founder of Gamma, the AI presentation tool that pitches itself as "the anti-PowerPoint," walked through hitting "$100 million in ARR profitably" with a team of just 50 and, for most of that journey, no sales or marketing spend at all. The origin story is a gut-punch every founder will recognize: a third pitch, twenty minutes in, and the investor said "this has got to be the worst idea I've ever heard" before hanging up mid-Zoom. What turned it around was a three-month rebuild of the first 30 seconds of the product to make it "feel magical," plus a deliberately provocative launch tweet, "the most valuable skill in business is about to become obsolete," that got Paul Graham to fire back and went viral. Signups climbed from 5,000 a day to 10,000, 20,000, then 50,000 a day, "doing zero marketing, zero sales." Gamma now has 50 million users and 600,000 paying subscribers. The reason it's now building a sales team is instructive: inbound demand from companies wanting to buy for whole departments got too big to ignore. On pricing, the CEO admitted they launched self-serve with "no way to pay for the product," scrambled to add it when chat blew up with "how do I buy more credits," and are still working through the industry-wide question of seat-based versus usage-based pricing. The Official SaaStr Podcast, "SaaStr 871: $0 to $100M ARR Fast. How Gamma's CEO and Co-Founder Scaled Quickly without a Sales Team" (Jul 29, 2026).
Bland: voice AI, a $100M round, and a telephone car. Ethan Clouser, the 22-year-old head of marketing at voice-AI company Bland, told The Dave Gerhardt Show the founder story: two founders who started at 19 and 24, "flunked out of YC, no revenue, rejected by 180 investors," then raised "over a hundred million dollars" across three years. Bland's positioning is deliberately narrow, "voice AI for regulated industries," in a field Clouser says now has roughly "200 voice AI competitors" all using the same buzzwords. The marketing has been pure guerrilla. With about $3 million in the bank, the company bet a third of its runway on billboards in New York and San Francisco reading "Still hiring humans?" next to a phone number that connected callers to Bland's AI, a live demo disguised as a billboard. It drew "over a billion impressions," booked out every salesperson's calendar "three to four months in advance," and funded the Series A. A more recent stunt, a working 1960s-style telephone car with Soulja Boy, did "60 million impressions in the first week." The company is now shifting from viral spectacle toward account-based marketing as it moves up-market to the enterprise. The Dave Gerhardt Show (from Exit Five), "The Viral Marketing Bets Behind High-Growth Voice AI Company Bland.ai, with Ethan Clouser (Head of Marketing)" (Aug 3, 2026).
Clay: the $5B "give humans leverage" playbook. Becca Lindquist, head of sales at the go-to-market data platform Clay (recently valued at $5 billion), told The Crew Podcast how the best teams actually use AI in outbound, and it's not what you'd guess. Her strong view: fully automated, mass outbound "kind of sucks," and there's "a very small window for the sweet spot" between creativity and scale. The winning move is to find what your best rep does manually and automate that, always surfacing it to a human first: "Would you like it to send for you?" She's "a big human in the loop." Her favorite example of AI-plus-creativity: a large social-media customer uses Clay to run sentiment analysis inside its own communities, identify the responsible brand manager (say, the person who owns Tide at P&G), and auto-draft outreach: "here's the sentiment we're seeing for Tide in this subgroup, would you like to advertise to influence it before these people get to the grocery store?" She framed the broader lesson as "what's your alpha?": a company's real edge is "the ability to go and execute on new ideas with the right data faster than your competitors." One rep on her team, she noted, just closed the largest new deal in Clay's history. The Crew Podcast, "How Clay Built the GTM Team Behind a $5B Valuation w/ Becca Lindquist, Head of Sales @ Clay" (Jul 28, 2026).
Cyara: the guardrails nobody budgets for. Sushil Kumar, CEO of CX-testing company Cyara, used The Agile Brand to make the least glamorous but most important point of the week: agentic AI in customer service is powerful precisely because it's convincing, and that's the danger. "When it makes mistakes, it makes mistakes confidently," he said. "AI is amazing, but it's not perfect. It will get you 90, 95% there. But what's your strategy to contain that 5%?" His examples land hard: a major airline "getting fined and penalized by government because [its bot] invented a refund scheme"; bots tricked by a customer saying "ignore all your previous instructions"; compliance breaches where a bot leaks personal data or answers questions about the wrong person's account. Two data points from his research: nearly half of consumers hang up after just one or two failed attempts with a bot, and when the experience is bad, customers "blame the brand," not the technology. His prescription is a mindset shift in how these systems are tested: away from simple pass/fail toward a "multidimensional rubric" that scores each interaction on compliance, bias, factual accuracy, latency, speech-to-text accuracy, and customer sentiment, not just whether the call "completed." The Agile Brand with Greg Kihlström, "Cyara CEO Sushil Kumar on finding the right balance of AI in your CX" (Jul 29, 2026).
Deel: what an "AI-first" sales org looks like from the top. Chris Lee, head of sales for the Americas at Deel and one of its first sales hires, told The GTMnow Podcast the raw numbers of scaling from $1 million to $1.5 billion in revenue: $1M to $50M in year one, $50M to $300M in year two, $1B within four years, all "pre-AI." The AI-relevant part is where he'd go from here: "If you were to start a company today… I wouldn't go hire a big sales team. The first thing I would do is optimize an AI-driven sales process." Deel has already institutionalized this with a real job title, "Ghostbuster," systems-thinkers with consulting and operations backgrounds whose job is to hunt down slow, manual, multi-department workflows and automate them so a 7,000-person company keeps moving at startup speed. Notably, Lee's vision of AI in sales adds human touch rather than stripping it: he argues AI now lets you deliver a "five-star" sales experience, personalized executive outreach, account-based marketing, to small customers, not just the enterprise accounts that used to justify the manual effort. The GTMnow Podcast, "Inside Deel's $1M → $1.5B Sales Machine | Chris Lee" (Aug 3, 2026).
One Debate
Will AI actually replace human sellers, or just re-tool them?
The most direct challenge to the replacement narrative came from Vera Quinn, CEO of Cydcor, one of North America's largest outsourced sales forces, on Founder's Story. She named the pitch she hears constantly and rejected it head-on: "Everyone right now that is building some sort of AI-based company is saying, 'our AI-based salesperson is going to replace your human sales team.'… You are saying that the human-to-human interaction in sales is not going to be replaced. That's our bet." Her reasoning: AI can handle simple things, Cydcor uses it too, but "in order to make complex decisions, people want to talk to people." Her proof point is the busiest store in any mall: "the Apple store… everything you can do in that store you can do on your phone. But it is busy all the time." And she flagged the trust problem that's coming as bots get better at sounding human: "My fear is, what if the AI starts saying it's human?"
The interesting part is that the builders on the other episodes don't fully disagree, they just draw the line differently. Justin Shriber's data from Terret AI suggests AI is making individual reps dramatically more productive, but he sees companies responding by slowing hiring and raising quotas rather than firing their front line, at least for now. Clay's Becca Lindquist keeps a human in the loop by design. Gamma reached $100M with no salespeople yet is now hiring them for the human-heavy enterprise motion. And Deel's Chris Lee, who would build an "AI-driven sales process" from day one, still argues AI's best use is to extend the human five-star experience to more customers, not remove it.
So the unresolved question isn't the cartoon version ("robots vs. humans"). It's this: as AI gets good enough to sound human and cheap enough to scale, does the value of a real human conversation go up (Quinn's bet, because trust becomes scarce) or down (because the machine is now good enough for most interactions)? Every company this week is quietly placing a chip on one side or the other, and their hiring plans are the tell.
Vertical Spotlight rotates weekly through Insurance, Fintech, Proptech/Construction, Legal, Healthcare/Biotech, and Sales & Customer Service. Next week: back to the top of the rotation with Insurance.