Apr 4, 2025 · 1h 9m · 20sales
20Sales: How the Best Sales Teams Use AI to Win Enterprise Deals | Sales Teams Will Be Dramatically Smaller | How to Ramps Sales Reps Way Faster | Why Unpaid Design Partners are BS | Why this Generation of Sales is Soft with Ishan Mukherjee @ Rox
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20 Sales, host Harry Stebbings interviews Ishan Mukherjee, co-founder and CEO of Rox, to discuss how artificial intelligence, disciplined go-to-market execution, and shifting enterprise buyer expectations are transforming modern sales teams. Ishan shares actionable strategies for scaling revenue, managing sales talent, and designing AI software that drives measurable business outcomes.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 27.1% of the talking time here. How this is scored →
Harry as informed peer 4.0 Guest teaching 5.5 Guest disagreement 1.7 Harry pushing back 3.3
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 12:51 Unpaid design partners are BS
Ishan strongly rejects the standard startup practice of free design partners, insisting that customers must pay cash to ensure true skin in the game.
Hardest push from Harry ▶ 18:02 Challenging single-rep consumption and commit ownershipHarry forcefully challenges Ishan on whether one account executive can realistically close mega enterprise deals while simultaneously driving wall-to-wall adoption and consumption.
Biggest teaching moment ▶ 5:00 Bits vs atoms in PLG and enterprise distributionIshan reframes the PLG vs enterprise debate by explaining that software companies must choose between investing in bits (product) or atoms (humans), requiring a unified leader to manage the funnel.
Harry holds his own ▶ 37:47 Personal LinkedIn triggers outclassing corporate updatesHarry demonstrates deep product and sales domain expertise by identifying personal LinkedIn posts as higher-converting outreach triggers than earnings reports, prompting Ishan to offer him the Head of Product role.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Episode Preview: The Future of Sales Teams and AI Integration | 0 | 0 | 0 | 0 | Introductory preview clips and sponsor ad reads for CapChase, ZoomInfo, and Gong. There is no active dialogue or power dynamic to evaluate. | |
| Scaling New Relic's PLG Motion from Zero to $100M ARR | 3 | 6 | 1 | 2 | Ishan details scaling New Relic's PLG model by focusing on developers and explains the structural difficulty of pursuing PLG and enterprise simultaneously. Harry asks targeted clarifying questions on C-suite alignment and sales roles. | |
| Tactical Channel Optimization: SEO, Performance Marketing, and Event Dinners | 5 | 5 | 1 | 3 | Ishan outlines shutting down event booths at New Relic to focus on SEM and SEO. Harry brings up Ramp's holistic event strategy, and Ishan agrees that post-event dinners beat conference booths. | |
| Search Engine Economics: The Reality of SEO and SEM ROI | 4 | 6 | 2 | 2 | Ishan explains the economics of SEM bidding wars, comparing it to an opaque market where ROI is rare outside narrow intent keywords. Harry draws a parallel to venture capital competition. | |
| Why Unpaid Design Partners Are BS versus Paid Pilots | 4 | 7 | 3 | 4 | Ishan forcefully dismisses unpaid design partners as useless and explains why Datadog beat New Relic through go-to-market consistency. Harry questions conventional wisdom around being plastic and reactive with pricing. | |
| The Evolution to Usage-Based Revenue Models and AI Power Laws | 5 | 6 | 3 | 6 | Ishan details the shift from commit to usage-based models and cites Frank Slootman's model of holding AEs accountable for consumption. Harry pushes back on whether a single rep can manage both closing and expansion on massive deals. | |
| Designing Early-Stage Sales Compensation and Identifying Outlier Sellers | 4 | 6 | 2 | 5 | Ishan discusses early-stage sales comp and finding outlier sellers. Harry challenges whether power-law compensation structures create toxic two-tiered cultures inside sales teams. | |
| Compressing Sales Rep Ramp Times Through AI Deep Research | 4 | 6 | 2 | 3 | Ishan explains how AI deep research compresses sales rep ramp times from two quarters to one. Harry prompts Ishan on whether the current generation of sales reps is too soft for tougher market realities. | |
| Evaluating Candidate Quality and Rising Enterprise Consumer Expectations | 3 | 6 | 2 | 4 | Ishan explains candidate evaluation and how personal AI usage elevates enterprise buyer expectations. Harry pushes back on whether enterprise buyers truly expect high-quality software given their baseline legacy tools. | |
| Unpacking the ROX Architecture: Real-Time Context and AI Agent Swarms | 4 | 6 | 1 | 2 | Ishan explains ROX's architecture, indexing CRM, warehouse, and public data to power seller agent swarms. Harry jokes about his investment style while seeking clarity on VC use cases. | |
| Macro Spending Shifts and ROX's 30-Day Value Verification Model | 4 | 6 | 2 | 3 | Ishan explains macro spending shifts from headcounts to software ROI, outlining ROX's 30-day Palantir-style value verification model. Harry asks why users fail to become daily active users. | |
| UX Evolution in AI Software: Replacing Chat Interfaces with Workflow UI | 6 | 5 | 1 | 2 | Harry offers practical product feedback that personal LinkedIn posts drive the highest outreach engagement, prompting Ishan to offer him the Head of Product role. Ishan then outlines transitioning from chat UI to workflow UI. | |
| Navigating Enterprise AI Churn Risks and Non-Deterministic Trust | 5 | 6 | 2 | 5 | Harry asks if enterprise AI tools face unprecedented churn due to non-deterministic output. Ishan agrees and predicts sales orgs will shrink by 80-90%, retaining only high-touch relationship builders. | |
| Data Infrastructure: AI Agent Swarms versus Legacy Providers | 5 | 6 | 2 | 4 | Harry directly asks if AI agent swarms kill legacy data providers like ZoomInfo. Ishan clarifies that public web research is free game, but walled-garden contact data still requires verified providers. | |
| Mastering the Transition from Founder-Led Sales to Rep-Led Sales | 3 | 6 | 1 | 2 | Ishan details transitioning from founder-led to rep-led sales, defining a true repeatable motion as list-building and first-meeting conversion without founder help. | |
| Ishan's 90-Minute Rep Interview Framework and Explainable Compensation | 3 | 6 | 1 | 2 | Ishan shares his 90-minute candidate interview framework and discusses compensation explainability to prevent internal organizational discord. | |
| Managing Trajectory Expectations: T2D3 Realities versus Enterprise Value | 5 | 5 | 2 | 5 | Harry challenges modern revenue expectations like T2D3, questioning if AI hype has set unrealistic benchmarks for founders. Ishan details focusing on 45-day land cycles and long-term sticky enterprise value. | |
| Capital Discipline, Preemptive Rounds, and Founding Team Lessons | 5 | 6 | 3 | 6 | Harry pushes back on taking preemptive venture rounds, arguing they destroy discipline. Ishan defends taking capital while maintaining $47M in the bank, and admits early mistakes in hiring AI researchers over applied product engineers. | |
| Quick-Fire Round: Sales Tactics, Perplexity, Quota Changes, and Exemplary Orgs | 4 | 5 | 1 | 2 | In the quick-fire round, Ishan calls automated emails dead, defends Perplexity's survival, and predicts sales OPEX will fall from 40% to 20% as reps manage larger quota books. |
Statements from this episode (52)
Mukherjee: Future sales teams will shrink to 10-20% of current sizes
“Dramatically smaller. It could be 10 to 20% of current kind of org sizes.”
Mukherjee: Top 10% of salespeople generate 90% of revenue
“Sales is a power law. The top 10% bring in 90% of the revenue.”
Mukherjee: Current salespeople are unprepared after four easy years
“Yes. We had it relatively easy the past four to five years, and I don't want to be morbid, but”
Mukherjee: Scale PLG by divesting everything except where you are the default
“So we basically shut everything down, reallocated investments to really go after them and earn the right for them to use the product again. And that's what the whole machine was built on. Divest everything else, invest where we are default.”
Mukherjee: Do not combine PLG and enterprise sales without aligned C-suite leadership
“It's only possible if you have a dialed in CEO Or kind of a C-suite who really gets it. If not, do not do it. You need a singular person who understands the emotion. If not, I wouldn't do it.”
Mukherjee: Amit Agarwal is the key architect of Datadog's product-led growth engine
“Not a lot of people know about him, but he's kind of the brains behind the product-led distribution engine at Datadog.”
Mukherjee: Enterprise buyers will always require human sales reps, not just self-serve software
“Fully believe that you need humans because customers want their problem to be solved. They don't give a F about products or services. They want their pain to be solved, and they want to build a non-transactional relationship with these vendors.”
Mukherjee shut down event marketing at New Relic due to poor ROI
“If you look at the data, organic content, and then performance marketing worked really well for New Relic. New Relic was the original current brand that did land and expand. Everything else, events, DevRel, other pieces, you do it because people want you to do…”
Mukherjee: Free tier software is just a better marketing website, not conversion
“People treat a free tier product as a converted customer. They're not converted to customers. Free tier product is just a better website. Like people come in, instead of browsing the website and marketing, they're just tinkering with the product. So just move …”
Mukherjee: Post-conference dinners yield higher ROI than sponsoring booths
“The dinner after the event around the conference is a lot better than the conference itself. Right. So like the best thing to do is take your dollars. We still did the events. Just don't do the booth. Don't do like the one of a hundred sponsors. Just have an a…”
Stebbings: Guillermo Rauch saw Vercel AI inbound grow 1% to 4%
“I've seen also Guillermo Rauch at Vassell's mentioned it's moved from like one to four percent of their kind of inbound.”
Mukherjee: SEM is a complete cartel and a messy market
“SEM, right, you want us to be proactive, like I came in first principles, I think it's a complete cartel. So if you really look at how you invest capital into the SEM game, you're just kind of selling your soul to a pretty messy market.”
Mukherjee: New Relic found SEM ROI on only 10 to 20 keywords
“But for us at New Relic, we had figured out there was 10 to 20 keywords where we were really good and we could put customers on a stitched path. They could see an ad, get to a landing page, use the product, get to an aha. And these could be very boring things.…”
Stebbings: Content marketing often takes two years to yield ROI
“I think this is the thing that everyone forgets with content, which is it takes two years often before you even see ROI come back at all.”
Mukherjee: Rox refuses unpaid design partners in favor of paid pilots
“We don't do any free design partners. Like we really focus on doing paid customers with like a pilot clause.”
Mukherjee: Unpaid design partners lack the commitment needed to build products
“If the customer is not separating with some sort of cash, there's just not enough skin in the game for them to put in the time to work with you for a few weeks, for a few months, for a few quarters to build a product. So I'm very anti unpaid design partners be…”
Mukherjee: Incumbent enterprise software takes 3 to 9 months to show value
“Incumbent enterprise software is gated by demos and implementation steps. This is three to six to nine month time to value.”
Mukherjee: Datadog won market dominance through six years of GTM consistency
“They never change segmentation. They never change comp models. They just ran a dialed-in machine for six to seven years, and they compounded value while everybody else kind of kept thrashing.”
Mukherjee: Enterprise software has shifted from commit to consumption revenue
“The way you sell, the revenue models, It's changed from commit-based revenue to usage or consumption-based revenue. So now you had the dollar values in the contract were just words. You only made money when people actually use the product.”
Mukherjee: AEs should hold both commit and consumption quotas
“So the student school of thought is only invest in the back carriers. Get the AEs, people who can hold a number. You eat what you hunt, like those folks, they should hold both the commit number, but also the consumption number. And it's an intense way to opera…”
Mukherjee: Modern software companies will shift to quota-carrying, high-variable compensation
“And I do believe that it is a hard path, but that is where modern businesses will go. You will invest in people who have full skin in the game and have this kind of owner mindset.”
Mukherjee: Early-stage startups must hire sellers who close without brand or PMF
“In early stage sales, the most important thing is to find these outlier sellers who can sell without A massive brand, a demanding engine, and perfect product market fit.”
Mukherjee: It takes 2-3 months longer to recruit outlier sales talent today
“Honestly, candidly, like we've done really well the last year, but it's taken me maybe two to three months longer to find those outlier sales reps is because it's just, it was pretty easy.”
Mukherjee: Most sales reps use consumer AI privately but lack enterprise access
“Most sales reps in the world use ChatGPT, perplexity, LLMs in their private life, but they can't use in their professional life.”
Mukherjee: Enterprise workers expect AI to perform any human work at scale
“When they come into the workforce, they actually want the software to do as much work as humanly possible. Specific word is humanly. The software should do anything that a human can at scale.”
Mukherjee: The alpha in sales AI is indexing full customer context
“The alpha is in being able to index all context about your customer.”
Mukherjee: AI should turn sales reps into orchestrators, not replace them
“My kind of view on automation is the human evolves to become an orchestrator, a quarterback, so we do all the work of composing your sequences, who to reach out to, but the human reviews and sends.”
Mukherjee: Fully autonomous AI outbound tools currently dilute company brands
“We're not at a point where a CEO or a CRO can just fire in the hole, kind of fill the pipe with these kind of outbound systems because it's going to dilute your brand.”
Mukherjee: Enterprise deals require 13 to 15 touchpoints to close
“It usually takes 13 to 15 touches.”
Mukherjee: Rox lets enterprise clients cancel if 30-day ROI metrics fail
“We only convert if we hit our ROI metrics. We run a 30 day, white glove, like, Palantir, like, kind of boot camp, where we work with the CFO, the CRO, and the C-suite on three metrics. The first one's dollar normalized outcomes, whether that's pipeline generat…”
Mukherjee: Rox reaches 75% weekly active and over 50% daily active usage
“We've hit 75% weekly active usages with more than 50% daily active.”
Mukherjee: Rox spent months building a chat AI agent before abandoning it
“The history of Farox is, it's a pretty experienced team. The first three to four months, we built a guard mode agent that knows everything about every customer with a chat interface. This does not work.”
Mukherjee: Enterprise AI tools face severe end-user retention and churn issues
“Is there going to be retention and churn issues from an end user perspective? A hundred percent.”
Mukherjee: Rox rejects experimental AI budgets, selling only from operational budgets
“We've made the decision of not tapping into it at all because of the focus on retention, usage, adoption, ROI. We only work with customers from their operational budget.”
Mukherjee: Top sales reps will handle 2x to 3x more output this year
“They will be able to do a lot more. They'll be able to do two to maybe three times more this year.”
Mukherjee: Legacy data vendors' 10–15 years of verified data is hard to replicate
“Having said that, the data providers, they provide a kind of source of value where they have 10 to 15 years of back-tested data of human kind of gathered kind of contact information, LinkedIn information, that they've been verifying through a lot of software, …”
Mukherjee: ZoomInfo has probably the highest quality B2B contact data
“Like Henry and ZoomInfo, for example, a phenomenal business. They have probably the highest quality contact data.”
Mukherjee: AI agents match data providers on company data, not contact data
“There's a split, like, company data is free game. Company data is public. Agents are, like, pretty much at par. Contact data is where the divergence happens.”
Mukherjee: Enterprise SaaS is currently undergoing a major rebundling cycle
“So what's going to happen is all the incumbents have obviously repositioned as AI companies now or agent companies, and they're also all bundling. Like as SaaS exploded, there was massing unbundling across sectors, definitely in the CRM space, right? You have …”
Mukherjee: Founders should hire five sales reps before adding a sales leader
“I back myself to recruit, train, and build a machine with five reps reporting to me directly, and then bring on a sales leader who kind of doubles the team and takes it from there.”
Mukherjee: Early sales hires need exactly 2 to 5 years quota experience
“Should have held the number at that ACV size for two to five years. Not less than two, highly not more than five”
Mukherjee: Many AI startups do not pay sales commissions to account executives
“You would be surprised that a lot of AI companies AEs are not compensated with commissions.”
Mukherjee rejects 30-minute interviews in favor of 90-minute working sessions
“I don't believe in 30 minute interviews. Either I believe in my kind of trust and honor network of folks telling me somebody's good or not. Or I do hour and a half working sessions.”
Mukherjee: Startup compensation must be logically explainable to avoid backlash
“Everybody's compensation, Should be explainable. There should be a rational reason as to why somebody is compensated a certain way. So I have guardrails, explainability guardrails that I don't want to kind of breach because if you're giving somebody some irrat…”
Mukherjee: Rox targets $100K initial deal sizes with expansion potential to $1M+
“A hundred K, like low six figures, and then really prove value, and we're doing, that is a no-brainer, quick land, we compete on time to value, and is there a path to high six figures, seven figures, and that's what we really care about.”
Mukherjee: Rox targets 45-day sales cycle for upper mid-market enterprise deals
“So if it's upper mid-market enterprise, we want to be at 45 days. From M-one to having kind of their first initial land basically done and signed.”
Mukherjee: Rox has $47M in the bank after raising $50M total
“We raised 50. I think we have Keep me honest, I say, 50, 47 in the bank. We put in our own money to the pre-seed, so it's like, we have full skin in the game.”
Mukherjee: I should have placed 5x more priority on early hiring
“In hindsight, I would have moved faster to build out that team, and then kind of going back after going to San Francisco, like, I need to spend more time. So the biggest mistake is I would have maybe five X'd the value I would put on just hiring.”
Mukherjee: Perplexity will remain a top-three player or dominate a large niche
“At the very minimum, they'll find a massive kind of niche, like a large Dow segment that they will be the best at. If not, they would be, like, a strong player in the top three for sure.”
Mukherjee: AI productivity will cut sales and marketing OPEX from 40% to 20%
“Instead of having large teams, you need fewer teams who hold more book, and management layers kind of compress, and then you need different kinds of leaders. And that completely changes, like, sales and marketing is 40% of your OPEX, right? Like, it could pote…”