Why I Left VC To Get Back Into The Operating Seat w/ Liam Mulcahy, Head of GTM @ Parallel Web Systems

This week Chris sits down with Liam Mulcahy, Head of Go-to-Market at Parallel Web Systems, who spent seven years in venture at Kleiner Perkins before taking an operating role and bringing seven people with him. Liam's claim is that the innovator's dilemma gap has effectively closed: Parallel has closed roughly 10 Fortune 500 enterprises in the last four months, one of them the fastest AI deal a major bank could remember, and those buyers are now evaluating with the same sophistication as agentic-native companies.

He also makes the case that a modern GTM org needs 50% fewer salespeople up to half a billion in revenue, walks through the payback-of-one standard he holds new reps to inside 45 days, and explains what happens when a rep shows up with a PhD-level brief built in Claude and can't say whether the prospect is based in New York or San Francisco. It's his second time on the show, and he spends most of it arguing against the way sales teams are still being built.

The classically trained question is who's in control of a deal - you, the prospect, or your competitor. Liam's addition: there's now a fourth party, how the buyer configured their eval harness. AI natives have gotten so used to passing decisions to their CLI that prospects will literally screenshot prompts: "my coding agent is telling me that you're better." The eval is the new POC, and buyers want it pure - a "demilitarized zone" that even verified champions won't let you enter, because they don't want you prompt-engineering your way to a win.

The counter is all upstream: nail use case, decision criteria, and what the harness needs to look like before they hit go - measure twice, cut once. Done right, it rips: six-figure deals tested in a day, a seven-figure deal over a weekend.

The premise of the Innovator's Dilemma is that big companies decide slowly and small ones outmaneuver them. Liam's claim: for enterprises building on the frontier, "there's almost no gap anymore." He and Graham modeled the enterprise clicking on around October - the typical quarter lag from his Kleiner years. Instead, Parallel closed roughly ten Fortune 50/500 deals in the first four months, with buyers evaluating as sophisticatedly as agent natives (many hired talent from them to build horizontal LLM/agent units).

One major financial institution went first conversation to mission-critical production in four months - the fastest deal the buyer could recall at the bank - and to Parallel's founders it felt slow. And since the biggest bank in the world isn't self-serving anything, the overlap is speeding up enterprise hiring, not slowing it.

Six months ago Liam said you'd need 20% fewer salespeople to scale. Updated take: 50% fewer, up until half a billion in revenue. Traditional capacity math says Parallel should have 70 reps; they have 24-25 and are on pace. But Jevons paradox cuts against the one-person-company meme: "if you have proof that five gets you output X, then why not have ten?"

Since revenue is a lagging output, he manages inputs. The standard: payback of one - a rep covers their base salary in closed deals within 45 days of joining, setting up a 7-8x yield on that rep over the year. Second gauge: days-to-close vs. deal size staying in the black - a $50K deal in 20 days signals product-market-sales fit; $30K deals in 90 days means something upstream is broken.

Parallel's internal GTM hub (built on their own API) pre-loads every rep's accounts: deep research, org charts matched to ICP, theoretical value props, sample messaging, and live monitors. "Giving that to sellers is like giving them an Iron Man suit... but inside of that is still Tony Stark." The one thing you can't automate is forming a real opinion about how a business works.

The failure mode: a rep presents a beautiful customer-facing doc built in Claude, then gets asked how the company makes money - and Liam hears them start typing. "You can't abdicate your unique ability to do the job" - otherwise a smart founder would be justified in not hiring a sales team at all. The proof it cuts both ways: a Parallel deployed engineer with zero classical sales training walking a top-five insurance company's CIO through their own strategic initiatives.

Liam builds org charts off the ideal customer journey, not SaaS silos - the old handoff chain was "eight cooks touching their food before it's ready." Parallel runs an AI-native team, an enterprise team, a product-powered self-serve motion, and no SDRs or BDRs, probably ever: inbound is "a drug that can dry up - then you go through withdrawal and die," so every rep owns their own pipeline generation. Also: imagine strapping a phone to a 22-year-old's head to cold-call staff AI engineers.

In the middle sit three flavors of deployed engineer (the closer, the PR-committing extrovert, the launcher). Post-sale, "the traditional CSM role's cooked" at the early stage - replaced by deployed strategists, and by consultant-profile hires, who Liam thinks are "most dangerous post-sale." His BDR replacement: a squire model - shadow an enterprise rep full-cycle for six months, then earn your way to customer-facing.

On his own prediction that every rep will have a GitHub by 2027: "If you're not steeped in these tools for personal and professional reasons, that ignorance is unemployment. You will be the 50% we don't need anymore." The bar isn't just professional - he's running his fantasy football lineup through agents and built a schedule app with his wife for their kids' school calendars.

Why GitHub specifically: GTM is now the fastest-rising use of Anthropic and OpenAI inside companies, and reps who build naturally start shipping one-off web apps that can leak data. A repo is both the secure path and the forcing function for real product proficiency. The end state: a repo per deal, decks and dashboards updating with live usage data - "what Figma did for design, but for go-to-market."

(4:12) Five years at Kleiner, and jumping back into the operator seat

(12:10) Picking your spot in the AI stack: labs, inference, apps, and the retrieval bet

(16:02) The eval harness: a fourth party in every deal

(27:43) The enterprise lag is dead: ten Fortune 500 deals in four months

(33:12) Capacity math: the Degnan debate & payback of one in 45 days

(44:26) "How does this company make money?": the Claude doc story

(50:51) The new org chart: three flavors of DEs & the end of the CSM

(56:37) Why Parallel will never hire BDRs

(1:01:50) Ignorance is unemployment: every rep needs a GitHub

(1:06:45) Where the talent went - and the Shawshank close