Stop Paying For SaaS: The AI Playbook Driving A $1B Trajectory | Ghazi Masood, CRO @ Replit
Ghazi Masood on building an AI-native GTM org at Replit
The most AI-native GTM org I’ve seen is hiring more salespeople. That sounds backwards until you understand what Replit is actually building. They’re automating inbound. They’re building apps on top of Salesforce. They’re creating their own CPQ, forecasting, customer health dashboards, and research tools. But they’re still buying Salesforce. They’re still using Nooks for dialing. And they’re still hiring humans who can create trust, teach the market, and explain the art of the possible.
That’s the real AI-native GTM lesson: don’t automate the org chart. Redesign the work.
I had Ghazi Masood on pod to talk through what this looks like in practice. Ghazi is the CRO at Replit. He previously helped build enterprise GTM motions at Auth0 and Retool, and now he’s building the revenue org for one of the fastest-growing companies in AI.
Replit went from roughly $2M to more than $150M in revenue in less than a year. They’re hiring 200+ people over the next 12 months. They’re building a B2B GTM company inside a PLG rocket ship.
This is the part I found most interesting: the playbook is more grounded than “replace everyone with agents.” It’s much more practical.
Use AI to remove drag. Use humans where judgment matters. Build the layer where your workflow is unique. Buy the systems where reliability matters more than customization. That’s the operating model.
A quick plug
I’m headlining Revstar in a few weeks (June 2) in Toronto and I’m going to be talking about a bunch of very familiar topics to you fine readers. There’s a ton of other great speakers like Jordan Crawford too.
I’m breaking down the 5 decisions that every founder and CRO has to get right to build an AI-native GTM org: who owns AI, what to build vs. buy, where to start, how org design changes, and where humans stay in the loop.
You’ll see what’s actually working inside Owner including how we quadrupled revenue per AE, doubled the rates at which we connect with decision makers, and AI workflows that increase selling time instead of creating more internal theater.
Hope to see all the Torontians there!
1. AI-native GTM makes human judgment more valuable
Most AI conversations in sales still start with the wrong question.
“What roles can we automate?”
That framing leads to lazy answers. Automate BDRs. Automate CSMs. Automate managers. Cut whatever headcount looks expensive in the board deck. Ghazi is taking a different path.
Replit is automating real work, but they’re also hiring aggressively. That only feels inconsistent if you think the point of AI is headcount reduction. I don’t. The better goal is capacity creation.
McKinsey estimates that about 75% of the value from generative AI will come from customer operations, marketing and sales, software engineering, and R&D. Their research also suggests generative AI could increase sales productivity by 3% to 5% of current global sales expenditures (McKinsey). That matters. But productivity is only useful if you know where to redeploy the time.
At Replit, the answer is trust, education, and category creation. Ghazi said the best reps are not showing up as traditional salespeople. They become the customer’s AI advocate.
“They are not viewed as a salesperson in front of the customer’s eyes. They’re advocate, they’re like their AI advocate, they’re like their AI trainer.”
That sentence is the whole shift. If you sell a mature category, you can optimize the funnel. If you sell a new way of working, you have to teach the market. AI can write the research brief. It can summarize the account. It can draft follow-up. It can suggest use cases. It cannot create conviction in a buyer who doesn’t yet understand what’s possible.
That takes judgment.
Harvard Business School research makes the same point from a different angle: AI access alone does not replace business judgment. Leaders still need training, context, and decision-making skill to use the output well (HBS). That’s why Replit is still hiring aggressively. They’re changing what those people do.
2. Build the workflow layer. Buy the plumbing.
The most useful part of the conversation was Ghazi’s build vs buy line. Replit is a company that helps people build software with AI. If any company was going to vibe code its entire GTM stack, you’d expect it to be them.
They didn’t.
They’re moving from HubSpot to Salesforce. Ghazi was direct about it:
“The big behemoth question is, hey, are we building our own CRM? No, we’re not.”
That matters.
Replit is building a revenue copilot on top of Salesforce. They’re building customer health views. They’re building their own CPQ and quote-to-cash workflows. They’re building forecasting. They’re building research and scripting tools for the BDR team. But Salesforce remains the system of record. That’s the right model for most scaled companies. Build where the workflow is proprietary. Buy where failure breaks the company.
Ghazi described the logic well. Large enterprise systems connect to billing, finance, accounting, renewals, true-ups, delinquencies, back office, and front office. Replacing that plumbing creates a ton of risk. And for what? A slightly more custom CRM?That’s a bad trade.
Gartner’s RevOps research points in the same direction. RevOps exists to unify the end-to-end buyer journey across people, process, technology, and data (Gartner). The key word is unify. If your core systems are brittle, every AI workflow built on top inherits the mess. This is where a lot of teams will get sloppy. They’ll confuse “we can build it” with “we should own it,” an those are different decisions.
The heuristic I took from Ghazi is simple:
If downtime stops the revenue org, buy it.
If governance, security, or financial accuracy matters, buy it.
If the workflow creates unique operating leverage, build it.
If the tool is mostly coordination, visibility, or internal experience, build it fast and iterate.
That’s how I’d think about it at Owner too. I don’t want to vibe code Salesforce. I do want to build the operating layer around Salesforce that makes our team faster, sharper, and more consistent.
That’s the real opportunity.
3. RevOps is becoming a product team
Ghazi said the first thing breaking at Replit is RevOps. That tracked immediately because when a company grows this fast, the pressure does not show up first in the close plan. It shows up in territory design, comp plans, routing, data hygiene, quoting, pricing, account ownership, and forecasting.
Ghazi’s words:
“Right now it’s the RevOps stuff. We are doing so much change management. Territories, comp plans, quotas, systems, hygiene, the way we structure deals.”
That’s what hypergrowth feels like operationally. Everyone wants to talk about the sexy AI workflows. The unsexy truth is that AI makes bad infrastructure louder. If your territories are unclear, agents will route faster into confusion. If your CRM is messy, copilots will summarize bad data with confidence. If your quoting process is duct taped together, AI will help you generate errors faster.
This is why RevOps has to evolve. The old RevOps job was admin, reporting, troubleshooting, and governance. The new RevOps job looks much more like product management.
What does the internal customer need?
Where does the workflow break?
What should be automated?
What needs a human approval step?
What should live in Salesforce, and what should live in a custom app on top?
Gartner has started using the language of revenue technology rather than sales technology, with emphasis on cross-functional workflows, data, user experience, and automation across the full revenue process (Gartner). That is the right frame.
The future RevOps leader will need product taste. They’ll need to understand systems design. They’ll need to sit close to sellers and customer-facing teams. They’ll need to know enough AI to spot leverage, and enough operational reality to prevent chaos. That’s a very different talent profile and it might become one of the highest leverage roles in the GTM org.
4. The best AI sellers may not look like traditional sellers
Ghazi is hiring salespeople from non-traditional backgrounds. One of Replit’s top sellers came from the Marine Corps. Another top commercial seller was a school teacher. That would have sounded reckless in the old enterprise sales playbook.
Find the top 1% seller. Make sure they crushed quota at Oracle, Salesforce, or another known training ground. Pay up. Repeat. Ghazi is still hiring experienced sellers too. But pedigree is only one signal. He’s looking for product passion, AI fluency, communication, and the ability to explain what’s possible.
“I’m optimizing for more people who are passionate about AI, that are passionate around our mission, and that really, really believe in what we’re doing.”
That has big implications. In a new category, product belief matters because the seller has to transfer conviction before they transfer information. Sometimes the best seller is the person who has built 10 weird apps for their kid’s Little League team and can explain the product like a practitioner. Give me that over clean MEDDICC vocabulary with no product conviction.
BCG’s AI at Work research found that only 30% of managers and 28% of frontline employees had been trained on how AI will change their jobs, even though regular users are saving meaningful time with GenAI (BCG). That gap is the market opportunity. Customers need guides. They need someone who can translate AI from concept into workflow. That person might be a former teacher. Honestly, that makes sense. Teaching is selling when the buyer doesn’t know what they don’t know yet.
Ghazi’s hiring process reflects this. He does 15-minute screens and looks hard at communication. Then candidates go into a role-specific panel with a prompt. He’s also comfortable with an 80% hiring bar if enablement can close the last 20% which is an interesting way to look at it.
A lot of leaders claim they want speed, then run hiring processes built for perfect certainty. You can’t hire 200 people that way. You need a clear trait profile, a fast signal process, and an enablement system strong enough to turn potential into performance.
5. The future GTM leader has to scale down
This was my favorite leadership point from the episode. Ghazi is a CRO at a company growing at a ridiculous rate, and he’s still doing first screens. He’s in customer conversations. He’s close to deals. He’s involved in the operating details.
He said it plainly:
“I’m not a chair CRO that’s looking at spreadsheets and doing orders. I’m actually doing the work.”
That’s the new standard. In an AI-native org, executives can’t hide behind dashboards. The dashboard will get better. The summaries will get better. The forecast will get better. That raises the value of firsthand judgment. If everyone has AI-generated summaries, the advantage shifts to the leader who can tell which summary is missing the point. That requires being close enough to the work to smell the difference.
HBR called out the risk of “dataism,” the false belief that more data and better algorithms can uncover the truth and make the right decisions on their own (HBR). I see this risk everywhere in GTM.
AI will make executives feel more informed than ever and some will confuse that feeling with understanding. The best leaders will use AI to scale down. They’ll inspect calls faster. They’ll review hiring signals faster. They’ll understand customer patterns faster. They’ll spend less time assembling information and more time making judgment calls. That’s a very different version of executive leverage.
The old model was delegation through layers. The new model is compression. A CRO can stay closer to the customer, the candidate, the workflow, and the system because AI removes the prep tax. But only if they choose to stay close.
The real AI-native GTM question
The easy AI question is: “What can we automate?” The better question is: “Where should human judgment get more leverage?” That’s what I took from Ghazi.
Replit is building an AI-native GTM org around a sharper division of labor. Agents can handle routing, research, internal tools, dashboards, drafts, and coordination. Humans still need to create trust, teach the customer, design the system, make the judgment call, and carry conviction into the market.
That’s the line.
Build the workflow layer. Buy the plumbing. Hire for judgment. Turn RevOps into a product team. Stay close to the work. Most teams will over-automate the easy stuff and under-redesign the important stuff. The winners will do the opposite. They’ll redesign the work instead of automating it.
If you enjoyed this episode, subscribe to The Revenue Leadership Podcast and leave a rating. It helps more revenue leaders find the show.