AI-Native CRMs: Doug Camplejohn and Patrick Thompson on Displacing Salesforce

"CRM has been a lie from the beginning. It is not about customers, it's not about relationships. It's really about management. It really should just be called M." That's Doug Camplejohn, who ran Sales Cloud as EVP and GM at Salesforce before leaving to build Coffee, an AI-native CRM that starts from the premise that the category was mis-built. He calls the last twenty-five years a forced data entry march, with reps feeding a tool whose actual user was the manager looking over their shoulder.

Recall Sessions is a Village Global Podcast subseries hosted by Somrat Niyogi, founder of Recall Capital and a Village Global Network Investor. This episode brings two AI-native CRM founders into one conversation: Camplejohn of Coffee and Patrick Thompson of Clarify. Both are rebuilding the category from inside it after running pieces of it at Salesforce, LinkedIn, Amplitude, and Atlassian.

Listen to the full episode on Apple Podcasts, Spotify, YouTube, or wherever you like to listen.

For insights from across the Village Global Network straight to your feed, follow us on X, LinkedIn, YouTube, Instagram, and TikTok.

Key Insights

CRM was built for managers. The reps were the data source, not the customer.

"CRM has been a lie from the beginning," Camplejohn says. "It is not about customers, it's not about relationships. It's really about management. It really should just be called M. It has been a forced data entry march for reps, mostly to help the manager have some visibility as to what everybody's doing."

The history backs him up. "If you go back to the Tom Siebel days, he would basically use the CRM as a stick to go to rep. Like, how come I knew you had this meeting? How come you didn't log this in?" The product was a surveillance instrument that asked the rep to type their day into a system so a VP could run a pipeline review on Friday. Reps complied because their compensation depended on it. They never loved the tool because the tool was never theirs.

Camplejohn references a frame from Marc Andreessen to explain what changes now. "We've gone from the Mad Men era where the secretary would prep you for your meetings. When email came around they would print out the emails and bring them to you. Now the work has shifted back to the individual, the manager, or the executive. And now we're taking some of those roles and they become agents for us." The support layer that disappeared in the move to email is coming back as software. Reps get an assistant that does the prep, logs the call, drafts the follow-up, and updates the record. The manager still gets visibility, but as a byproduct of work that actually happened, not as the reason the rep was typing in the first place.

Camplejohn and Thompson are betting that founders building rep-first tools can take the manager's visibility along for the ride. The rep gets a tool that prepares the meeting and updates the record. The manager gets the same dashboard, populated by work that actually happened. Reps had no exit before. Now they do.

You can vibe code a CRM in a weekend. You'll get 1987's ACT.

Camplejohn is direct about the build-it-yourself fantasy that has become fashionable in AI circles. "You can vibe code a CRM in a weekend. Congratulations, you just created a contact manager, 1987 version of ACT. It's like an iceberg. The hard part about CRM is really all the data and all the architecture underneath that you're not gonna do on a weekend."

Thompson agrees from a different angle. He uses a brewer's heuristic he picked up earlier in his career: "Do what makes your beer taste better. If you're not a CRM company, why build a CRM?" The point is the same one Camplejohn is making. The visible UI of a CRM is the small fraction of what a CRM is. The data model, the integration surface, the workflow logic, the permissions structure, the deduplication, the activity capture, the reporting layer: that is the iceberg. Salesforce spent twenty-five years building it, and the moat is real.

A solo founder can ship a working contact manager by Sunday night. Shipping a system of record that a 500-person revenue org can run on is a different problem, and a better LLM does not solve it.

AI on bad data inherits the bad data. That's why Agent Force underwhelms.

"One of the tenets we said from day one was you can't have good AI with bad data," Camplejohn says. "The problem that Salesforce has had with Agent Force in the sales world. You look at the average sales cloud instance and how crappy the data is in there. It's really hard to get good AI out of that." The output of any agent is bounded by the substrate it's reading from. Bolt an LLM onto a CRM whose data was entered under duress by reps gaming a forecast, and the agent will reflect that corpus back to you with high fluency and low accuracy.

Camplejohn has a phrase for the architectural divide. "I always call it BC. There's before ChatGPT and after. So all of those players, I just say, yeah, if you want a BC CRM, you can do that. Slapping an LLM on the side of a relational database CRM, that's BC, is very different than what platforms like us have built." The distinction is not marketing. A CRM designed before generative AI assumed humans would be the source of structured data. A CRM designed after assumes agents will capture, structure, and maintain that data, and the schema, the permissions, and the activity layer are all built around that assumption.

Salesforce can ship better agents on top of Sales Cloud, but they can't fix the corpus underneath without rebuilding how reps enter data in the first place. Rebuilding that breaks thirty years of customer workflows. Camplejohn watched the room where that tradeoff gets debated. He knows which side wins.

Salesforce is the most expensive database in your company. That's a pricing problem.

"Salesforce being the most expensive database that organizations use, I had a query via SQL at Amplitude, about half the company didn't have access to it. This is primarily a pricing issue. So for us, we give the CRM away for free and we charge for the work that's done on top of it," Thompson says.

Per-seat pricing for the customer-data system of record means the customer-data system of record is the smallest accessible database in the company. Marketing buys a license to see it. RevOps gates access. Product never logs in. The data that should be the most widely circulated asset in the company becomes the most tightly rationed, because every seat costs $150 a month.

Clarify charges for output: the agent runs, the data work, the sourcing, the enrichment, the outbound drafting. The CRM itself is free. Buyers pay for what gets done on the data, and the system underneath has to do the work, which is where Camplejohn's architecture argument meets Thompson's pricing argument from the other side. Pricing the work instead of the seat is what turns Clarify into a Salesforce alternative at the category level instead of the feature level.

Salesforce can't rewrite its data layer. Clari can't grow into one.

Camplejohn has internal knowledge of what a pivot would require. "I've been inside of the Valley of the Beast of Salesforce. They can't turn that ship. They can continue to glue things onto the Frankenstack that they've delivered. But to fundamentally change that, we're just at a very different inflection point in this market." The "Valley of the Beast" framing is access talking, not an outsider's potshot. He ran Sales Cloud. He knows what it would take to rewrite the data layer, and he knows the org chart that would have to agree to it.

The same constraint runs in the other direction for adjacent players trying to grow into CRM. "It's hard to get to the place you want to get to if you didn't have an idea you wanted to get there in the beginning. So suddenly Clari waking up and saying, oh, we want to be a CRM, you're basically gonna rebuild the plane mid-flight, the fuselage, the engines, everything." Clari and Gong started from workflow tools. Their engineers built the activity capture to serve forecasting accuracy. Retrofitting that into a system of record means their teams have to rewrite the schema, the permissions, and the activity layer at the same time the existing customers are still running on top of the old version.

Salesforce poured its data layer before generative AI existed. Coffee and Clarify poured theirs after. Salesforce protects $300B of market cap and twenty-five years of customer lock-in. Clari protects an install base of forecasting customers. Camplejohn and Thompson have neither to defend, and that's the bet.

An AI-native CRM is a low-odds bet on a generational outcome.

"The way I talk about this category is the opportunity for success is low, but if you're successful, it's a big business," Thompson says. "You can go build a better point solution where the opportunity for success is higher, but if you're successful, it's a smaller business." That's the honest expected-value frame for any founder weighing the category.

Displacing Salesforce is a low-probability, high-payoff bet. A point solution that drafts email or scores leads has higher odds and a lower ceiling. Camplejohn and Thompson picked the harder side because the prize, if they get there, is a generational company.

About the Guests

Doug Camplejohn is the founder and CEO of Coffee, an AI-first CRM. He previously served as EVP and GM of Sales Cloud at Salesforce and led Sales Navigator as VP of Product at LinkedIn, where he joined after Fliptop, the predictive analytics company he founded, was acquired in 2015. Earlier exits include MI5 Networks (acquired by Symantec) and MyPlay (acquired by Bertelsmann).

Patrick Thompson is the co-founder and CEO of Clarify, an AI-native CRM that has raised $22.5M. He previously co-founded Iteratively, a customer data platform acquired by Amplitude in 2021, where he served as Director of Product and GM for Amplitude CDP. Earlier in his career, he led design for Jira Software at Atlassian.

About the Host

Somrat Niyogi is the founder and GP of Recall Capital and a Village Global Network Investor, writing pre-seed and seed checks into AI-native companies. He previously served as General Manager and Head of Business Development at Gusto, with earlier operator roles at Clari, Miso, Stitch, and Salesforce. Together, Village Global and Somrat back early-stage founders across AI, fintech, and B2B SaaS.