Startups to Join: Rox
The startup taking on Salesforce
The Quick Rundown:
Rox is building an AI-native Salesforce that uses AI agents to augment sellers and support them with planning, research, and engagement.
They’ve raised $50M in funding from Sequoia, GV, and General Catalyst, and their product is already being used by 35+ top enterprise sales teams including MongoDB, OpenAI, and Ramp.
They’re actively hiring engineers in both SF and India.
This is the first edition of “Startups to Join” – a blog for engineers and designers to discover early-stage (pre-seed through Series A) startups that are reimagining industries and on the path to becoming enduring companies.
Every week, I review hundreds of startups and spend hours talking to VCs and founders to write about one company that’s building something special. Whether you’re looking for the next unicorn to join or just curious about what top VCs are investing in, subscribe below to keep updated:
The Death of Salesforce
Last year, I had dinner with an ex-founder turned VC who has built and invested in multiple unicorns over the last two decades. Toward the end of the meal, I asked them what three factors make a great product. Their response: “something with network effects, high switching costs, and that serves a core workflow so critical that teams are entirely dependent on the product”.
Salesforce is the dictionary definition of this.
With each additional company, Salesforce’s ecosystem becomes more attractive for developers and its AI insights product (Einstein) also experiences increasing returns. It also costs close to a million dollars to just implement Salesforce which makes it extremely difficult, both financially and emotionally, to rip & replace. And, on top of that, CRMs are simply critical to the workflow of any team that manages customers at scale.
In other words, despite all the problems with the product, Salesforce has proved remarkably hard to displace.
That’s why I’ve always been pretty skeptical anytime I come across a new company claiming to be building the next-gen Salesforce. But, after speaking to 100+ founders and sales teams over the last few months, I’ve repeatedly heard a few key trends that made me realize why now is a unique moment in time to displace Salesforce and build a new core system of record.
Why Now & Why Rox
First, sales processes have completely changed.
The old enterprise sales playbook of sales-led growth used to follow a structured process: meet with an executive, run a POC for a few months, and then negotiate a large upfront contract. In this old playbook, teams focused on closing large initial deals after long sales cycles and revenue growth was driven by new logo acquisition and annual renewals.
This sales motion - prospect, close, renew - aligned perfectly with Salesforce's CRM architecture. Salesforce was – and still is – the best source of record for structured data. Its architecture is purpose-built to ensure data consistency making it the ideal product to track discrete deals, manage pipelines, and log customer interactions. This reliable structured approach is one of the reasons Salesforce has been so sticky and defensible.
But sales processes are no longer linear, they are circular.
Instead of focusing on signing large upfront contracts, the modern sales motion has shifted to a continuous cycle of land & expand. As a result, companies have moved from large upfront contracts to consumption/ usage-based pricing to cater to this new reality. In 2022, 61% of SaaS companies offered some form of usage-based pricing, up from 34% in 2020.
And, as this shift has played out, the “expand” component of this new sales motion has become critical to revenue growth – SaaS companies with $20M+ in revenue now see nearly 50% of their new ARR coming from expansion revenue within existing accounts. This data is also backed up in the public market: public SaaS companies with usage-based pricing models experience a 50% revenue multiple premium compared to their peers.
As a result, the focus of sales teams has shifted. Sales teams now care less about initial deal size and more about whether customers are seeing value in the product. Kyle Poyar does a great job of summarizing this new era of sales: “Every day the customer is making a decision about whether to use the product, which means every day they could potentially decide to stop paying altogether”.
So, as companies started to switch to usage-based pricing, the big question for sales & revenue teams became: how can we expand existing accounts? The answer was in the data: leverage dynamic data, not static. Instead of just tracking pipeline and renewal dates, revenue teams realized they needed continuous visibility into product usage patterns, customer health metrics, and support tickets. Most of this data is unstructured, existing in silos across ERPs, CRMs, and customer support tools, and managing it requires an entirely different architecture than what Salesforce was built for.
Therefore, what was once Salesforce’s greatest strength – a structured data model – now acts as a key limitation. As investors at a16z put it: “While incumbents often adapt to new platform shifts, they are rarely able to completely rethink their architecture”. Salesforce’s architecture is explicitly not designed for unstructured, usage-based data which is the data that high-performing SaaS sales teams now rely on.
This new reality has led to a migration of customer data from CRMs like Salesforce to data warehouses like Snowflake because these data warehouses are built to support both structured data (contracts, deals) and unstructured data (product usage, support tickets, and customer communications). According to some estimates, 40% of the data that now exists in modern data warehouses is actually just customer data. This is where the data that matters for sales teams lives.
So, put simply, the data that made Salesforce so valuable is leaving the platform which means that its data network effects are slowly eroding. And, more importantly, sales teams are spending less time in Salesforce because the data they rely on no longer lives there.
Instead of being in Salesforce, sales teams are now spending their day navigating the custom tools that their internal engineering teams have built to access the data they need. Building this custom infrastructure has resulted in an expensive engineering effort for companies. Of the sales teams and founders I’ve spoken to, the majority currently build complex data pipelines to pull information from multiple sources into Snowflake – product usage data, customer support tickets, CRM data, and communication logs. Then, they create custom data models and logic to unify this data and develop health scoring models. Sometimes they also add a BI tool like Looker to create dashboards on top of their data warehouse.
As a result, the new reality for sales teams today is a fragmented workflow: checking Looker dashboards for usage data, searching LinkedIn for contact research, browsing news for account updates, reviewing meeting recordings, and updating Salesforce for pipeline tracking. All these context switches take up a lot of time, and for high-velocity sales teams, time is the most important asset.
This is where I believe the opportunity lies to build a “system of intelligence” wedge that enables the displacement of Salesforce.
The best salespeople will always enter a meeting informed with the latest information. But instead of having to spend 10 minutes combing through all these data sources, AI agents can automate this. AI is especially good at dealing with unstructured data, meaning that sales reps can spend more time on what actually matters and companies can spare themselves the engineering efforts of building complex data pipelines and integrations. This opportunity to build the “system of intelligence” also avoids the high switching costs of Salesforce – you are providing value to companies from day 0 without requiring them to rip out Salesforce (creating a perfect, low-friction initial wedge).
With this new system of intelligence, the data has already left Salesforce to data warehouses (no more data network effects), the intelligence system now becomes the core workflow, and the switching costs of leaving Salesforce are avoided in the short term as companies don’t have to rip it out. Put simply, the factors that made Salesforce defensible no longer exist in this new reality.
But, building an AI sales insights tool alone doesn’t displace the 800lb gorilla that is Salesforce. It does, however, lead to a very key asset: access to the data that matters.
At the end of my dinner last year, I asked the VC what some of their favorite questions were to ask founders. One of them was about what the second-order effects are if their initial product is successful. Here, the second-order effects are clear.
If you build a system of intelligence that becomes so core to the daily workflow of sales teams, they will start to add more data integrations (calendars, mail, meeting transcripts, etc.). And, as models improve over time, the insights provided by AI agents will only become more valuable. This means that this intelligence platform will become so entrenched into the core workflow of sales teams that it gains access to the data that matters, creating a clear path to become the de facto system of record. In this new reality, the last defensible moat that Salesforce has – high switching costs – will become a secondary factor; the CRMs of today will just become the backend systems of tomorrow.
This strategy and product is exactly what Rox is building and it’s why I chose them to be the first company I highlight in this new “Startup to Join” series. It’s also why sales teams at companies like Ramp, OpenAI, and MongoDB rely on them and why VC funds like Sequoia, Google Ventures, and General Catalyst have already invested $50M in them.
Product Overview
Rox is an acronym for revenue operating system and it acts as an agentic CRM to help at-scale businesses acquire, retain, and grow customers. When you log into the product, you can add accounts and an AI agent is automatically assigned to each account – one agent per account. These accounts have basic fields (like competitors and headcount), but you can also add in custom fields with queries like ”have they raised >$20M in funding” and the AI agent will surface relevant sources.
Each agent is tasked with researching its account using both internal and external data sources, ranging from news articles to previous meeting recordings that you provide. With this knowledge, the agent can create a company report to help AEs plan what accounts to prioritize, provide insights ahead of meetings, and optimize the timing and type of engagement. Teams can also tune their agents by specifying key things to look out for when monitoring public and private data for insights.
Importantly, the product doesn’t do the actual selling – it augments the seller instead of replacing them. One way it does this is by providing a feature called “Pipe Gen Actions” that suggests which accounts to reach out to and provides a draft email tailored to specific contacts. While the feature still has flaws, as base models improve and agents have access to more emails to train on, I can see a world in which teams begin to heavily rely on Rox for outreach and engagement too.
Technical Architecture
Behind the product, Rox has built a technical architecture that is purpose-built for both unstructured and structured data sources. At the foundation is their data lakehouse, which processes multi-modal data at a terabyte-plus scale. This layer integrates unstructured sources like news and job postings with structured data from CRMs, support systems, and product usage time-series data. Rox also has bi-directional API-based connectors for tools like email, calendars, and Slack which means that it can act as a single, unified system for customer data.
On top of its data lakehouse, Rox has a unified knowledge graph that maps companies and people across distributed data sources. This is not just a simple database of relationships; it's a dynamic system that updates in real-time as new information becomes available. Ultimately, it is this knowledge graph that enables AI agents to access and process information contextually.
This bi-directional architecture and flow means that Rox’s intelligence layer can read from the knowledge graph and write back into the system of record, allowing the agents to support sellers.
Team
Alongside its exciting product roadmap and long-term goal to become the de-facto source of record for sales and revenue teams, Rox’s founding team also stands out.
Before starting Rox, Ishan Mukherjee (CEO) served as the product lead for Siri, co-founded Pixie which was acquired by New Relic, became the Chief Growth Office of New Relic, and then grew their self-serve business from 0 to $100M+. Shriram Sridharan (co-founder) previously led the Kafka data infrastructure org within Confluent and was also one of the founding members of Amazon Aurora. Avanika Narayan (co-founder) was a Knight-Hennessy Scholar at Stanford, has spent time at both Sequoia Capital and Palo Alto Networks, and previously created a non-profit that spread CS curricula to 10,000+ low-income, first-generation students. Diogo Ribeiro (co-founder) has also previously scaled GTM and AI systems at Premise, ThousandEyes (acquired by Cisco), and Lacework. A pretty strong founding team all things considered :)
Closing Thoughts + Jobs at Rox
As you can probably tell, I think now is an incredibly exciting time to be building in GTM tech. With customer data leaving Salesforce, AI’s ability to handle unstructured data, and base models improving, this is a unique moment in time to attempt to displace Salesforce. And, while there is still a long way to go, if Rox’s system of intelligence can earn the right to manage the data layer, they are perfectly positioned to improve the interaction layer alongside existing customers and ultimately have a chance at becoming the de-facto core system of record.
If being part of a team building the next-gen AI Salesforce sounds exciting to you, Rox is actively hiring across a lot of engineering roles.
Resources
https://www.sequoiacap.com/article/partnering-with-rox-every-seller-needs-an-agent-swarm/
https://a16z.com/ai-transforms-sales/
https://www.gv.com/news/rox-ai-sales
https://www.generalcatalyst.com/stories/our-investment-in-rox