Davis Martens

I recently left BCG to explore new ideas. We all sense that what's happening with AI agents and tool use is a big deal, but the real opportunity lies in how we actually apply them. AI is changing the rules of software, unlocking new leverage across every category. Many legacy providers won’t adapt fast enough, creating openings for a wave of new startups. From accounting and legal to ERP, we’re entering a new internet age. I’ve been exploring how agents can shake up revenue and customer operations, especially in the $300B+ CRM market .

As a previous founder , and through my work on deal teams at BCG , I've seen the immense effort it takes to build relationships that actually win deals. But nowhere did this inefficiency hit home harder than earlier this year. After moving to New York, I spent a few months helping a friend run business development for his successful construction business. What I encountered was a harsh reality check: before I could even send a single outreach email, I was signing up for hundreds of dollars' worth of subscriptions to tools like Apollo, Clay, Salesforce/HubSpot/Attio, and Instantly.

Then, instead of building pipeline, I spent all day researching customer profiles, drafting templated emails, and agonizing over proposals, just to sound at least a little differentiated. It was overhead, not leverage . And sales is just the beginning. Once you win deals, customers expect a white glove experience : constant follow ups, clarifying questions, getting quotes and pricing from suppliers, and providing regular updates. It's a full time job in itself, except it isn't—my friend still needed to manage crews on site, which was his actual day job.

The Core Issues with Today's CRMs​

For most sellers, CRMs have become synonymous with tedious tasks like updating fields, generating reports, and writing endless notes. These systems were largely designed to manage the processes and complexities of large enterprises. Actually using them to grow was never their objective. As a result, millions of individual sellers and small teams, like startup founders and small businesses, are left underserved. The promise of seamless integration with other tools often falls flat, requiring additional platforms or custom development. Furthermore, the lengthy and complex configuration processes demand significant money and time investment, locking users into platforms for months before they even know whether they deliver any real value. That's simply not an option for top revenue teams.

Scaling templated emails and LinkedIn outreach is no longer the answer; it just creates noise. Customers expect more. They ignore irrelevant outreach, tune out bad ads, and unsubscribe from lifeless newsletters. But creating an experience that feels deeply personalized and relevant is still incredibly manual. That’s the bottleneck holding growth back.

However, we have entered a new era where lean teams are rewriting the rules . Companies like Cursor, Midjourney, Adept, Anysphere, and Anthropic are reaching $2M+ ARR per employee , scaling at lightning speed. That’s the new benchmark: small, hyper effective teams scaling faster than ever. To compete in this new world order, others need to give their employees 10x+ leverage ; to do more, move faster, and deliver better outcomes.

Here's a revised version of that section, focusing on the founder managing teams of agents and delegating tasks where their direct involvement isn't critical for value.

A Vision For Revenue Ops​

My vision for sales teams is to augment founders with AI agents and turn the CRM from a passive system for logging activity or pulling reports into the central hub where founders orchestrate AI agents. These agents don’t just summarize; they act. They work like an extension of the founder: monitoring accounts, launching campaigns, engaging prospects, managing relationships, and surfacing insights in real time, adding value through every interaction.

One agent might identify a new target segment based on win patterns. Another could test 100 versions of a subject line to see what converts. A third might watch for signals (a funding round, a strategic hire) and prepare the perfect message before you even notice.

And it's not just about acquisition. These agents stay with customers through the full lifecycle. They maximize value, resolve issues, and deepen relationships. Imagine an AI that responds to RFQs within the hour or keeps clients updated without a human lifting a finger.

Instead of implementing complex, static workflows, teams will leverage dynamic agentic workflows to tailor the experience to the customer, not the other way around. Need to generate invoices? Build a tool for an agent to generate and send them. Need to integrate data from five different source systems? Give the agent access to those systems via tool usage and let it query the data dynamically, instead of building slow integrations or copying and pasting manually.

This vision was science fiction a year ago. Now, with LLMs, RAG, and tool use, the future becomes clear: agents will 10x a team’s impact by augmenting their work and radically improving the quality of the customer experience.

Consider a founder’s day in this new reality, experiencing the profound multiplier this orchestration layer provides. When the founder logs in, an agent has been digging through databases to identified and qualify new leads based on a deep understanding, surfacing profiles that perfectly match ideal customer criteria. The founder reviews these insights and, delegates an agent to draft a highly personalized outreach for a top prospect, recapping their specific problem, outlining its implications, and proposing a tailored solution. However, perhaps this buyer isn't quite ready yet; the founder simply hands off to an agent that follow-ups regularly, every two weeks, ensuring they stay top of mind with relevant updates without any further manual effort.

When the deal closes and it's time to onboard, an onboarding agent seamlessly handles all the basic tasks, like ensuring documents are signed, sharing the company's security policies, and providing all necessary documentation the customer needs to get started. Even after the customer starts to use the product, no matter how small the account, and agent continuously monitors product usage or buying behavior. If a customer becomes less engaged, the agent automatically flags it and proactively reaches out to schedule a check-in, allowing the founder to understand what’s happening and intervene.

This is the kind of customer experience that builds loyalty and advocacy, driving the rapid scale required to grow 5x faster than in the pre-AI era.

Rebuilding the Stack​

Getting there is challenging, and retrofitting AI into legacy CRMs is like bolting a jet engine onto a bicycle. These systems weren’t designed with autonomy or intelligence in mind. They’re structured around static data models, hardcoded workflows, and manual input. Trying to add AI into that foundation limits what agents can actually do.

In the age of AI, we need to redefine what a CRM is. Traditionally, a CRM has been a system of record: a collection of tables and fields designed to store data. This static, ledger like approach is fundamentally orthogonal to how agents should be designed. Instead, the CRM will evolve from a mere database into the central operating system for customer, product, marketing, and insights. It becomes the ultimate orchestration layer, with agents constantly updating between systems, pushing and pulling data dynamically. This is no longer just a system of record; it's the intelligence hub that powers your entire customer facing operation.

Building an agent native CRM platform from the ground up unlocks a different kind of architecture; one where agents are first class citizens and core to the system. Agents can natively interact with objects, orchestrate tools, and adapt workflows in real time. Instead of being limited by rigid permission structures and fixed data schemas, we can build a dynamic context layer that reasons over emails, notes, calls, and activity history.

Crucially, this new architecture must embrace composability to maximize the benefits of code generation, advanced tool use, and the Model Context Protocol (MCP). The future of software is moving away from monolithic SaaS as the primary value layer. Instead, orchestration will be the new value layer. Users won't just configure; they'll create. This means that instead of spending millions on costly, lengthy sales implementations, users will be able to simply prompt new workflows and customize experiences to precisely meet their customer and business needs. The value will no longer reside in a static, functional SaaS layer, but instead in the dynamic orchestration layer that will allow users to truly customize and adapt to every unique customer interaction.

Rather than relying on heavy, hard coded integrations, we treat tool execution as a native primitive , so agents can query APIs, submit forms, or write updates across systems with real authority via a composable MCP first tool layer . This enables agents to call APIs, write to databases, compose customer specific tools, and take action directly. It creates a full feedback loop: agents don’t just track or suggest; they observe, act, and learn based on interactions and outcomes. For example, agents should have the ability to run thousands of micro experiments each week (from messaging to timing) continuously refining customer segments and profiles. This level of dynamism is simply impossible in legacy CRMs that require admins to design and maintain complex workflows.

Rebuilding the stack also shifts the economic model. Legacy CRMs are optimized for large seat based contracts, services revenue, and long implementation cycles. A modern CRM for AI first teams is built for small, fast moving orgs, where the value isn’t in how many people you add, but how much each person can do. That means outcome first design, not usage first pricing .

The incumbents will bolt on copilots to old interfaces. We’re building the operating system for modern revenue teams , where agents aren’t an add on, they’re the engine.

Product Beliefs​

1. Agents are autonomous and goal oriented

Agents should not follow static, prescriptive, step by step processes, but autonomously solve problems. Users set goals, principles, and boundaries; agents reason to identify the path forward.

2. Asynchronous orchestration by default

Agents operate independently based on schedules, triggers, and context. Humans are pulled in only when needed (e.g., for approvals or exceptions). We minimize bottlenecks and optimize collaboration.

3. Permissions as a feature

Access control isn’t static. Agents should be able to request, justify, and escalate permissions dynamically. If an agent can't complete a task due to access limitations, it should be able to ask for temporary scope (e.g., single use permission for tool) or route requests to more privileged peers.

4. Agent workspaces

Agents need a shared space to collaborate, delegate, and escalate. Think of it as a virtual office where agents plan amongst each other, assign tasks, or ask other agents for help.

5. Experimentation is core to progress

Agents must be allowed to experiment across variants, channels, and timelines. This includes layered feedback mechanisms: short term results, long term signals, and cumulative learning. For example, a campaign agent should track open rates, but also sales outcomes weeks later.

6. Programmable memory and state

Agents have layered memory: a short term context window (what I’m doing now), a manifest of generalized know how (how things usually work), a state map (where we’ve been and how it went), and long term memory stored as a knowledge graph and retrieved via RAG. Memory must be editable; sometimes slowly over time, and sometimes instantly after a breakthrough or error.

Building For Founders​

My initial focus is on founder led GTM teams in sub $10M ARR SaaS startups , partly because I understand their problem, but also because they are motivated to move at speed and take risks in search of differentiation and alpha. These typically are high agency teams that don't have time to waste. They’re not trying to configure Salesforce; they’re trying to get the next 100 customers. What they need isn’t another tool to manage, or introduce process to slow down velocity, but something that acts and drives pipeline. They want leverage, not overhead . Chuff is built for people who prefer shipping product to planning, who value outcome over process, and who treat software like a teammate, not a reporting system.

Where I’m Headed​

This is just the beginning. We’re building a new kind of CRM, one that replaces static forms and workflows with autonomous agents that think, act, and grow alongside your team. We have a bold vision: to become the operating system for revenue, powering every conversation, insight, and customer touchpoint with intelligence and intent. The ambition is aggressive, because the opportunity demands it.

To get there, we’re assembling a world class team of builders, operators, and founders who want to redefine what software can do. The path ahead is fast and high stakes. We’ll ship fast, learn publicly, and compete hard. If we get this right, we won’t just improve CRM; we’ll replace it.

AI agents are going to reshape how businesses interact with their customers. The promise is real: faster resolution, lower cost, scalable conversations, and systems that can operate 24/7. But there’s a growing risk that’s not getting enough attention—the temptation to over-agentify everything.

There’s a fundamental flaw in how most agentic systems are built today. Too often, agents are treated like glorified workflow engines, designed as a series of step-functions that chain together discrete reasoning tasks. One step labels an email, another analyzes a document, another triggers an action.