Inside the AI agent stacks at SaaStr, Owner.com & Klaviyo
Three companies at three different scales rebuilt their operations around AI agents in 2025-2026. Here's what each one is actually running.
At SaaStr AI 2026, three companies walked through exactly what they've built on top of AI agents: not the marketing slide, the real stack. SaaStr is a sub-10-person media and events operation. Owner.com is a $100M ARR business. Klaviyo is a $1.4B+ public company. The approaches differ. The direction is identical.
SaaStr runs 20+ specialized agents handling marketing, customer success, events, inbound, outbound, and dead-lead revival, built on headless Salesforce. Owner.com leads with a free AI product that converts 83% of new customers before any human touches the account. Klaviyo is rebuilding its product development process from the inside using agents. All three have shifted from humans doing routine work to agents doing routine work with humans overseeing output.
SaaStr: two humans, one dog, 20+ agents
The full SaaStr go-to-market team is Amelia Lerutte (Chief AI Officer), Jason Lemkin, David on sponsor sales, and Ginger the dog. The rest of the work is agents.
None of these agents were designed as agents on day one. They became agents through 600 to 1,000 commits each, seven to eight commits a day, over a few months. That cadence matters: the agents that work best are the ones that got iterated constantly, not architected upfront.
Here's what each one does:
10K (AI VP of Marketing). Built on Replit, first commit January 2026, around 1,000 commits, 18K+ lines of code. Started as a dashboard to stop copy-pasting numbers from Marketo and Salesforce into Notion. Now owns daily revenue tracking, forecasting, campaign performance, and pushes three new marketing ideas per day via Slack and email.
QBee (AI VP of Customer Success). Started as a project management replacement. Now manages 150+ sponsors with personalized email outreach, asset collection, and real-time risk flagging. QBee doesn't have full Salesforce integration yet and already outperforms 85% of human CSMs at force-ranking sponsor health.
Annie (event producer agent). Rebuilt on Replit in November 2025 from a basic Squarespace site. Now has 46K+ lines of code and runs the parking pass app, the agenda, the attendee newsletters, and active website visitor targeting.
Amelia AI (qualified inbound). The most-trained agent in the stack. 2.2M sessions, 442K chats, 614 booked meetings, and roughly $85K average sponsor ASP this year. Replaces three BDRs SaaStr would otherwise need to hire.
Agent Force (dead lead revival). Runs inside Salesforce. Has the highest open rate of any agent in the stack because it has the most context on each contact.
Ava/Artisan (warm outbound). Handles past attendees, past sponsors, and lapsed contacts that humans wouldn't prioritize. Recovered roughly $500K of sponsor revenue this year.
Monaco (cold ICP look-alikes). Pulls close-won history, builds look-alike accounts, books meetings without a human touching it.
The connective tissue across all of these is headless Salesforce. None of the agents would function the way they do if they had to go through the Salesforce UI. They use the API directly, in real time.
The pattern worth noting for operators in our audience: SaaStr's agents started as boring internal tools. A dashboard. A project tracker. A website. The "agent" label came after hundreds of commits, not before. If you're waiting to find the right agent use case, you're probably already sitting on it.
Owner.com: build the free AI product, then bundle from there
Adam Gild was at SaaStr three years ago talking about being a Shopify for restaurants. He came back this year with $100M ARR on the horizon and a stat that reframes how they got there: 83% of new customers start their journey by using an AI product before a human ever touches the account.
The pivot in early 2023 was the company. Here's what they built:
Gradr (free AI restaurant website generator). Got 2M+ views on X in two weeks. Costs Owner roughly $1 in compute per restaurant. Free for the first three months, then $1 per month. A restaurant owner types in their name and within five minutes the agent has crawled their Google Business Profile, pulled every nearby competitor and review, run an AI photo shoot, and produced a functioning website.
The economics of Gradr are the point. At $1 per restaurant in compute, Owner can give away the product, let it demonstrate value before any sales conversation happens, and then sell the bundle. The conversion from free AI product to paid customer is where the 83% number comes from.
This is a distribution strategy dressed up as a product strategy. The free AI product fills the top of funnel faster than any paid acquisition channel could, and it pre-qualifies buyers by getting them to experience value before they see a price. The operators we work with who are trying to figure out LinkedIn inbound are running a version of the same logic: demonstrate before you pitch.
What Owner.com built at scale with Gradr, operators can approximate on LinkedIn through consistent, specific, expertise-driven content that solves a problem before anyone's asked for a proposal. The free-value-first mechanic is not unique to $100M ARR companies.
For a closer look at how founders are building this kind of inbound engine, see how founder-led brands are driving LinkedIn inbound without relying on ads or cold outbound.
Klaviyo: rebuilding product development from the inside
Klaviyo is a $1.4B+ public company. They're not using AI agents to build new revenue lines. They're using them to change how the existing product gets built.
The internal shift at Klaviyo is about the product development process itself: agents embedded in the workflow that handle the parts of engineering and product work that slow humans down without requiring human judgment. Code review support, ticket triage, documentation, test generation. The work that fills the calendar but doesn't require a senior engineer's brain.
What makes the Klaviyo case interesting is the scale at which you have to operate before the ROI on internal process agents becomes obvious. For a company at $1.4B+ ARR, shaving 20% off engineering cycle time is worth more than most companies make in a year. For a 10-person startup, the same 20% improvement is meaningful but not transformative.
The implication isn't that small operators should ignore internal agents. It's that the right agent for a small operator probably isn't a code-review tool. It's a QBee (customer health monitoring) or an Amelia AI (inbound qualification) or a Gradr (lead generation through free product). The unit of work the agent handles should match the unit of work that's actually bottlenecking growth.
For operators building inbound pipeline on LinkedIn specifically, the bottleneck is almost never content creation. It's finding the right posts to engage with, at the right time, before the comment section gets crowded. That's a job AI can help with. See how LinkedIn inbound signals work in practice.
What these three companies have in common
Beyond the surface-level "they all use AI agents" story, three things cut across SaaStr, Owner.com, and Klaviyo:
They started with the boring problem. QBee started as project management. Annie started as a static website. Gradr started because restaurants needed a web presence and Owner saw a cheap way to give them one. Nobody built a "VP of Marketing Agent" from scratch. They automated the thing that was already annoying, then kept committing.
They're all running agents on top of their real data. SaaStr's agents are useful because they have full Salesforce context. Owner.com's Gradr works because it has access to Google Business Profiles, competitor reviews, and real photos. Klaviyo's internal agents work because they're embedded in the actual product development workflow. Agents without proprietary context are marginally better than a good prompt. Agents with proprietary context are genuinely hard to replicate.
The agents that work are the ones that got iterated. The common thread in the SaaStr stack is commit count. 600 to 1,000 commits per agent, over months. That's not a deploy-and-forget system. That's a product. The operators who will get the most out of AI agents in 2026 are the ones treating them like products, not experiments.
The piece of this that matters most for operators building LinkedIn inbound: the same principle applies to content strategy. The accounts that build durable inbound on LinkedIn are the ones that iterate on what works, week over week, rather than the ones that have the cleverest initial strategy. We've seen that pattern across the profiles we audit. The ones with the most traction are rarely the most sophisticated at the start. They're the most consistent over time.
For a structured look at what consistency produces on LinkedIn, Justin Welsh's LinkedIn strategy is one of the clearest examples in our dataset of what 600+ iterations on a content approach actually produces.
Frequently asked
SaaStr has 21+ agents in production as of 2026, handling functions across marketing, customer success, event production, inbound qualification, warm outbound, and cold prospecting. The stack is built on top of headless Salesforce, with each agent using the API directly rather than the Salesforce UI.
