squad

Landra AI

A Solo Founder, a Full Product Team, and the Final Say on What Ships

Ezra Rufino, Founder · getlandra.com · Updated

Squad makes me 10x more effective at running my business.
Ezra Rufino, Founder, Landra AI

A customer writes into Landra's Intercom chat with a problem. Ezra Rufino does not read the message first. One of his agents does. It works out whether the problem is a bug, a confusing screen, or a bad answer from Landra's own AI chat, drafts the fix, builds it, and opens a pull request. It also drafts the reply to the customer.

When Ezra gets to it, there is nothing left to investigate. The fix is already in a pull request, and the reply is waiting for his yes or no.

10x

more effective at running the business, in his own estimate

19

AI agents working for one founder

105

pull requests merged in 5 weeks, each reviewed and tested

93 to 116

pages on the examples swipe file, a new batch every week

1 a week

YouTube video, scripted and built into a full deck

Sep 2026

on Squad since

The challenge: every department, one founder

Landra is a software product, so it has the problems of a whole company. Customers write in. Bugs show up. The AI that writes pages for customers sometimes gets an answer wrong. Search traffic needs new content every week. Videos need scripts and slides. Someone has to watch the numbers.

Ezra is one person. Each of those jobs is reasonable on its own. Together, they are more than one founder can keep track of, let alone do well.

Why a squad, and not one more AI session

Ezra already builds with AI coding tools. What he wanted was a process he could trust on bigger work, not just a faster way to write code.

A squad is a team of AI agents with separate jobs, not one assistant waiting for the next prompt. That separation is what made the difference for him on engineering work:

Because my Squad works as a team too, for larger dev projects I feel more confident in the work. Because I'm able to have one dev build it, another code review and security review, another run e2e testing and sign off on it.

Ezra Rufino, Founder, Landra AI

He compares it to running the same project in a single Claude Code session and then testing every edge case himself. With a squad, the checking is somebody's job.

The checking does not even run on the same AI. The agents that write Landra's code run on Codex. His senior engineer agent plans the work and does the code review and security review on Claude Opus.

The setup: a chief of staff and a team of specialists

Ezra's lead agent is his Chief of Staff. It manages the other agents and his projects, helps with his calendar, and sends him a briefing every morning: the previous day's numbers, how Landra is tracking against its goals, and the priorities for him and the team.

It is his favorite agent, and the reason is simple: "rather than having to always keep tabs on everything myself."

Around the chief of staff sit the specialists. One reads everything customers say and turns it into fixes. One runs SEO and content. A social media agent works with him on videos. Engineering work moves between agents that build, review, and test.

The list keeps growing. In a post on X, he added cold emails, weekly newsletters, Reddit posts, bookkeeping, and carousel posts cut from his videos.

It did not all click on day one. He wrote that it "took me a minute to figure out exactly how to get the most out of it." Now he uses it, in his words, "like crazy."

Workflow one: from customer message to pull request

  1. The trigger. A customer writes in through Intercom or the support inbox. The same agent also keeps reading Landra's websites.
  2. The work. An agent reads the websites, the Intercom chats, and the customer service emails. It flags bugs, UX problems, and weak answers from Landra's AI chat. Then the squad drafts the bug or feature change, builds it, and opens a pull request.
  3. The handoff. Ezra gets the pull request and a drafted reply to the customer, in Intercom or in his email.

No reply reaches a customer without him. In his words, "I get to decide whether or not to send it."

Three lanes, sorted by risk

Not every change waits for Ezra. Every change goes into one of three lanes, based on what could go wrong.

  1. Ships without him. Blog posts, SEO pages, the weekly examples batch, small bug fixes, and small behind-the-scenes fixes. Ezra hears about them afterwards. Code in this lane still needs a code review, a security review, a QA pass, and passing tests before it ships.
  2. Waits for his approval. Visible product changes, like layout or onboarding copy. Ezra approves them from before and after screenshots in Slack.
  3. Waits for him to test and merge. Templates, sign-in, billing, database changes, and customer emails.

That split is what lets a solo founder approve the risky work without becoming the bottleneck for everything else.

Workflow two: the SEO engine

Landra already uses Outrank for blog writing. Ezra wanted more than that: more control over what Landra writes about, and a second pass on the posts Outrank produces.

  1. The trigger. The squad runs a standing process for SEO writing, from finding new topics to building out new posts.
  2. The work. Agents pick topics, write posts, and review the posts from Outrank, making changes where they are needed. Every week, one agent also adds new pages to Landra's examples page, a swipe file of DTC landing pages that Ezra calls one of the most important parts of the site for SEO.
  3. The handoff. New content and new examples, without Ezra writing them himself.

The examples page is the part he highlights: it gets a lot of traffic, it is directly relevant to what Landra sells, and it only works if it keeps growing. It went from 93 pages in mid-September to 116 by early November, with 23 new teardowns added in weekly batches.

Since mid-October the squad has also shipped 10 new blog posts, a free guides section with 6 guides, and rewrites of 8 older posts, plus a monthly changelog post that goes out with Landra's product update email.

Now using [Squad] to automate the process of finding landing pages and adding them to our /examples page. Very effective.

Ezra Rufino, on X

Workflow three: a partner for videos

Not everything is automated. Ezra's social media agent works more like a collaborator. They go back and forth on ideas and scripts for his YouTube videos, and the agent builds entire slide decks for him to walk through on camera.

Since October 20 that has been a steady rhythm: one YouTube video a week, scripted and built into a full deck, which Ezra records on Mondays. The same agent also scripts six short videos for him every Monday.

The results

Ezra does not measure the change in hours. He measures it in how much a solo founder can keep moving.

I'm able to manage a lot more than I ever would have in the past. Everything on the marketing side and continuing to make sure we're improving the product and responsive to customers.

Ezra Rufino, Founder, Landra AI

He also points to the volume of work: "We've shipped a huge number of things to be honest."

The numbers back that up. At the start of October his squad moved to reviewed pull requests. In the five weeks after that, it merged 105 of them, 47 of them fixes. Lately it is 22 to 30 a week.

Before that, Ezra pushed straight to main himself, about 35 to 40 commits a week, with no separate review or QA step. The pace stayed high. What changed is that every change now gets checked by someone other than its author.

One result was money, not time. His chief of staff now handles chargebacks. On a $79 chargeback, it wrote the response, Ezra said yes to fighting it, and it submitted the case. Landra won. It is the only case decided so far.

What makes it work

Three things in how Ezra runs his squad carry over to almost any small company.

The agents prepare, and he keeps the customer-facing call. Customer replies arrive as drafts, and he decides what gets sent. Code goes through review and testing by other agents before it ships. The work is done before he looks at it.

Checking is a separate job. On larger projects, the agent that writes the code is not the agent that reviews it, and neither is the one that tests it. That is the same reason human teams separate those roles.

One agent keeps track of the rest. With a team this size, the hard part is knowing what is going on. The chief of staff's morning briefing and project management mean Ezra does not have to hold the whole picture in his head.

If you want the same setup

Start with the place where your customers already talk to you. Write down what should happen when a message comes in: who reads it, what counts as a bug, and what you want waiting for you at the end. That is the brief for your first agent.

Then add the check. Give review and testing to a different agent from the one that does the work, and keep the final send for yourself.

Figures in this story are Ezra's own, as he and his chief of staff reported them in November 2026. Quotes marked "on X" are from his public posts in October 2026.