Key Takeaways
- Gavin Baker built 17 specialized Grok bots at Atreides Management to handle distinct, continuous investment research workflows.
- Baker previously attempted to glue together custom research scripts with coding tools, but abandoned the effort because turnkey agent bots eliminated user friction.
- Asset management succeeds or fails on information routing: ensuring the exact right data reaches the right decision-maker at the exact moment of decision.
- John Coogan pointed out that AI agents unlock deep research at 10 p.m. on Sunday nights without crossing managerial boundaries with junior analysts.
- Baker maintains an internal culture where team members can call him anytime on weekends, but he avoids calling downward; autonomous bots eliminate that operational bottleneck.
Atreides Traded Custom Code for 17 Specialized Bots
Most investment firms treat AI as an enterprise search bar. Gavin Baker took a different route at Atreides Management. Instead of relying on a single general interface, he created a fleet of 17 distinct Grok bots, with plans to add more as new workflows emerge.
Baker did not start there. He tried building custom setups first. “I clued together all this stuff with cloud code and it was really, I'm sure it was user error on my part,” Baker said. “But Grokbot, it just makes it so easy. I think I'm up to like 17 Grok bots and every day I think of a new one to add.”
He calls it the single most effective AI tool his firm has deployed. The difference comes down to specialization. Instead of writing multi-paragraph prompts every morning to guide a general model, Baker assigns one narrow, recurring research objective to each bot. One bot monitors specific filings, another tracks supply chain changes, and another watches executive commentary. Each runs constantly in the background.
Removing the Managerial Friction from Late-Night Research
Every founder and fund manager knows the Sunday night dilemma. You find an anomaly in a balance sheet or a sudden shift in customer churn at 10 p.m. You want three hours of data compiled before the market opens on Monday.
Calling a junior analyst at that hour destroys firm culture. As John Coogan observed, “It can be 10 p.m. on a Sunday and you can kick off a real research effort that, yes, you could probably call someone on your team and say, 'I need you to,' but that would be crossing a professional line.”
Baker agreed, pointing out his own management boundary: “Well, my policy is actually anybody in Atreides is always welcome to call me anytime on the weekend. I try not to call people on the weekends.”
Autonomous agents solve that operational tension. They allow a manager to execute an aggressive, spontaneous inquiry at midnight without burning out human talent. The team returns on Monday morning to find the answers already compiled, without having worked an eighty-hour week.
Information Routing Beats Raw Data Crunching
Many investors assume the value of machine learning lies in crunching giant datasets faster than an Excel model. Baker views the real problem differently: asset management is an information routing failure.
“There is so much of investing that is making sure that the right information gets to the right person at the right time,” Baker said. “And Grokbot is so good at that.”
When a portfolio manager misses a material shift, it rarely happens because the data did not exist. It happens because the filing, tweet, transcript, or price signal sat unread in a queue of hundreds of daily alerts. Specialized bots act as intelligent filters. They do not just summarize text; they direct signal straight to the person responsible for allocating capital to that exact asset.
What to Do With This
List your top three recurring Sunday evening questions. Build three separate bots dedicated solely to those specific inquiries instead of relying on a single general chat window. Let the bots run your weekend prep so your team never receives a late-night work request again.