Key Takeaways
- William Imoh gives autonomous agents unrestricted read and write access to his email, calendar, Linear, and Notion, restricting only raw customer data and company bank accounts.
- When agents hit integration walls, Imoh bypasses manual approval by turning on autonomous computer use and selecting "allow always" on browser prompts.
- Imoh replaced traditional recruiting software with Three Pipes, an internal database where an agent processes candidate calls, extracts notes, and scores applicants without human data entry.
- Imoh argues software is consolidating into custom, problem-specific systems of record managed entirely through conversational interfaces.
The Case for Total Access
Most founders treat AI tools like untrusted interns. They sandbox permissions, mandate human approval for minor edits, and spend half their workday copying context between tabs. William Imoh took the opposite approach.
His reasoning is practical. An autonomous agent stripped of write permissions is just a search bar with latency. If you force an agent to ask permission before every calendar update or task creation, the cognitive overhead defeats the purpose of running solo operations. Imoh draws two hard boundaries: the agent cannot touch raw customer data, and it cannot touch the company bank account.
“Now, I know it is counterintuitive in some parts and some conversations that you want to be skeptical about what AI has access to and what it doesn't, what it has read or write access to,” (play from 30:32) Imoh says. “For me, I just let all that fly out the window. The only thing it doesn't have write access to is the bank.” (play from 0:38)
When standard Model Context Protocol (MCP) connections fail, Imoh does not stop to write new API integrations. He activates computer use and browser access inside ChatGPT Codex. When prompted to approve individual actions, he selects "allow always." The machine runs his desktop so he does not have to.
Software Shrinks to Databases and Prompts
Full agent access exposes a quiet shift in how internal software works: point-and-click user interfaces are becoming dead weight. Imoh argues that SaaS categories like traditional recruiting tools exist only because humans need graphical buttons to update rows in a database. When agents can read and write across tools, the interface disappears.
“I have a firm belief that we are going to be left with systems of record, like a custom database that is designed around a problem,” (play from 32:18) Imoh says. “That's what is going to be left, and an AI interface.” (play from 32:27)
To prove it, Imoh built Three Pipes, an internal hiring tool for Hackmamba. He never logs into it to enter data. After an interview, he sends a single command to his bot fellowship: “I just spoke to this person, get them into Three Pipes.” (play from 32:49) The agent scans his calendar for the event timestamp, pulls meeting notes, searches his inbox for the candidate's resume, populates the custom schema, generates a candidate summary, and calculates an initial fit score. The database receives the record without a single manual form submission.
The Trade-off Founders Ignore
Surrendering control carries an operational trap. When an agent manages your calendar, triages your email, and assigns tasks in Linear, you gain founder leverage at the cost of operational visibility. If an agent misinterprets an email thread and cancels a sales meeting, you will not discover the error until the prospect disappears. Full read-write permission works only if you run a solo operation or an agile team where the blast radius of a mistake is small and easily reversed.
What to Do With This
Pick the internal workflow where you waste the most time doing repetitive administrative data entry, such as logging sales notes from your calendar into your CRM. Build a lightweight custom database schema in Notion or Postgres, grant your agent API or browser write access, and automate the sync from your call transcript. Keep human approval on anything involving cash transfers or contractual commitments, and remove human confirmation for everything else.