Daisy Hollman works on the Claude Code team at Anthropic. At her session at Anthropic's Code with Claude conference in London, she gave a room of engineers one instruction.

Build copies of yourself. "Your job as a software engineer at this point is to make little clones of yourself so you can scale up your abilities and scale up your work across many agents," she said.

Every stuck founder is running through much in their one brain. You answer the messages. You fix the problems. You check the output. Then you wonder why the calendar never clears. Claude can take a large share of that load. You run it like a chatbot when it could run like a colleague. Hollman's session shows what to do differently.

How to make Claude take on more of your job: advice from the Code with Claude conference

Give it the same access you have

Out of the box, Claude sees a fraction of what you see. "If Claude can't do everything you can do, it can't do your job with you," Hollman said. Your decisions live in Slack threads, email chains, shared docs, and dashboards. Connect Claude to one of them and leave out the rest, and it can only guess at the parts you never shared.

Hollman runs a simple test. Spend a full day working without leaving Claude. "Every time you have to reach for another tool, every time you have to alt-tab to something else and copy-paste into Claude, that's something Claude is missing," she said. Write down each gap, then wire Claude into those sources. "The gap is much bigger than you notice until you make all of the connections."

Hand it the reason behind the task

Claude produces better output when it knows the reason behind a task. After internal meetings, Hollman feeds the notes straight in. "I will go right after the meeting and feed the meeting notes into Claude and say, is there any low-hanging fruit from this meeting that you can address? And I'll get two or three PRs per meeting," she said. A PR (pull request) is a finished change packaged up and ready to ship to the product.

Do the same with your own calls. Drop the transcript or your notes into Claude and ask what it can action while the conversation is fresh. The constraint you forgot to mention, the reason a client wants something a certain way, the history Claude cannot see, all of it produces better results.

Build a tighter feedback loop

Founders re-prompt when output disappoints. Hollman points somewhere else. "The fastest way to make your agent better…is a tighter feedback loop," she said. Claude improves when it can check its own work against a defined set of standards.

Give it the raw material to do that. A few examples of strong output, the rules you apply, a checklist it runs before handing anything back. Tell it to review its draft and fix what falls short. The correction loop makes a difference.

Let it run while you sleep

Claude does not need you watching. Anthropic shipped a command called loop that keeps a prompt running on a timer. "It literally just runs a prompt every fixed interval of time," Hollman said. Set a task going at the end of the day and let it work. "Once it gets to CI (continuous integration), even if your CI takes two hours to run, you can just leave it for the next day and a half, and it'll fix all of the CI bugs." CI is the automated testing that runs on new code.

One engineer on her team moved through roughly 1,000 PRs in a single month this way. You check in from your phone. "It is a great way to do a 30-second check-in after dinner to make sure that your agents are still running overnight and aren't stuck on something dumb," she said. Making progress while you rest is the closest thing to cloning yourself.

Stop doing the parts it does better

Some of your daily tasks belong to Claude now. "You should not be fixing CI failures yourself at this point in time. Agents are very, very good at that," Hollman said. The same logic covers any repetitive job you do from habit.

Write the list. The follow-up emails, the formatting, the first-draft research, the status updates, the reconciliations. Each one you keep doing by hand is a clone you didn't make. Pass them over.

Make Claude do more of your job for you: according to Anthropic

What changes when Claude does more of your job? Access, reasons, a feedback loop, work that runs overnight, and the discipline to step back. Those five moves turn Claude from a tool you prompt into a colleague you direct. The founders getting ahead build copies of themselves while everyone else keeps doing every task by hand, freeing the hours for the dream life they started the business to fund.

Grab my free AI playbook for founders to put one of these into practice this week.