Most sales reps lose hours every day to work that has nothing to do with selling: drafting the same kinds of email replies, hunting for account context before a call, writing up notes after one. A recent case study from Anthropic shows what happens when one of those reps stops treating that load as fixed. Jared Sires joined as an account executive in 2024 with no coding experience. Within a year he had built his own assistant on top of Claude, and the rest of his team adopted it inside a day.
What one rep actually built
Sires was carrying a book of 600 to 700 accounts, taking 10 to 15 customer calls a day, and answering email late into the night. Using Claude Code, he built a Gmail-resident app he called CLAFTS, short for Claude Drafts. It reads an incoming email, pulls the relevant context from shared Google Drive, internal tools, and public documentation, then writes a reply in his own voice. The application runs to roughly 4,300 lines of code, almost all of it written by Claude rather than typed by hand.
He did not set out to become an engineer. He described a problem, reviewed what Claude produced, corrected it, and repeated until the tool worked on real emails. After he shared it in a team channel, the wider sales organisation picked it up within 24 hours. He has since moved into a role building Claude-powered tools full time and packaging them as a Cowork plugin so colleagues can install them without ever touching the code.
Why this matters for Australian sales teams
The detail worth sitting with is not the app itself. It is that the person who built it could not write software a year earlier. The barrier that used to sit between a sales team and custom tooling, a developer with spare capacity and a budget to match, has largely gone. A sales manager in Sydney or Melbourne can now describe the workflow they want and have a working version in days rather than quarters.
The patterns Sires built are not exotic. Most sales teams run the same handful of repetitive jobs, and each one maps cleanly onto something Claude Code can assemble:
Inbox triage and drafting: replies written in your voice, with the account context already pulled in.
A daily account brief: each morning, a ranked list of which accounts to focus on based on usage or engagement signals.
Pre-call research: a short, current summary of an account assembled before each meeting.
Follow-up from transcripts: action items and a draft follow-up email generated straight from a call recording.
None of these requires a data science team or a long procurement cycle. Each is a small tool that does one job well, built and refined by the person who actually feels the pain. That ownership is the part that travels well to a small Australian sales team, where the person with the problem and the person who could fix it are usually the same person.
The real cost of admin time
Put a number on it. Sires estimated the inbox tool alone saved him two to three hours a day. On a typical Australian sales salary, that recovered time is worth roughly $25,000 to $40,000 a year per rep, before you count the deals that close faster because follow-ups go out the same day instead of three days later. For a team of ten, the arithmetic moves into the hundreds of thousands of dollars a year. The build cost, by contrast, is measured in days of one person's time.
The figure assumes the time saved goes back into selling rather than quietly disappearing, which is a management question more than a technical one. But even at half the estimate, the return on a few days of building is hard to argue against. The teams that see the full benefit are the ones that decide upfront what the recovered hours are for.
How to start without a developer
If you want to copy the pattern, the order of operations matters more than the tooling:
Pick one job that eats time every single day. Inbox replies and pre-call research are the usual first wins.
Write down what good output looks like, in plain language, with two or three real examples to test against.
Build a rough version with Claude Code, review what it produces, and correct it until it holds up on real cases.
Share it narrowly first, then package it as a Cowork plugin once it has earned trust across the team.
The teams that get value from this treat it as a series of small, owned tools rather than one large platform project. If you want help scoping a Claude Code build for your sales team, or working out which workflow to start with, that is the kind of work we do at Automata AI. You can book a short call to talk it through on our contact page.
This piece is based on Anthropic's published case study of how its own go-to-market team uses Claude.
