How Anthropic Uses Claude · Field case

The person who rebuilds the workflow can't code

A seller who had never opened a terminal built a 4,300-line tool with Claude Code and rebuilt his team's entire GTM workflow. Months later, 80% of the sales org runs on what he made.

4,300 lines of code, almost all written by Claude Code
by someone who'd never written a line before
Jared Sires AE → GTM Architect claude.com/blog · 2026-06-05
Following Part 1

Where did those Skills come from?

The first piece was Eleanor Dorfman on the SaaStr stage, looking down at how Anthropic rebuilt its whole sales org around Claude: keep the stack, make Claude the connective tissue between tools, turn Slack into a single front door for support functions, and, crucially, encode top-rep patterns as Skills so every new hire gets the whole set on day one.

That piece left one question in a single passing line: those "top-rep patterns" that became Skills, who built them, and how?

This is the answer underneath that line. And it's a little counterintuitive. Not an engineer. A seller who had never opened a terminal.

The grind

A sales rep buried in admin

Before joining Anthropic in 2024, Jared Sires had never written a line of code. And why would he? He was a startup account executive.

As is often the case at fast-growing companies, his book quickly grew to 600 or 700 accounts. With 10 to 15 customer calls a day and an expanding list, he found himself at his desk answering emails until 9 or 10 p.m. every night. "It was almost impossible to manage my inbox," he says. "And doing outbound on top of that, you don't really know where to focus."

600–700
accounts in his book
10–15 /day
customer calls
9–10 pm
how late he answered email
24–48 h
cadence of Anthropic product changes

Email volume wasn't the only problem. Anthropic ships product changes every 24 to 48 hours, and customer questions tend to land on the latest details: batch API SLAs, prompt caching discounts, model pricing, SDK behavior. Answering them well meant searching across Slack, Google Docs, internal knowledge bases, and the developer docs, day after day, with a slightly different set of facts each time.

His first experiment with Claude was narrow and practical: using Apps Script (Google's lightweight dev platform) plus Claude, he pulled product usage data from internal systems and had Claude rank his accounts each morning by how fast they were growing. With 700 accounts, the daily brief told him where to spend his outbound time.

CLAFTS · Claude Drafts

The inbox, and those late nights

The harder one was the inbox. With Claude Code, he started building something that drafts replies to customer emails in his voice. "Claude Code, having the terminology 'code' at the end of it, made me feel a little bit intimidated just to even start," he says. "But after a certain time frame, I understood the power of it being able to hook up to my computer and answer things about files on it."

That something is CLAFTS (Claude Drafts): an app that lives inside Gmail and uses the Claude API to draft replies. It pulls context from a shared Google Drive folder and other third-party tools, references Anthropic's public documentation through web search, and matches his writing style. By the time he opens his drafts folder at the end of the day, the responses are waiting for review.

Gmail · CLAFTS Draft · awaiting review
Tocustomer @ acme.co
ReHow does the prompt caching discount work?
Re: pricing on prompt caching

Hi Dana,

Good question. Per our latest docs, cache writes bill at 1.25× the base input rate and cache reads at 0.1×, so once you're hitting the cache, the cost of those tokens drops sharply…

(Drafted in my voice, facts pulled from today's public docs, waiting for one review pass before it sends.)

context from Google Drive web search · docs writing style ⌘ review → send

Here's the detail that's easy to miss, and it's exactly where sales tools usually break: whenever Anthropic ships a change, the docs update, and Claude picks it up on the next draft through web search. "Claude is able to use web search to understand our latest documentation from our website and reference that material when generating emails," Jared says. "I don't need to keep all of that in my head." What the customer gets is always tied to whatever shipped most recently, not to whatever he happened to remember.

And it wasn't effortless. Out of the box, Claude's writing ran long and hedged, so he reworked the system prompt until the drafts matched his own voice.

4,300
lines of code, almost all written by Claude Code
hundreds
of system-prompt iterations to match his voice
10–15 hrs/wk
CLAFTS saves him every week

As for impact, the shift he cares about most isn't the hours, it's accuracy. "Before CLAFTS, I felt like I was doing more administrative work than actually spending time with customers," he says. "After CLAFTS, I was actually able to do more of what I wanted to do, which is sales."

CLAFTS Tones

Even the guardrail, he found by accident

Next he built CLAFTS Tones, which uses pattern matching to mimic his voice across different relationships, customers, peers, family. To test it, he wrote himself a sequence of increasingly angry emails on his personal account. Claude picked up the tone, then refused to keep going.

test draft · tone rising
"I'm a bit disappointed by the delay here and would appreciate a quick reply."
test draft · sharper
"This is the third time now, and I'm finding it really hard to accept."
test draft · full anger
"This is completely unacceptable, I demand…"
CLAFTS · Claude
Got your tone, but I won't write this one. Generating angry emails to customers isn't something I should do.

"That was when I knew CLAFTS Tones was working." — Jared Sires

Diffusion

How it spread on its own

CLAFTS wasn't a top-down program rolled out by memo. He shared the first version in Slack the next morning, and within 24 hours others in the sales org had started using it with similar results. One co-builder was John Albert, a BDR regularly working past midnight emailing customers. The rest of the business development team came on board once they saw their teammate getting hours back each day. "From there, they did most of the evangelism themselves."

Jared 1 seller Slack within 24h teammates adopt it themselves 80% of the sales org within months · shipped as a Cowork plugin
one rep with the worst pain teammates use it on sight no top-down program

Then he kept building: two skills bookending his calendar, a daily brief and a daily recap. Each morning, the daily brief reads his calendar, runs a web search on whoever he's meeting, and produces talking points before the first call, connecting to Google Calendar and CRM through MCP servers. At day's end, the daily recap drafts follow-ups from Google Docs and meeting notes. "You couple those together and you get Claude managing your daily tasks, which essentially becomes an agent." He's now experimenting with the Agent SDK, chaining workflows where one Claude run feeds the next.

To scale these across the team he supports, he packages skills and MCP connectors into a Claude Cowork plugin anyone can install in minutes. Within months of launch, roughly 80% of Anthropic's sales org was using it. The remaining 20% are largely new hires, which he considers the next challenge, since the skills were built specifically to help people ramp faster.

/customer-context~90 sec
Pulls a 360-degree account view across every system below.
/pipeline-management
Surfaces at-risk deals, forecasting guidance, and progression recommendations.
A new hire installs the plugin on day one, with 20+ skills already wired to these tools
SalesforceIntercomGongGoogle CalendarGmailGoogle DriveBigQuery
The arc

From AE to GTM Architect

2024 · before
Account Executive
a startup seller who had never written code or opened a terminal
after CLAFTS
GTM Product Manager
finding problems in how sales operates and building Claude-powered fixes
today
GTM Architect
sits in design reviews with product engineers, shaping new sales tools

"I feel like with the technical barrier dissolving, I'm almost able to design more products and have senior engineers help me implement to the final stretch," he says. "I'm able to augment and do more things."

His advice to sellers wondering whether they could build something similar is simple: open Claude Code, find one task that's slowing you down, and ask Claude how to build a solution for it.

With Claude, I'm able to design and build things that don't just improve my own day-to-day workflows, but also those of my broader team. I have space to work more creatively and strategically, and there's no turning back.Jared Sires · Anthropic
Not a one-off

The bottleneck was never the code

Pull the camera back to Part 1 and Jared isn't the exception, he's a sample of a pattern. Eleanor's org-level playbook already had a line about encoding top-rep patterns as Skills. CLAFTS shows you how that actually happens: the pattern isn't "encoded" out of thin air, it's run first by the rep with the worst pain, on their own problem, then packaged and copied. Eleanor's "Create an Asset" Skill mentions a rep dropping a Gong transcript into Claude Code to spin up a custom prototype. Same thing. The person building the tool is a seller, not an engineer.

Put the two pieces together and you have the same story from above and from the ground. Eleanor answers "how should a sales org rebuild around AI." Jared answers "where do those Skills come from, and who should you hand Claude Code to first."

The sharpest thing about the CLAFTS thread is how it sets the bottleneck straight: what stalled Jared was never writing code, it was knowing what to build. And on that, the person who lives the pain every day knows better than any engineer. Once Claude Code flattens the coding barrier, the only thing still scarce is the business judgment he already had.

Original source
How one Anthropic seller rebuilt his team's workflows with Claude Code

claude.com/blog · "How Anthropic uses Claude" series · 2026-06-05 · featuring Jared Sires (GTM product manager / GTM architect)

Companion piece: How Anthropic rebuilt its sales org with Claude as connective tissue (Eleanor Dorfman @ SaaStr AI 2026)