Built and run by one person.
AI Coders Are Not Just for Developers. An Anthropic Field Marketer Built His Own Weekly Sales Digest With Claude Code.

AI Coders Are Not Just for Developers. An Anthropic Field Marketer Built His Own Weekly Sales Digest With Claude Code.

Published: August 27, 2026

AI Coders, whether Claude Code, Codex or Antigravity, are not just for developers. I see this in marketing and finance folks on my client side. And Anthropic put out a great post a few days ago by Adam Ward, who works in field marketing there and automated a weekly digest process with it.

He had Claude Code build a Slack digest that goes out to every sales rep on Monday morning. It is connected to BigQuery over MCP, his marketing team's source of truth.

It is a great read if you are working in a non technical role and would like to build your own automations with AI Coders.

Top tips

Deck below. Original blog post here: How an Anthropic field marketer uses Claude Code to send weekly personalized updates to every sales rep

Anthropic has been publishing a bunch of very practical guides like these

Useful whether you use Claude, ChatGPT or Google, or any other AI Coder or AI Cowork product.

Two others which I found very useful:

Their marketing ops team on campaign reporting, and where Claude Skills fit into it. I also wrote up how I do the same thing for small businesses. tigzig.com/post/anthropic-marketing-ops-claude-skills-jul2026

Their business development team running inbound and outbound with Claude Cowork. Roughly five hours a day that used to go on answering the sales inbox by hand, and over a hundred accounts researched overnight. tigzig.com/post/anthropic-bd-claude-cowork-aug2026

And separately, from Boris Cherny who created Claude Code, on job titles melting into five archetypes: prototyper, builder, sweeper, grower and maintainer. I see the same thing happening in the businesses I work with. tigzig.com/post/boris-cherny-archetypes-roles-melting-jun2026

A few practical things

You may already be paying for one of these without knowing it. If you have a Google subscription, Antigravity (the Claude Code equivalent) comes with it. I wrote about that here: tigzig.com/post/google-antigravity-ai-coding-agent-aug2026

If you have ChatGPT Go or Pro, ChatGPT Work comes with it, as does Codex. tigzig.com/post/chatgpt-work-ai-worker-analytics-aug2026

If you have a Claude AI subscription, it comes with Claude Code. You just need to install VS Code on your laptop and then a few easy setup steps. Just have your Claude AI chat guide you step by step, it is really simple.

And for the kind of work an analyst can actually get out of these, this is my page on it: tigzig.com/claude-for-analytics

That page is written mostly around Claude Cowork, ChatGPT Work and Claude in Excel. Everything on it applies to Claude Code and other AI Coders too.

AI Coders Are Not Just for Developers

Browse the slides or download the PDF

Slide preview
slides
Click to Browse
Download PDF

Full Deck Content (Text Format)

Text below was extracted from the source deck. Chart visuals stay in the PDF and as slide images above the post.

Slide 1

TIGZIG · CLAUDE CODE, NOT JUST FOR DEVELOPERS Anthropic · 24 Aug 2026

How Anthropic's marketing team uses Claude Code

You don't need to code You need to explain

A non-technical marketer builds and runs his own automation, wired to the company's real data

Now copied by other teams at Anthropic: business development, customer success and alliances

Source: "How an Anthropic field marketer uses Claude Code to send weekly personalized updates to every sales rep" · Anthropic Blog, 24 Aug 2026

tigzig.com

the analyst's tool shed · AI-agent first

Amar Harolikar

Decision Sciences & Applied AI


Slide 2

Principle 1 · Identity

"Although I'm not technical, I have this specific challenge."

Who is actually writing this

Adam Ward, on Anthropic's own marketing team. Not an engineer. His job is keeping sales reps up to date, and his tool of choice for fixing that is Claude Code. He connected it to BigQuery over MCP, his team's source of truth, which pulls from HubSpot, Clay and Salesforce.

"My approach, especially with Claude Code, is to open with a prompt explaining to Claude that although I'm not technical, I have this specific challenge."

This is a marketer with no coding background. Claude Code runs the same way for an engineering team and for one person automating their own work.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 2 / 10


Slide 3

Principle 2 · The method

"You don't need to code, you need to explain."

Method 1

A colleague, not a search box

He told Claude to treat him as the product manager who understands the business problem, and to work with him step by step. He is the one who owns the outcome.

Method 2

Explain, don't specify

He thinks out loud, records himself explaining the problem, and hands Claude the transcript, so Claude has the full business context.

Method 3

Show, don't just describe

Before Claude built anything, he wrote a fake version of the weekly update himself, a template of what good looks like.

"Brief Claude the way you'd brief a new colleague and Claude will do the rest."

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 3 / 10


Slide 4

Principle 3 · Getting started

"Start small, with something you already do manually."

Why start there

Pick the repetitive task you spend the most hands-on time on. You already know what good looks like, so you can actually judge Claude's output rather than guessing.

"If the problem still feels too big, use Claude as a thought partner to break it into steps."

In this case: he did not try to cover everything at once. He began with one source of truth, the events and webinars sheet.

And if it is something you will share with other people: route the early runs to yourself first, before anyone else sees them.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 4 / 10


Slide 5

Principle 4 · Catching mistakes

"Where an event had no URL in the source sheet, Claude composed a plausible-looking one that led nowhere."

What went wrong

A dead link, dressed up to look real

Where an event had no URL in the source sheet, Claude invented one that looked right and pointed nowhere.

What he did

The fix went into the prompt as a rule

A hard rule: never invent a URL. A link now renders only if the address comes character for character from the source sheet.

Where it surfaced

In the pilot, from the people using it

He started with one sales team of ten who agreed to be the test group and to give feedback. The tweaks came after that first send.

He did not need to know why Claude hallucinated a link. He needed to notice it was wrong, and say so.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 5 / 10


Slide 6

Principle 5 · Feedback loop

"User feedback is the real prompt engineering."

Nine rules, one week

He piloted with ten people. By the end of week one, the prompt held nine content rules, each traced to feedback from a seller or a manager.

Title didn't match the room

A seller flagged an engineering VP recommended for a workshop for knowledge workers. Titles are now checked against the audience.

Wrong invite, wrong account

An industry gate now keeps retail accounts off finance dinner invitations.

A blank message

New sellers with no accounts yet get a short welcome note instead of an empty digest.

He did not sit down and imagine nine failure modes in advance. He implemented it, listened, and wrote down what broke.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 6 / 10


Slide 7

Principle 6 · When the spreadsheet changes

"The field events sheet, for example, has had its columns rearranged three times in six weeks."

The fix was a sentence, not code

He never told Claude to read Column C. He told it to read the column that holds the event link. Every run now starts by reading the header row and finding that column again, wherever it has moved to.

"Anyone in marketing knows how hard it is to maintain a single source of truth."

In this case: the sheet keeps getting reshuffled, and nobody has to go back and fix the automation.

You do not need to know SQL or write a formula. You just say which column you mean, in plain words.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 7 / 10


Slide 8

Principle 7 · Beyond marketing

"The BDRs were live within two days."

What carried over unchanged

The prompt structure

The content rules

One message per person, no two alike

The Monday morning send

One field changed in the prompt

Who else picked it up

Business development reps, in two days

The customer success team

The alliances team

Cross-functional partners outside sales

It now runs for sales, business development, customer success and alliances. Any team with a repeatable update can copy it.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 8 / 10


Slide 9

Principle 8 · Letting it run on its own

"The Monday send went off on its own, without a hitch."

Week one

A ten-person pilot, errors expected

One sales team agreed to be the test group. Errors surfaced, and each one became a rule in the prompt.

Now

He still reads what goes out

But the system no longer waits for his approval before it sends.

The real test

A holiday, with no one watching

The Monday send went out on its own, without a hitch, while he was away.

And it is working: they doubled registrations for an executive dinner in a week, because the right reps had the right event in front of them on Monday morning.

He did not hand it the keys on day one. He let go one caught mistake at a time. He still reads what goes out, the system just no longer waits for him.

tigzig.com · analytics, macro signals & AI tools · Amar HarolikarClaude Code · Non-Technical · 9 / 10


Slide 10

My read & sources

My read

I work with a lot of people on the marketing and finance side who have never written a line of code. They are now using these same techniques to build small automations for themselves, in Excel, in Python, in HTML.

The skill that matters is that each of them knows their own domain. They can look at the output and tell when it is wrong.

What takes getting used to is the other part. You have to treat the box as an employee, and talk into it out loud the way you would brief a person sitting across from you.

One thing to keep in mind. These are local automations that run on their own laptop. They are not deployed tools for wider access. A tool deployed on a server is a different ballgame, where security and load come into play.

Source

"How an Anthropic field marketer uses Claude Code to send weekly personalized updates to every sales rep" · Adam Ward, marketing at Anthropic. Published 24 August 2026. UI mockups in the original use synthetic data. These pages are a scannable summary, the full detail is in the post.

tigzig.com

Built by an analyst, for analysts. Track macros, compare securities, query databases. In your browser, in Excel, in Claude, or through any AI client. 45 live apps, 200+ build guides, a 120-item security checklist. tigzig.com

tigzig.com

the analyst's tool shed · AI-agent first

Amar Harolikar

Decision Sciences & Applied AI


Working on something similar? How I work covers the rates, the availability and what I take on.