Why Smart Operators Are Still Doing Dumb Work

Last week: Andrew Mewborn, founder, spending 10+ hours a week manually A/B testing emails. Subject lines, hooks, body copy, CTAs. One variable at a time. Tracking in spreadsheets. He's the scientist, the lab tech, and the data analyst. Manual GTM optimization that should be automated.

This week: Katie West, Customer Success leader. Fifteen years of experience across Rakuten and Rudderstack.

Her job? Writing the same email 100 times. Asking engineers for data exports. Rebuilding dashboards she already built at her last company.

Smart operator. Dumb work.

Here's why this is still happening and what you should build.

The Work That Shouldn't Exist

Katie manages hundreds of customers. She needs to know: Who's using the product? Who's not? Who needs help?

Seems basic. It's not.

She can't tell by looking at logins (set-and-forget users don't log in often, but the product works fine for them). She can't tell by support tickets (high tickets could mean engaged users, not unhappy ones). Usage data exists, but it's scattered.

So what does Katie do?

She spent 6-9 months building a health score dashboard at Rudderstack. Combined product usage, feature flags, consumption data, support tickets, Slack messages. Red, yellow, green indicators.

Then another 3-4 months getting her team aligned on what those indicators actually mean.

She's rebuilding the exact same thing.

We are recreating exactly what we had at Rudderstack.

Different company. Same problem. Same manual work to solve it.

The Email Problem

Once Katie knows which customers need help, she has to reach out.

Generic CS emails don't work. "Hope you're enjoying the product!" lands in spam mentally, if not literally.

She needs personalization:

• Your usage dropped 20% last month
• You don't have this feature enabled
• People aren't logging in - would you like training?

Her current process: Giant Excel sheet. Tracks next steps for each customer. She manually writes every single email.

Takes hours. She can only prioritize enterprise customers and the biggest ACV gaps. Everyone else? Doesn't get reached.

Template emails that could fill in with usage data automatically. Categorize customers, generate custom emails, make it easy to send.

She's not asking for AI magic. She's asking for mail merge with usage data. In 2026.

The Data Export Problem

Katie needs to email users. Not just the people who signed the contract - the actual users. The 10+ people added to the platform after the sale.

Those users live in the product database. Different permission levels (admin, biller, poster). She can't query it herself.

Every single time she needs to send an email: Site maintenance notification. Rate change announcement. Adoption campaign.

'She asks an engineer: "Can you pull an email list for me?"'

I have no way of seeing that. So right now I end up just pulling a list and then I manually email everybody myself. And that takes a long time.

Engineering time: wasted on data exports. Her time: wasted waiting, then manually sending emails.

This happens weekly.

The Spreadsheet Problem

Katie's data lives in five places:

• Sigma: Product usage
• HubSpot: Sold-to contacts, contract amounts
• Product database: Actual users, permissions
• Support tickets: Customer health signals
• Google Sheets: Where everything ends up

She hired someone to connect Google Sheets to Sigma. Now they're manually building out feature flag tracking, pulling contract data from HubSpot, trying to get it all in one place.

All roads lead back to spreadsheets at the end of the day.

Why spreadsheets? Because CS tools are too rigid. Her company is in growth mode. Product might change. Permissions might change. She needs flexibility.

Her strategy: "We're going to do everything in spreadsheets for 6-12 months until we know what we want and we know that this works. Then we'll invest in building this stuff out."

Translation: She'll rebuild this infrastructure for the third time in her career.

The Training Video Problem

Users don't adopt products by reading documentation. They adopt by seeing how to do something once, then doing it themselves.

Katie knows this. So she manually records training videos. Does live calls with customers to walk them through the UI.

Why is this necessary? Because in the sales process, executives saw the demo. But the actual users - the people who have to use the product every day - never saw it. Now they're expected to figure it out.

'What users want: "How do I add a teammate?" → 2-minute video showing exactly where to click.'

What they get: Engineering documentation. Or a calendar invite for a group training session with strangers.

It's 2026. Why am I reading docs? I just want a video that shows me where to click to solve this problem.

She spends hours creating these videos manually. One at a time. For common questions she's answered dozens of times.

Why This Is Still Happening

Katie is smart. She knows what she needs. She's built systems before.

So why is she still doing this work manually?

Three reasons:

1. The tools are too heavy.

Enterprise CS platforms require months of implementation. They're rigid. They're expensive. For a growth-stage company that might change direction in six months, that's not viable.

Katie doesn't want a platform. She wants a workflow.

2. The data isn't connected.

Her CRM has sold-to contacts. Her product database has actual users. Her analytics tool has usage data. Her support tool has tickets.

No tool bridges all of this. So she ends up in Google Sheets, manually combining everything.

3. AI tools aren't solving her actual problems.

She doesn't need a chatbot. She doesn't need an AI assistant that summarizes her emails.

She needs:

• Automatic customer categorization based on usage patterns
• Auto-generated personalized emails with real data
• Queryable access to her product database without asking engineers
• One-click training video generation for common UI tasks

These are repetitive, data-driven tasks. This is exactly what AI should do.

But nobody's building it for her.

The Pattern

Katie isn't unique.

Every CS operator at a growth-stage company faces this. So do sales ops people. So do product ops people. So do marketing ops people.

The pattern:

• They're highly skilled
• They know exactly what they need
• They spend their days doing repetitive work that should be automated
• They rebuild the same infrastructure at every company
• They end up in spreadsheets because nothing else is flexible enough

Why builders miss this:

You're building AI assistants. AI copilots. AI chatbots.

Katie doesn't want conversation. She wants automation of repetitive tasks.

She's writing 100 versions of the same email with different data. She's rebuilding the same dashboard at her third company. She's recording the same training video for the 50th time.

This is dumb work. AI should do it.

What Andrew and Katie Have in Common

Last week: Andrew manually A/B tests emails, tracks results in spreadsheets, optimizes GTM one variable at a time.

This week: Katie manually categorizes customers, writes personalized emails, rebuilds data pipelines at every company.

The connecting thread:

Both are doing work that scales linearly with growth. More emails to test = more manual tracking. More customers = more manual work.

The economics didn't make sense to automate this before. Building software to eliminate Katie's manual work would've required a team, 18 months, meaningful investment.

Now? Two people, six weeks, modern AI tools.

The constraint shifted. The problems stayed the same.

What You Should Build

Five opportunities from Katie's work:

1. Adoption Intelligence Without Infrastructure

Katie spent 6-9 months building a health score dashboard. Then 3-4 months getting her team aligned on what it means.

The opportunity: Plug-and-play adoption intelligence. Connects to product database, analyzes patterns, automatically categorizes customers (engaged, at-risk, set-and-forget). No infrastructure work required.

The signal: "We're recreating exactly what we had at our last company."

2. Usage-Driven Email Generation

'Katie manually writes emails with usage data: "Your usage dropped 20% last month."'

The opportunity: Pull usage data, generate personalized email, create send list, send automatically (or queue for review). Not generic templates. Actual data-driven personalization.

'The signal: "I end up just pulling a list and manually emailing everybody myself."'

3. Product Database Access Layer

Katie asks engineers for email exports weekly. Users live in product database, not CRM.

'The opportunity: Queryable interface on top of product database. "Show me all users who logged in the last 30 days with admin permissions." Export instantly. No engineering time required.'

'The signal: "I have to ask engineering to pull this for me."'

4. Flexible Data Pipeline for Growth-Stage Companies

Katie's data: Sigma, HubSpot, product DB, support tickets. Everything ends up in Google Sheets because CS tools are too rigid.

The opportunity: Lightweight data pipeline. Connect your stack → Google Sheets. Pre-built connectors for common tools. Easy to modify as company changes.

'The signal: "All roads lead back to spreadsheets."'

5. Query-Based Training Video Generation

'Katie records videos manually for common questions. "How do I add a teammate?" "How do I change permissions?"'

The opportunity: User types question → system generates 2-minute video showing UI clicks. AI captures screen, adds voiceover, delivers instantly. Updates automatically when UI changes.

'The signal: "I just want a video that shows me where to click."'

How to Find This

Katie's problems weren't obvious in the interview. She didn't say "I need an AI tool."

She said:

• It takes a long time
• I have to ask engineering
• We're rebuilding this
• I manually email everybody
• I spend hours recording videos

These phrases signal repetitive work.

When an operator describes their day and you hear:

• I manually [X]
• It takes hours to [Y]
• I have to ask [someone] to [Z]
• We're rebuilding [A]

They're telling you: This is dumb work. I shouldn't be doing this.

That's your opportunity.

Last week: Andrew Mewborn, manual GTM optimization in spreadsheets.

This week: Katie West, smart operator doing dumb work.

Next week: Another operator. Another pattern.

The opportunities aren't hiding. They're sitting in plain sight in every operator's day.

You just have to listen.

• Mike & Govind

📺 Watch the full conversation with Katie West