1 / 15

The Operator-Engineer

Why Services Companies Have Never Had a Bigger Advantage

Sam Gaddis  ·  Runpoint  ·  Feb 24, 2025
  • Hold for ~15 seconds while people settle in
  • "First, thank you to the Super Step team for having us — really appreciate the invite."
  • "I'm Sam Gaddis, I run a company called Runpoint — we build software for companies across services, manufacturing, cybersecurity, media, you name it."
  • "Today I want to talk about a shift that's happening right now — the people who run businesses are becoming the people who build the tools, and that changes everything."
  • NEXT: "Before I explain the framework, let me show you what we actually built."
Case Study

Runpoint OS

Live Demo

Our internal operating system — built by operator-engineers, used by everyone.

Airtable (11 tables) Claude AI + OpenAI SendGrid Stripe Linear GitHub Jira Slack Google Calendar Google SSO Todoist

Sales & Finance

  • Opportunity pipeline
  • Deal forecasting (probability-weighted)
  • Multi-scenario cash flow modeling
  • Resource & headcount planning
  • Margin calculation per deal
  • Billing schedule management
  • Stripe payment processing
  • Campaign attribution
  • Email verification
  • Global search

Delivery & Ops

  • Proposal editor (TipTap)
  • AI proposal generation from transcripts
  • E-signature workflow
  • PDF generation & delivery
  • Proposal view analytics
  • Relationship intelligence (9 dimensions)
  • Account health scoring
  • Linear / GitHub / Jira sync
  • Contract template library
  • CLI for project ops

Recruiting & AI

  • Job posting management
  • Application intake + screening
  • AI resume assessment
  • Google Calendar interview scheduling
  • Automated interview reminders
  • AI chat with function calling
  • Transcript insight extraction
  • Account win identification
  • Expansion opportunity finder
  • Demo mode (data obfuscation)
119 API routes  ·  11 database tables  ·  77 components  ·  32 views  ·  11 integrations
  • Lead with this. "Before I explain the framework, let me show you what we actually built."
  • "I'm going to show you four things today — a mix of internal tools and something I built for another company I own. All of our client work is under strict privacy requirements, but we build so many internal tools that we have plenty to show. That's what I'll focus on."
  • Walk through the live app — sales pipeline, proposals, recruiting, financial modeling
  • The density of the slide should make the audience lean in — proof before theory
  • ~5-7 min demo
  • NEXT: Close the demo. Pause. "That took us three weeks."
The Gap

That took us 3 weeks.

3 weeks
Runpoint OS — 119 routes, 11 integrations, used daily by the whole team
vs.
6 months
What the same scope would have taken a traditional dev team — if it got funded at all

The gap isn't AI capability. It's knowing what to build and why it matters.

  • Let the demo still ring. "That took us three weeks."
  • A year ago it would have taken six months — if it got prioritized at all
  • The point isn't speed. It's that this was never going to get built the old way — the budget, the headcount, the prioritization would have killed it in the backlog
  • AI didn't make us faster. It made the thing possible.
  • Make them feel it: "How many tools like that are sitting in your backlog right now?"
  • NEXT: "But here's the thing — building it was never the hard part."
The Reality

The hard part isn't
the build.

What most companies focus on

Can AI write the code? Can it build the thing? The technology isn't the bottleneck anymore.

Where the real work is

Knowing what to build, weighing impact vs. effort, assessing data accessibility, and prioritizing across every part of the business AI can now touch.

The bottleneck was never the technology. It's the prioritization and the imagination.

  • The hardest part of our job is never the build. It's figuring out what to build.
  • This technology can touch any part of the business — sales, ops, finance, recruiting, delivery. That's what makes prioritization so critical.
  • Most companies' answer to "what would you do with AI?" reveals their ceiling — they'd do the same things, slightly faster
  • The real answer: build everything you've been putting off. The imagination to ask the right question is the scarce resource.
  • NEXT: "I want to walk through three concepts that frame everything else — who to invest in, how to think about automating your workflows, and why services companies specifically are leaving the most on the table. Let's start with the first one."
Concept 1

The AI Competence Matrix

Uses AI Does not use AI
Slop Cannons Lots of output. None of it good.
Turbo Brains Fast, correct, compounding value.
Dead Weight Slow and wrong.
Steady Hands Reliable, but limited throughput.
Bad judgment Good judgment
  • Walk each quadrant. Start bottom-left, go clockwise
  • Pause on the arrow: "AI doesn't fix judgment, it amplifies it."
  • Turbo Brains = where operator-engineers live. These are the people to invest in — through hiring, through internal development, through giving them room to build.
  • Key line: "AI moves people up the chart. The question is which column they're in."
  • NEXT: "OK, concept two — once you have the right people, what do they actually do?"
Concept 2

Workflows as Skills

Every repeatable process in your business can become a skill that AI handles.

Linear
More clients = more people =
more coordination overhead
vs.
Compounding
Turn workflows into skills =
each client costs less to serve

Automate your workflows once. Every future client benefits.

  • "If you've used the CLI coding tools like Claude Code, you'll know about skills — they're a way to teach AI to automate things you do all the time."
  • The same idea applies to every part of a business. Pretty much anything in accounting can be turned into a skill that an AI can handle. Same with project intake, invoicing, reporting, onboarding — anything repeatable.
  • The mindset shift: stop thinking of work as tasks people do. Start thinking of work as skills you can teach.
  • Linear growth means every new client costs the same to serve. Compounding means the second client is cheaper than the first, and the tenth is nearly free.
  • NEXT: "So why don't more companies do this? Concept three."
Concept 3

The Client Trap

Services companies are so focused on solving client problems that they never automate their own workflows.

100% of energy goes to client delivery
Internal ops run on spreadsheets and willpower
Tribal knowledge instead of systems
"We'll build internal tools when things slow down"
Every new client means more manual overhead
Revenue scales linearly with headcount

The cobbler's children have no shoes. But now the cobbler has AI.

  • "So we know who to invest in, and we know the goal is to turn workflows into skills. The obvious question is — why doesn't everybody just do that?"
  • "This is the answer. If you run a services company, look at this list and tell me how many of these sound familiar."
  • Pause and let them read. Give it a beat — the room will start nodding.
  • "The cobbler's children have no shoes. You solve these exact problems for your clients every day. You just never turn the lens inward — because there's always another client to serve. But now the cobbler has AI."
  • NEXT: "OK, enough concepts. Let me show you another build." → CNC demo
Case Study

Austin Custom CNC

Live Demo

One person. Two weeks. A complete business OS.

QuickBooks Online Airtable (22 tables) SendGrid Claude AI Tiller Gmail

Money

  • Accounts Receivable
  • Accounts Payable
  • Liabilities tracking
  • Cash position (Tiller sync)
  • P&L with drill-down
  • P&L cleanup (check OCR)
  • Cash flow forecasting
  • Recurring expenses
  • Payroll settings
  • 88 API routes

Operations

  • CRM pipeline (Kanban)
  • Opportunity tracking
  • Project management
  • Contract upload + AI analysis
  • Expected payment extraction
  • Team management
  • Time tracking
  • Crisis task management
  • Data quality grading (A-F)

Communication

  • Email campaigns (bulk send)
  • AI email composition
  • Contact management
  • Gmail contact sync
  • Email verification
  • Feedback system (voting + comments)
  • AI feedback clarification
  • Universal audit trail
  • Change log with revert
  • Weekly AR summary (cron)
88 API routes  ·  22 database tables  ·  52 components  ·  6 integrations
  • Let the slide land for 5 seconds before starting the demo
  • The density IS the point — the audience should feel the weight of what one person built in two weeks
  • Walk through the live app
  • ~5-7 min demo
  • NEXT: "So what kind of person actually builds this? Let me introduce the role."
The Solution

The Operator-Engineer

Not a developer who learns business. Not a business person who learns to prompt. The intersection of all three.

Business
Taste
Technical
Acumen
AI
Passion
Operator-Engineer
Knows which problem is worth solving
Comfortable at the code level, not just the prompt level
Ships fast, iterates faster, throws away freely
Treats AI as a power tool, not magic
Thinks in systems, not features

Any two of three makes you useful. All three makes you dangerous.

  • The Venn is the core idea. Walk through missing-circle combos:
  • Business taste without technical chops = great specs nobody builds
  • Technical acumen without AI = steady hand, reliable but slow
  • AI passion without business taste = slop cannon
  • You need all three. Traits on the right are what it looks like in practice
  • "Any two makes you useful. All three means you can go from problem to shipped solution without waiting on anyone else."
  • NEXT: "OK, so how do you actually create these people?"
The Playbook

How to Build an
Operator-Engineer

You don't hire for this title. You create the conditions and find the people who light up.

1

Give them unlimited model access

The $20/mo ceiling on AI tools is the dumbest bottleneck in your org. Remove it.

2

Data accessibility

Get your data into one place and make it available to the operator-engineers in your organization.

3

Streamline deployment

It should not take more than a button to deploy. No tickets, no third-party IT.

4

Give them permission to experiment

If your project plan involves a steering committee or a timeline longer than a month, it's probably not going to work.

The person is probably already in your org. The environment is what's missing.

  • Actionable slide. The audience should be thinking about specific people in their org
  • "You know who I'm talking about — the person who built that internal tool nobody asked for but everyone uses."
  • Model access — please don't make them use Copilot. Don't put them behind a usage cap. If a financial institution can use Claude Code, your organization can too. This is the cheapest, highest-leverage unlock you have.
  • Data accessibility — these are one of the two things that get in the way most at our clients. Build your data warehouse, get everything into one place, and make it available to the operator-engineers in your org. You can do all of this with Claude Code now.
  • Deployment — the other big blocker. With tools like Port.io you should be able to deploy new environments — sandbox databases, dev, staging, production — with a click of a button. You should not have to go through a third-party IT service. You should be able to spin up a prototype deployment account within 24 hours.
  • Executives: your job here is clearing red tape. Focus on giving your people access so they can build the tools they need. Most of your time should be spent removing the obstacles, not reviewing the output.
  • Permission to experiment — if the project plan involves a steering committee or the timeline is longer than a month, it's probably not going to work. The best operator-engineers build 10 things and keep 2. Let them.
  • NEXT: "And here's the proof that these people exist..."
The Signal

The Talent Signal

We posted a job for "business-minded CLI experts" — unlimited model access, no red tape.

800

applications in one week

People are desperate for environments that let them work this way.

  • Let the number land.
  • "This is what happens when you do the things on the previous slide. We posted a job description that promised pushing the limits of AI — a maxed-out Claude Code plan, a maxed-out Codex Pro plan, no red tape."
  • "800 people. For one role. At a company most of them had never heard of."
  • "Every one of the best applications we got talked about that promise as the reason they wanted to change companies. These people are out there. They're just stuck in environments that won't let them work this way."
  • NEXT: "One more example — and this one might surprise you."
Case Study

Runpoint Magazine

Live Demo

A print quarterly and digital publication — built and shipped by a small team.

Stripe Shippo (USPS) Airtable SendGrid Runpoint CRM

Content & Editorial

  • Markdown article system
  • 6 published articles (Issue 01)
  • Featured article promotion
  • Category & author taxonomy
  • Unique SVG hero art per article
  • Author pitch submissions
  • AI-assisted writing framework

Commerce & Fulfillment

  • Stripe checkout (promo codes)
  • Physical magazine orders
  • Auto shipping label gen (Shippo)
  • USPS rate optimization
  • Duplicate shipment prevention
  • Order webhook handling
  • Newsletter subscription

Design & Distribution

  • Dark/light theme
  • Smooth scroll (Lenis)
  • Dynamic OG images (Satori)
  • Page transitions
  • Responsive editorial grid
  • CRM contact sync
  • Campaign member tracking
4 API routes  ·  6 articles  ·  5 integrations  ·  print + digital
  • The point isn't the tech — it's the breadth
  • A small team ships: print magazine + digital publication + e-commerce + automated fulfillment
  • Walk through 2-3 articles showing range, then show the physical magazine if you have a copy
  • ~5 min demo
  • NEXT: "And actually — one more. You're looking at it right now."
Case Study

This Presentation

Meta

Built this week. From a call transcript to a full slide framework with synced presenter notes.

Claude Code GitHub Vercel

Slide Framework

  • Scroll-snap navigation
  • Keyboard controls (arrows, space)
  • Progress bar + nav dots
  • Dark/light theme toggle (T)
  • Mobile responsive
  • OG image auto-generation
  • PDF export on deploy

Presenter Mode

  • Synced second-tab view (P)
  • Speaker notes with formatting
  • Up-next preview
  • Elapsed timer
  • BroadcastChannel sync
  • Keyboard nav from either tab
  • Transition cues between slides

Workflow

  • Partner call transcript → outline
  • Claude Code iterative builds
  • Design system tokens (JSON)
  • Auto-deploy on push to main
  • Email gate (Airtable + API)
  • Self-contained single HTML file
  • No build step, no dependencies
1 HTML file  ·  0 dependencies  ·  15 slides  ·  synced presenter mode  ·  built in a week
  • The presentation you're watching right now is the case study.
  • Last week, my partner and I had a call about what to present at Telluride. I took that call transcript, fed it to Claude Code, and we went back and forth building this.
  • It's a single HTML file — no React, no build tools, no npm install. Just HTML, CSS, and JS.
  • The presenter mode is synced across browser tabs — I flip slides here and my notes update in a separate window. Hit P right now to see it.
  • Auto-deploys to slides.runpoint.ai on every push. OG images and PDFs generate automatically.
  • This is what operator-engineers do. The tool didn't exist, so we built it.
  • NEXT: "So — what can you do with all of this?"
Taking Action

What You Can Do

1

Find the operator-engineers in your org

They're already there — building things nobody asked for but everyone uses.

2

Fix the environment so they can actually work

Remove the friction. Unlock the tooling. Measure outcomes, not process compliance.

3

Contract out the push

Sometimes organizations need someone to come in and show what's actually possible.

  • "So — what can you do with all of this? Three things."
  • Card 1: "The people are already in your organization. You saw the matrix — you know who your Turbo Brains are. Find them. Give them room."
  • Card 2: "Fix the environment. Everything we talked about — model access, data accessibility, deployment, permission to experiment. That's where executive energy should go. Clear the red tape."
  • Card 3: "Sometimes organizations need a bit of a push. That's something you can contract out — to companies like ours or others. For a lot of our clients, especially the ones having trouble getting this going internally, we've seen big shifts just by coming in and showing how fast and how differently things can be delivered. It puts people's feet to the fire a little — especially the folks who might be more old-school in their thinking about what's possible."
  • Keep it casual. This isn't a close — it's an observation. The demos already did the selling. You're just saying the door is open.
  • NEXT: "Let's open it up."

Questions?

sam@runpoint.ai
runpoint.ai
Download PDF
  • ~10 minutes for Q&A
  • If questions are slow: "What's the biggest organizational barrier you're seeing to AI adoption?"
  • Fallback: "Dario Amodei says 1-3 years to human-level AI. What does that timeline mean for your business?"