AI systems / Sales and marketing

I build the AI systems behind sales and marketing teams.

Like a pipeline that finds, researches, scores, and emails a specialty coffee buyer in about 35 seconds, built in Python with Claude.

View the work Download resume
Daniel KorrieIndianapolis, working remotely

Selected work

Four systems, in use.

01Prospecting pipelineManna Origins2026~35 seconds per prospect 02Content generatorElevate Indianapolis2025Drafts in the organization’s voice 03CEO daily briefingElevate Indianapolis2025About an hour saved each morning 04Branded graphics workflowElevate Indianapolis2025On-brand graphics without a designer

Case study 01

Manna Origins / Contract / 2026

A prospecting pipeline that does the repetitive part.

The problem

Manna imports specialty coffee and sells it to cafés and roasters. Finding those buyers was manual: searching maps, reading websites, writing every email by hand.

What I built

An eight-stage Python pipeline. It finds businesses, researches each one, scores the fit, drafts a personal email, sends it, reads replies, and hands warm leads to sales in Notion.

Python / Anthropic SDK / Google Places / Apollo / Gmail API / Notion API / SQLite / YAML

Follow one prospect through.

Claude agentPlain Python

192Prospects checked by hand before automating anything
35sPer prospect once it runs on its own
14/37Top-tier fits in the first pilot

Decisions

What made it hold up.

  1. 01Claude only where judgment helps

    Research, writing, and reading replies use Claude. Scoring is plain rules in a YAML file, so it stays predictable, cheap, and easy to change.

  2. 02One failure doesn’t stop the run

    Each stage handles its own errors. Failed prospects go to a retry queue with the error saved, and every run ends with a summary.

  3. 03No duplicates, full trail

    The database blocks repeat prospects, and every stage writes to a log, so any lead can be traced end to end.

  4. 04Built for more than one client

    Each client gets its own database, settings, and logs. A learning module suggests changes, and a person approves them first.

Case study 02

Elevate Indianapolis / Freelance / 2025–now

Tools a nonprofit team runs without me in the room.

Elevate runs mentoring programs in Indianapolis schools with a team of about 30. Every tool ships with written instructions and handoff notes.

elevate-content-generator.vercel.app
The Elevate content generator: fields for content type, event name, date, what happened, quotes, photos, and call to action
02 / Content generatorOpen the live app

Content generator

Staff pick a content type, add what happened and real quotes, and get a draft in Elevate’s voice: blog posts, newsletters, appeals, social posts, and CEO letters.

The voice comes from a style guide built from their newsletter archive. A stronger model handles the important pieces, so they only pay more when it matters.

CEO daily briefing

Claude connects to the CEO’s email, Slack, and other approved tools. Each morning he gets one briefing with what needs his attention, and scheduled tasks remind him who he still owes a reply.

Used daily / About an hour saved each morning

Branded graphics workflow

Staff paste a post description and a Claude Design workflow picks the right branded template. They can regenerate the wording or the layout until it’s right.

Non-designers make on-brand graphics on their own

Approach

How I work.

  1. 01Learn the work before automating it.

    I checked 192 prospects by hand before the pipeline ran on its own. The rules came from that, not from guessing.

  2. 02Use AI where judgment helps.

    Models write, research, and read replies. Plain code scores, routes, and logs. That keeps systems cheap and easy to trust.

  3. 03Build for the people who run it.

    Every tool ships with instructions and handoff notes. If a team needs me to use it, it isn’t finished.

Portrait of Daniel Korrie
Daniel Korrie, Indianapolis

About

Before I built automation, I did the work it replaces: 80+ calls a day in B2B sales. I build for the pipeline and the people in it.

  1. Business development and sales automation, Manna OriginsContract. Built the eight-stage prospecting pipeline.
  2. AI automation and marketing systems consultantFreelance. Websites, SEO, and Google Ads for small businesses; AI tools for Elevate Indianapolis since 2025.
  3. Carrier account manager, Spot Freight80+ calls a day, 30–40 active carrier accounts.
  4. B.S. Marketing/Systems, Taylor UniversityInformation systems, web design, and digital marketing.
Earlier work

Websites and landing pages for Columbia Culinary Tours, Abide Services Group, and Go2DJs, in Webflow and WordPress. More on Contra.