↓ Skip to main content
Recruiter Bot: Part 6

Recruiter Bot: Part 6

·496 words·3 mins
Author
Shanti Korporaal
Recruiter Bot - This article is part of a series.
Part 6: This Article

Apparently, Doing It the Hard Way Wasn’t a Waste of Time
#

Captain’s Log. Stardate 2026.314.

I have two Recruiter Bots.

Neither is finished.

Both have project plans.

And somehow, both keep assigning me homework.

I WOULD LIKE TO SPEAK TO MANAGEMENT.

Unfortunately, that’s also me.

🤣

Welcome back to Recruiter Bot.

The original plan was simple.

Build Recruiter Bot the hard way in n8n, learning how to design workflows, test things and troubleshoot problems.

Then build another version with Claude doing more of the development.

Two approaches. Same objective.

I was supposed to finish the n8n version first.

But I got curious.

PROJECT SCOPE: UNCHANGED.

PROJECT SCHEDULE: REARRANGED.

AVAILABLE RESOURCES: STILL ONE HUMAN.

🤣

I’ve worked alongside software developers before, but this is my first time designing and building a software system from scratch.

And n8n has been brilliant for learning.

I can see the workflows, follow the data and investigate when something breaks.

It’s slower because I’m doing so much myself.

Meanwhile, Claude is building its own version.

I’m deliberately not giving it my entire n8n workflow because I want to see how it approaches the same problem.

It’s faster, but I have less immediate visibility into what’s happening.

Which is exactly what I wanted to explore.

Then we hit a problem.

Evidence matching.

Recruiter Bot needs to match job requirements against actual evidence from my career history.

Not invent experience.

Not make assumptions.

Find the evidence.

And Claude wasn’t getting the matches right.

I kept looking at the results thinking…

How are you not seeing this?

Then I remembered.

I’d already spent ages working through this exact problem with ChatGPT in n8n.

We’d developed a prompt that was working well in my tests.

So I offered it to Claude.

Claude incorporated the prompt into its system.

And the matching improved.

N8N DEVELOPMENT TEAM: WE’VE ALREADY SOLVED THIS.

CLAUDE DEVELOPMENT TEAM: PLEASE FORWARD THE SPECIFICATIONS.

PROJECT MANAGER: I AM BOTH DEVELOPMENT TEAMS.

🤣

And that was a satisfying moment.

Because without all that n8n experimentation, I wouldn’t have had a tested solution ready.

Or necessarily understood why the matching was going wrong.

The two approaches are teaching me different things.

n8n is teaching me how the machinery works.

Claude is teaching me how to work with AI that’s doing more of the building.

And understanding the first is already helping me with the second.

Neither version has produced a finished résumé yet.

And apparently, Claude would also like me to do more career archaeology.

CAREER ARCHAEOLOGY III: THE DIG THAT NEVER ENDS.

😭🤣

But I’m starting to see why building the hard way first was worth it.

The point was never just to build a Recruiter Bot. It was to understand enough to know what I was asking AI to do.

And now I have two unfinished systems, two project plans and nobody else to delegate to.

CAPTAIN’S LOG: REQUEST FOR ADDITIONAL STAFF DENIED.

Reason: Budget.

Approved by: Me.

🤣


Originally published on LinkedIn.

Recruiter Bot - This article is part of a series.
Part 6: This Article