UnHacked the podcast with hosts Justin Shelley, Bryan Lachapelle, and Mario Zaki.

A client could not hire a $95K to $125K role, so they built an AI agent instead. It now monitors bids, files docs, and even runs go or no-go decisions.

In this episode, Justin Shelley, Bryan Lachapelle, Mario Zaki, and Joshua Holloway break down what “vibe coding” looks like when it actually solves a business bottleneck, not just a tech demo. Josh walks through a real proof of concept: an agent that watches a mailbox, ingests bid emails and attachments, mirrors a consistent folder structure, extracts key data into Excel, and keeps iterating toward a real dashboard.

The business impact is not hypothetical. The team describes how this automation started by saving 6 to 24 hours a week of manual parsing, then evolved into decision support. The company can move faster, reduce errors that happen when someone is “drowning,” and even pursue bigger opportunities because the bottleneck process is no longer slowing everyone down.

They also get honest about the risks. AI workflows can “go off the rails” or degrade over time, especially when you rely on broad prompts. The group explains why narrow tasks, hard-coded guardrails, and quality checks matter, plus what can go wrong if you blindly approve permissions or accidentally give people super-admin access.

If you are wondering whether AI can actually replace a painful, repetitive process in your company, this is the clearest real-world example in the series so far.

What you’ll learn:

  • How an email-watching agent can capture bid responses, read attachments, and file everything into a consistent project folder structure
  • Why “consistency is key” for automation, and how standard folders made this build easier
  • How the workflow progressed from intake automation to a go or no-go decision based on company criteria, with a human proof step
  • Why starting with Excel can be a practical transition step before replacing it with a dashboard
  • How to add guardrails and checks so AI does not delete data, drift over time, or create unsafe access