Insights | 8 West Consulting

AI and Deveolpment

Written by Simone Medas | 2026 - June 3

"We knew our research might lead to big changes for the world, but we felt like we couldn’t stop because innovation was so exciting."

 

— Stanisław Lem, “His Master’s Voice”

 

This article is intended for developers interested in integrating AI into their workflow, and for engineering managers who want to evaluate the real-world impact of tools like Cursor, Windsurf or other AI-based coding assistants.

 

Recently, 8 West gave me the opportunity to try out Cursor, a tool that helps with software development using AI. I was so excited to get started that I ended up spending the whole weekend experimenting with it — a side project entirely done in my own time. I wanted to see if it could help me create something that wasn’t just a prototype, but a real tool used in production: a solution that allows sending emails with HTML content using JSON as the data source. I quickly put the tool to use by generating and sending reports — formatted in HTML — containing Azure DevOps test results. At first, it looked like a simple task. But as I worked through it, I started asking deeper questions:

 

How much can AI really help in software development? And when do we still need human judgment and experience?

 

What I Learned

 

Working with Cursor showed me both what AI can do well and where it still struggles.

 

  • It's especially useful for quickly building proofs of concept (POCs) or starting a project using new technologies, libraries, programming languages, or infrastructure components you want to explore. It lowers the entry barrier and encourages experimentation — offering a valuable opportunity to get hands-on exposure with unfamiliar tools.
  • However, the more you understand the technology you're working with, the better you'll be able to guide the AI and evaluate its suggestions. If you're unfamiliar with the domain, the AI might lead you down paths that are functional but not well-suited to your actual goals — producing solutions that are inefficient, incomplete, or not fully aligned with the intended use case. That’s why human oversight and subject-matter knowledge remain essential.

 

What worked well.

 

  • It wrote working PowerShell scripts to get test results from Azure DevOps.
  • It built an HTML/CSS email layout that matched the JSON data.
  • It used Handlebars.Net to help fill in data automatically.
  • I got a basic version of the tool working in under an hour.

 

These wins showed how helpful AI can be when setting up repetitive or technical tasks.

 

What didn’t work so well.

 

  • The code had to be modified to work with PowerShell 5.1 instead of 7. I realized only after the first requests, that I should have specified the Powershell version, since it has big differences. So the question remains: if I had been more specific from the beginning, would I have had fewer problems?
  • It had trouble making custom helpers in Handlebars and kept repeating mistakes.
  • It couldn’t easily forget wrong ideas or avoid trying things that had already failed.
  • Emails didn’t always look right in email apps, even if they looked good in a browser.

 

These problems didn’t mean the tool was useless — but they showed that AI still needs human support and corrections. Having a developer check the results and make final fixes was very important. In fact, if I had written the entire solution manually from scratch, without the help of this tool, I estimate it would have taken me more than twice the time overall. The tool clearly helped speed up development significantly by handling many repetitive and technical tasks, allowing me to focus on the logic and final touches.

 

What “AI-Better” Means.

 

Thanks to 8 West, I have seen firsthand that AI cannot replace developers, but it makes their work easier and faster. It also creates opportunities to learn new things and explore unfamiliar technologies, helping developers grow as they create.

 

With the right approach, AI can be a valuable teammate. It works best when people guide it, analyze what it creates, and use their experience to improve it.

 

I’m especially glad to see that 8 West’s organizational vision aligns so closely with this perspective. The emphasis on augmenting human expertise rather than replacing it — and on empowering teams to use AI responsibly and creatively — reflects my own experience and values. The recently published AI-Better vision statement captures these principles clearly and, in my view, will strongly resonate with many in the development community.

 

I’d love to hear what others have done with AI. Have you tried solving real problems with it? What worked, and where did you still need to step in?

 

To close the article, I’d like to quote Andrew Ng — a respected authority in the field of artificial intelligence and the founder of DeepLearning.AI — to balance the opening reflection from one of science fiction’s greatest voices with the perspective of one of today’s foremost AI practitioners:

 

“Some people today are telling others not to learn programming because they think AI will take over that job. I believe this is some of the worst career advice ever. I disagree with a famous scientist who once said that programming will probably disappear and computers will just program themselves. Telling people not to learn coding is harmful.”

 

— Andrew Ng

 

 

 

Technical Breakdown – Cursor AI Usage Summary

 

📌 Project Overview

 

Objective: Develop a PowerShell-based tool to send HTML-formatted emails containing Azure DevOps test results, using JSON as the data source and Handlebars.Net for templating. Total time invested: ~10 hours across two Cursor usage sessions.

 

🧱 Tool Structure & Components

 

  • PowerShell-based script (initially version 7, later adapted to 5.1).
  • HTML Template for email rendering.
  • JSON file with ADO test results.
  • JSON file with recipient list.
  • Handlebars.Net for databinding.
  • README.md auto-generated.
  •  

✅ Functional Requirements Delivered

 

  • Extraction of ADO test results via REST API.
  • HTML output generation from template and JSON.
  • Multi-recipient email support via external JSON file.
  • Successful email delivery on first test run.

🕒 Time spent: 30 minutes

 

🛠️ Improvements & Quick Fixes

 

  • PowerShell 7 → 5.1 compatibility.
  • Refactoring into two scripts: JSON data extraction Email rendering + sending
  • Request for cleaner HTML layout was fulfilled.

🕒 Time spent: 15 minutes

 

⚠️ Challenges & Technical Issues

 

🧩 Dependency Handling (Handlebars.Net)

 

  • Multiple failed attempts to resolve package dependencies.
  • Final solution achieved after trying various workarounds.

🕒 Time spent: 2 hours

 

⏱️ Custom Helper for Duration Formatting

 

  • Cursor failed to generate a valid helper for milliseconds → readable duration.
  • Required GitHub source reference and manual coding.
  • Repeated loops and regressions in output.

🕒 Time spent: 3 hours

 

📦 Adding External Helpers Package

 

  • Attempted to include library with predefined helpers.
  • Cursor failed to resolve references and dependencies.
  • Solved manually with a simplified approach.

🕒 Time spent: 3 hours

 

🖼️ HTML Template Correction

 

  • Minor syntax issue caused binding to fail.
  • Identified and fixed manually.

🕒 Time spent: 30 minutes

 

📬 Final Rendering Consideration

 

Discrepancy between browser and email client rendering due to CSS in <head>. 📝 Recommendation (from Cursor): use inline styles. ✅ Solution tested and considered valid.