23 years of Linux administration and shell scripting. 6-7 years of accumulated automation work. Senior role managing pan-India telecom servers.
Problem: automation plateau. Next-level skills (Python, data transformation, advanced visualisation) required time I did not have.
AI solution: not learning those skills end-to-end, but using AI to leverage my existing domain expertise for implementation tasks I previously lacked the bandwidth to acquire.
TOOLS USED IN THIS STORY:
ChatGPT — AI-assisted shell script review and optimisation
Power Query (Excel/Power BI) — AI-guided data transformation
Python learning stack — domain-contextualised AI tutoring
AI dashboard tools — pan-India server statistics visualisation
Technique 1: AI-Assisted Legacy Script Optimisation
The advanced use case that most developers have not considered: using AI to review code written iteratively over years by a single expert, specifically to surface inefficiencies that proximity has made invisible.
The Prompt Framework for Legacy Script Review
"You are a senior Linux/shell scripting expert specialising in
telecom infrastructure optimisation.
Review this shell script: [paste script]
Context:
- Environment: [Linux version, server type]
- Function: prepaid charging platform automation
- Current performance: [baseline metrics]
- Scale: pan-India, [N] servers
Analyse for:
- CPU load inefficiencies (rank by impact)
- Race conditions or timing vulnerabilities
- Memory usage optimisation opportunities
- Redundant operations that can be batched
- Monitoring gaps that should be instrumented
Provide optimised version with inline comments explaining each change.
Flag any changes that require testing in staging before production."
"I had spent 6-7 good years automating many things. After joining Be10x and being introduced to ChatGPT and other AI tools — I was able to manage, optimize, and streamline all of that in just one week."— Nagaraj
The CPU load reductions surfaced by this review had not been identified in years of manual expert review — because the expert was too embedded in the logic to see it from the outside. AI provides that external perspective on demand.
Technique 2: Power Query — AI-Guided Data Transformation
The pan-India dashboard challenge: raw server statistics data in a format too complex to visualise directly. The transformation requirement that Nagaraj had previously called impossible.
The Data Transformation Prompt
"I have server statistics data from [N] servers across India.
Format: [describe raw format — columns, data types, structure]
Problem: [specific transformation challenge]
I need a Power Query M script that:
- Normalises server IDs to consistent naming convention
- Pivots [specific columns] for regional comparison
- Calculates [specific KPIs] from raw fields
- Handles missing data by [specific rule]
- Outputs a table suitable for slicer-based filtering in Excel"
The 'impossible' project had not become possible because the data changed. It became possible because the data transformation capability changed — AI provided the Power Query expertise that years of manual work had not developed.
"I told my boss this is impossible. The data we get is too complex. Transforming it into something easy was very difficult."— Nagaraj (before AI upskilling)
Technique 3: Domain-Contextualised Python Learning
The standard Python learning approach — generic tutorials, exercises unrelated to your domain — is slow for experienced professionals because it lacks the contextual relevance that accelerates retention.
The Domain-Contextualised Learning Prompt
"I am learning Python. My background: 23 years of Linux admin
and shell scripting in telecom. My goal: automate server health
checks and alarm analysis for a prepaid charging platform.
Teach me [specific Python concept] by:
- Explaining it in terms of how it maps to shell scripting concepts I already know
- Giving an example from telecom server management
- Showing the equivalent shell script and the Python alternative
- Highlighting where Python is genuinely better and where shell scripting remains more appropriate"
Learning with your specific domain context embedded produces dramatically faster competency development than generic tutorials.
"Now I have entered a world where I feel I can achieve anything."
Key Takeaways for Developers
Use AI for legacy code review — it surfaces what proximity obscures. Iteratively built code contains inefficiencies that the author cannot see. AI external review finds them systematically.
Power Query and AI is an underexplored combination. Complex data transformation requirements become tractable when you can prompt for specific M script logic rather than learning it end-to-end.
Domain-contextualised prompts accelerate skill acquisition by 3-5x. Asking AI to teach you in terms of your existing knowledge, with your domain's examples, is categorically different from generic tutorials.
// Watch Nagaraj's full walkthrough
▶ https://youtu.be/E25x7VXVXWM
Top comments (0)