top of page
image.png

Case Study

AI-Suggested &
Human Reviewed

Designing how performance engineers used an AI suggestion model which reduced manual effort of converting hardcoded values to dynamic value.

AI-Feature Embedment
Enterprise Tool

Let's have some glimpse first...

image.png
image.png

My Role

Research + FA +PM + {Design} + Engineering + Data Engineers

Discovery

  • Identified core parameterization pain points with support from performance engineering SMEs and support ticket analysis.

 

  • Strategised a phased rollout plan with PM, including time for user testing, explainability validation, and governance review

Development

  • Collaborated with the VuGen engineering team to define and ship the Parameter Candidate AI feature.

  • Collaborated with OpenText's AI Design Principles working group to align the feature with published UX-for-AI standards.

  • Validated UX through moderated testing with performance engineers.

  • Ensured implementation fidelity working directly with engineers on confidence-score thresholds and override behavior.

  • Worked with the data science team to define KPIs measuring suggestion accuracy and adoption

⏳ 6 weeks explorations.. in progress.

🖥 VuGen Desktop, integrated within the DevOps toolchain.

image.png

Setting some context, before we begin…

Ecosystem - DevOps (ADM)

  • VuGen is the script recording and editing tool at the front of OpenText's performance testing pipeline.

​

  • It lets performance engineers record real user transactions and turn them into repeatable, data-driven load test scripts, which then flow into LoadRunner Enterprise (LRE) for execution.

What is VuGen?

image.png
image.png
image.png

Current Experience

More manual and time consuming...

image.png

Validating the problem

Why AI-Assisted feature was required?

image.png
image.png

Current user flow..

image.png

Current opportunities...

image.png
image.png

Validating opportunitines

Brainstorming & decisions...

image.png
image.png

Along with feature addition

Redesigning the space...

image.png
image.png

Output

AI-suggestions in the user flow..

image.png
image.png

What next...in progress

Questions we are asking and validating...

image.png

Can AI show occurrence as well, if the name is repeated?

When the user discard, how can we undo it?

Can the AI understand how sure it is or can they show the confidence in their suggestion?

How to provide further filtration options for the script, when the script is data heavy?

What other setting does a user need, or what can they customized in 
(AI Suggestions) while the AI scans?

and a lot more...

What next...in progress

Questions we are asking and validating...

Detection earns trust only when paired with visible preview before commit.​​​​

Transparency

​​​Discovery was the real bottleneck not the replacement mechanics.​​​

Discoverability 
Flexibility

Suggest naming templates; don't enforce team conventions.

Scalability

One shared model tuned by team feedback scaled better than per-workflow patches.

I will be more than happy to walkthrough you with the process, validations, assumptions we made... :)

bottom of page