Tactic 2: Automated Variant Generation (Copy, Design, & More)
Tips:
•
Be specific in prompts:
e.g., “Generate a lighter,
friendly version of this text” or “Suggest a
completely different layout for the signup form.”
When to use:
The more detail on what kind of change you
want, the better the variants.
Employ AI variant generation during the experiment design
phase, right after you decide what to test. It’s especially useful
•
Quality-check and edit:
AI content is a first
when you need many creative options fast – for instance, testing
draft. Review generated copy for brand
personalized messages for different audiences, or localizing
compliance, tone, and factual accuracy
content to different regions without hiring translators. In high-
(especially for product details). Edit any quirks
velocity testing, AI can supply a steady stream of variations so
before deploying to real customers.
you’re never stuck waiting for new creative.
•
Use multi-modal AI for design:
Some advanced
Why it works:
tools (or internal models) can generate images
GenAI (especially large language models) excels at producing
or design mockups from text descriptions. If
natural-sounding text in different styles, tones, and languages.
available, leverage these to get visual
It can mimic your brand voice or even a famous personality if you
variations (e.g., different hero banner images)
prompt it to, ensuring variations stay on-brand. AI doesn’t get writer’s
to test, accelerating graphic design work.
block – you can instantly get tagline alternatives, UI text tweaks, or
color/theme suggestions.
According to industry experts
, today’s AI
•
Don’t forget edge cases:
Ask AI to also suggest
can go beyond one-off suggestions and recommend complete
a “radical” variation outside your usual comfort
experiments with multiple variations built-in optimizely.com. This
zone – this could surface an out-of-the-box
means when the AI suggests a test idea, it can provide the specific
idea that yields a surprising win. Just ensure it’s
variants to try, saving your team creative effort.
still practical to implement.
GenAI Q uick- W in P laybook for Experimentation
5
Powered by FlippingBook