Last week, my co-founder Rhys sent me a LinkedIn post in Slack. It was from Nikolai Golos at Fluently AI, arguing that generic CTAs like "Start now" don't convert because users don't know what happens next. Rhys wondered if we made our CTA more descriptive like the examples in the post, would it help signups?
This is something I knew from years of A/B testing but had forgotten – hence the lazy ‘Get started’ button on the Rally homepage. A few days later with a specific proposal: change our homepage CTA from "Get Started" to "Start Simulating with AI Personas."
My gut reaction? Resistance.
I was busy, wasn't 100% sure about the new text, worried about mobile display, and frankly, risk-averse. We were launching a major publication feature the next day, expecting significant traffic. The last thing I wanted was to mess with a working system right before a big spike in signups.
But Rhys persisted, and I recognized an opportunity to dogfood our product. Let’s test it in Rally and see what AI personas based on our customers think.
A/B testing a change like this would be impractical. We’re an early stage startup, and don't have the traffic volume for statistical significance, and it would have taken weeks to get meaningful results. Instead, we used synthetic research to validate the concept.
Five minutes later, we had our answer.
The number of AI personas that decided to click on the button increased by 10%, from 51 to 56 out of 64 personas. Not earth-shattering, but meaningful for an early-stage startup where every conversion matters. Most importantly, it removed my fear. I know it’s not statistically significant, but knowing digital twins of my customers didn’t hate the idea, made me realize I was being overly protective of our landing page copy, becoming a bottleneck for Rhys. I could see concrete evidence that the change wouldn't harm performance—and might actually improve it.
We’ll see how the launch goes tomorrow, but I’m not worried. In fact, I’ll probably test a few more variations in the morning, to see if I can beat Rhys’ score.
The Meta Learning
I didn't feel strongly about either CTA option, so I instinctively wanted to stick with what I had. But my inaction would have prevented any change from happening, that could have a positive impact. Classic status quo bias in action.
As a founder, it doesn’t matter what headline is your favorite: you are not your customer. You need to get out of your own head, and be open to testing other options.
The broader lesson isn't about CTA optimization—it's about decision-making velocity and quality. How often do we avoid potentially beneficial changes because we can't quickly validate them? For early-stage startups especially, this kind of rapid validation can be the difference between iterating toward product-market fit and getting stuck in analysis paralysis.
Sometimes the best way to overcome your own hesitation is to just ask your audience directly. Even if that audience happens to be synthetic.
Want to test your own product decisions with AI personas? Try Ask Rally and see what your virtual audience thinks in minutes, not weeks.