This article was based on the interview with Narek Vardanyan, CEO of PreLaunch.com by Greg Kihlström, AI and Marketing keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
Synthetic personas sound like a marketer’s dream—or perhaps a data scientist’s hallucination. What if you could talk to your ideal customer without the logistics of scheduling a focus group or the potential bias of a survey? It’s a tantalizing promise, and thanks to platforms like PreLaunch.com, it’s inching closer to reality. But as Narek Vardanyan, CEO of PreLaunch.com, makes clear in a recent interview, not all synthetic personas are created equal—and their utility depends heavily on what you’re asking, and why.
Vardanyan brings a deep background in go-to-market strategy and product validation. After launching a successful marketing agency and helping over 1,000 products generate more than $100 million in sales, he saw a persistent gap: many products fail, not because they’re bad ideas, but because the assumptions behind them aren’t properly tested. PreLaunch.com is built to bridge that gap, and their newest innovation, Customer Persona, takes things a step further—by letting marketers talk directly with AI personas modeled on real audience data.
So are synthetic personas the future of customer insight—or just another shiny object? Let’s take a closer look at four takeaways from our conversation with Vardanyan.
Synthetic Personas Are Only as Smart as Their Training Data
“The biggest unknown… is how reliable is this persona for their purposes,” said Vardanyan, emphasizing that synthetic personas must be built on credible, statistically significant data to deliver any real value.
This is where most enterprise marketing leaders should begin their analysis—not with a cool demo, but with a sober look at the foundation of the model. Are the personas trained on first-party behavioral data, or scraped assumptions from a public subreddit? Vardanyan suggests the difference is not subtle. If the underlying data is weak, no amount of clever prompting will generate actionable insights. For companies with deep historical data, building proprietary personas may yield highly predictive results. For everyone else, buyer beware.
Synthetic + Human Is the Magic Formula for Innovation
“If this is something that hasn’t been in the market… we are basically mixing real humans with AI,” explained Vardanyan.
This is where PreLaunch.com’s real differentiation shines. For truly novel products—say, a flying car—generative AI alone can’t predict how customers will react, because it has no training data. But PreLaunch’s approach places new ideas in front of real people, asks them for micro-commitments (like $5 deposits), and then uses that intent data to fuel machine learning. Only after clusters of interest emerge do they generate personas. These personas aren’t hallucinations—they’re abstractions of verified customer motivation.
Marketers in innovation-driven organizations should take note: this hybrid model of human intent plus AI summarization offers the best of both worlds. It’s fast, scalable, and most importantly, it’s grounded.
Speed and Scalability—With a Caveat
“The biggest advantage is the time… You can talk as long as you want. Nobody gets tired,” Vardanyan said, with just the faintest smirk in his voice.
That’s the pitch that makes synthetic personas so appealing: 24/7 access, instant feedback, zero scheduling conflicts. In a world where product teams are expected to iterate quickly, and marketers are constantly under pressure to “move fast without breaking too many things,” AI-generated personas feel like a tempting shortcut.
But Vardanyan warns not to let the speed seduce you. The utility of synthetic personas is entirely contextual. Testing a new cereal box design? Sure, use synthetic personas. Testing market viability for a wearable AI assistant that reads emotions? Probably not without real-world validation.
You Still Need to Do the Hard Work
“There’s no ultimate benchmark… you need to do your own fact checking,” Vardanyan emphasized when asked how to measure the reliability of AI personas.
Enterprise marketers who have grown comfortable outsourcing thinking to dashboards and models may want to reread that quote. Vardanyan encourages users to triangulate: test your synthetic persona outputs against traditional research. Cross-check survey results. Compare predictive outcomes. The best results come when synthetic insights are part of a broader research and validation ecosystem.
That also means marketers can’t treat these personas as black boxes. Ask how they were built. Ask where the data came from. Ask whether the model was fine-tuned or simply off-the-shelf. If you wouldn’t make million-dollar decisions based on a random Twitter poll, don’t do it based on an unvetted persona.
There’s a quiet revolution happening in how brands interact with customer insight. For years, we’ve built personas from patchwork: a few quotes from a focus group, a handful of analytics reports, a sprinkle of aspirational fiction. Synthetic personas, particularly when trained on real behavioral intent, can turn that fiction into function.
But as Vardanyan reminds us, this isn’t a shortcut to truth—it’s a tool, and like any tool, it can be used well or poorly. “They will become more useful, more convenient,” he said, “but you need to be really careful not to just fully rely on it”.
Marketers, especially those working in high-growth, innovation-centric environments, should see synthetic personas not as an oracle, but as an amplifier—one that becomes more accurate the more real-world insight you feed into it. The best marketers will use synthetic personas to scale their instincts, not replace them.
And for those wondering how to stay agile amidst all this change? Vardanyan’s advice is refreshingly pragmatic: “Use all the current innovations… and try to implement them into the platform and into the vision that we’re having”.
In other words, experiment early, test often, and keep humans in the loop. Especially when you’re trying to guess what kind of person buys a flying car.
