When Joahne Carter commissioned AI-assisted research to reposition women’s health brand Semaine Health, she didn’t tell her research consultant the whole plan.
While Yogesh Chavda built synthetic personas and ran 15 positioning concepts through an agentic AI workflow, Joahne quietly conducted a parallel study with real consumers. She wanted to know if synthetic data could actually be trusted before she would stake the brand on it.
The results mirrored each other almost exactly. That’s when Joahne knew she had something.
Her experience points to a question every marketer is now asking: Can you trust AI to guide high-stakes brand decisions?
Joahne, chief marketing officer at Semaine Health, and Yogesh, founder of Y2S Consulting, recently joined the Content Marketing Institute and its sister brand TMRE for the webinar, Can You Trust AI for Brand Decisions? A Real-World Test from Strategy to Results.
Together, they shared how AI-assisted research alongside traditional validation helped reposition the brand with confidence and measurable success.
Watch the webinar or read on to hear their story:
Big-brand marketing research on a startup budget
Joahne heads up both marketing and insights for Semaine Health, which offers vitamins, minerals, and supplements through retail stores and direct-to-consumer channels.
It traces its roots to a scientific process: A bioengineer and his wife set out to create natural supplements to address the root cause of her endometriosis.
“From there, Semaine was really born out of a need to help someone, and that is still so much of the DNA of the brand: to just bring better solutions to women’s health,” Joahne says.
In some ways, it’s still very much a start-up mentality, with a commitment to testing and iterating everything from products to messaging. But Joahne is no startup newcomer. Before joining Semaine, she spent years at Procter & Gamble and Johnson & Johnson, where she worked on billion-dollar brands including Olay, Herbal Essences, and Listerine.
She knows exactly what a fully-resourced research process looks like — and what it costs. Wearing multiple hats at a lean organization, she faces constant pressure to get research right, quickly, and affordably.
Why traditional market research wouldn’t work
When Joahne wanted to validate the brand’s marketing to make sure it resonated with the target audience, she enlisted Yogesh, a fellow P&G alum, for his market research expertise.
She knew firsthand that a traditional research plan entails months of extensive work with a large team, significant resources, and a variety of tools. A typical process includes:
Related:How To Use Agentic AI To Create Marketing Workflows
Internal and external meetings with stakeholders
Several rounds of focus groups to narrow down ideas
Quantitative and qualitative studies exploring everything from the initial positioning to advertising, packaging, activation, and more.
Joahne wanted to evaluate 15 ideas, but limited resources meant the traditional research path wasn’t an option. A fully comparable process at a major CPG brand would cost significantly more and take three to six months.
So, she and Yogesh decided to use agentic AI for the data foundation, while keeping human insights at the forefront to judge any branded strategic directions.
How AI market research actually worked
Joahne already had an idea of Semaine Health’s target audience and brand positioning.
So, Yogesh suggested using agentic AI models — autonomous systems built on large language models (LLMs) — to pressure-test her 15 ideas using synthetic data instead of recruited consumers.
“An agentic model basically is a series of steps that you’re taking in a workflow, of how you go from start to finish to do brand positioning work,” he explains.
Synthetic market research uses AI-powered simulations to model human behavior, preferences, and data. Generating synthetic personas based on real-world data lets businesses test concepts and survey designs to inform their marketing strategies faster than traditional recruiting. For example, a focus group AI module can replicate the questions and structure of a human focus group with synthetic respondents.
Related:What If Your Audience Could Talk Back? Custom GPTs Can Help
Using these modules, Yogesh built a sample of synthetic respondents that could be used qualitatively and quantitatively, as well as to create personas.
“I really appreciated that the steps that Yogesh included in the workflow definitely paralleled what you would do in a traditional process,” Joahne says.
Building the system wasn’t without friction. Yogesh describes the process as architectural — designing the plan before building it, a lesson he learned the hard way.
“Once I built the plan, then it was easier for me to go and execute it and actually simplify and operationalize the system,” he says. “My suggestion is just jump in, start planning and start building, and you’ll learn through trial and error what’s actually working and what’s not.”
Testing the synthetic data without telling anyone
The agentic workflow winnowed the 15 ideas to three in a matter of weeks. In a traditional market research process, that step alone would have taken months.
Yogesh began with Joahne’s target audience definition and built a detailed persona from it, covering demographics, psychographics, path to purchase, key touchpoints, and the emotional and functional tensions around category usage.
From that persona, he generated approximately 300 synthetic respondents, crafted with enough specificity that the likelihood of duplication was less than one in 2,000 or 3,000.
Joahne and Yogesh then agreed on six questions to reflect the decisions she needed to make as CMO — evaluating each concept’s distinctiveness, relevance, and emotional resonance.
All 15 ideas, plus a competitive benchmark concept, were run through the model, producing a data file of 300 respondents across six questions across 16 concepts. Batching that volume through the system was a key technical challenge for Yogesh to solve, but the output gave both qualitative and quantitative signals from the same synthetic sample.
A synthetic qualitative round was conducted to probe the why behind the numbers. The top performers weren’t the concepts Joahne had originally expected to win. One idea she hadn’t favored surfaced as a frontrunner. Without the rapid qualitative follow-up to understand why consumers responded favorably to it, she says the team likely would have set it aside and moved forward with a less optimal choice. The speed of the synthetic process made that extra investigative step possible in hours rather than weeks.
Still, Joahne wanted to test the benchmark concepts validated in the AI-powered focus group. Without telling Yogesh, she conducted a smaller quantitative study with real people, featuring the same buyer personas and questions. It focused on the top three ideas.
The validation study proved fruitful. The studies mirror each other with a standard variability that Joahne would expect in two traditional research studies.
The entire process — from 15 ideas to validated positioning — took approximately six weeks at a fraction of the cost. Joahne estimates the AI-assisted approach ran roughly 80% less than a comparable traditional study would have cost at a major CPG company.
Joahne was ready to place (some) brand trust into the strategic decisions.
Taking the new positioning to market
The synthetic data with a dose of human insight allowed Joahne to make definitive brand marketing strategy decisions: “We’re going to market with this. This is what our brand stands for. It’s our tagline … It’s like it is living, breathing, truly, truly what the brand stands for.”
The new positioning rolled out across advertising, retailer presentations, and health care professional outreach. Semaine Health also redesigned its website around the new strategy.
In a market test comparing the new creative against the control, the brand saw double-digit increases across every channel.
“What we are seeing with the new positioning is that we’re increasing the average order value, which is behind the new positioning,” Joahne says. “So, people are actually buying into it and spending more on the brand, and we’re actually growing in terms of new consumers.”
Making the case for trusting AI in the research process
Joahne’s takeaway isn’t that AI replaces the hard questions of marketing strategy. It’s what gives marketers a better shot at answering them.
“What should our positioning be? Who should we target? Those questions don’t go away,” she says. “Understanding the consumer journey, AI helps you with all those things. But those are business situations that you will continue to encounter. It’s just important to think about, can this tool help me get to a better outcome?”
For marketers looking to expand their use of AI, Joahne and Yogesh recommend a few points to consider:
Start with an experiment. Gain confidence in it by progressing to bigger projects before you use it for the highest-stakes decision.
Define what success looks like. Without a clear understanding, you won’t know when to move onto the next step.
Validate your findings with targeted traditional methods (i.e., real consumers) to make sure the synthetic findings are on the right track.
Use AI for rapid research, but always incorporate humans into the evaluation process.
Ask the questions: Does the AI research resonate with what you’re trying to solve for? Does it reflect the rigor of what you would’ve done from a traditional perspective?
The Semaine Health case demonstrates that trusting AI-assisted research doesn’t have to mean taking it on faith. Joahne didn’t. She tested it, validated it against real consumers, and only committed once the data earned her confidence.
AI-assisted research puts your traditional research instincts to work to support faster decisions that drive real results. So, start small, validate results, and keep human oversight at the forefront.