By Pia Ostos, EVP & President, MarTech at Inmar Intelligence
1. AI can generate recommendations in seconds, but not every recommendation is a good one. What should marketers look at before trusting an AI-generated recommendation?
The first thing I’d look at is what the recommendation is actually built on. Machine learning trained on real purchase history behaves very differently than generative AI, which can produce confident outputs that aren’t always grounded in the same level of behavioral data. Accuracy and transparency are actually the top concerns the industry cites for AI adoption right now, and that gap is exactly why the distinction matters. So before trusting any recommendation, I’d want to know what data sits behind it and whether the output can be traced back to real consumer behavior.
2. How can marketers tell the difference between AI that is simply finding patterns in data and AI that is actually helping them make better business decisions?
Finding a pattern is easy; AI is very good at that part. The harder question is whether the pattern is grounded in something real, like an actual transaction, or if it just sounds plausible. Generative AI only performs as well as the data feeding it, and it’s built to produce a fluent answer whether or not that answer is accurate. If a recommendation can’t be traced back to real data, I’d treat it as a starting point for a conversation, not a decision.
3. Household penetration is often treated as a measure of brand reach. Why is looking at who isn’t buying yet just as important for finding growth?
Penetration tells you how many households bought a brand, but it doesn’t tell you who’s about to stop. That’s an equally important group to watch. Pairing purchase history with signals like coupon activity and predictive intent helps catch shoppers who may be starting to drift. Growth is just as much about the next new buyer as it is holding onto the households you already have.
4. With more consumer and purchase data available than ever, how can brands turn those insights into more relevant and timely marketing?
There’s more purchase data available than most brands know what to do with, so the real opportunity isn’t collecting more of it, it’s connecting what already exists. Purchase history is still the foundation, but pairing it with intent signals like coupon activity gives a much fuller picture of who’s ready to buy versus who’s starting to drift. In testing, that approach has delivered noticeably more efficient media spend for the brands using it. It’s less about a new dataset and more about making the data you already have work harder.
5. Promotions have traditionally relied heavily on discounts to drive sales. Why do you believe better targeting can create more value than simply offering a bigger discount?
A bigger discount doesn’t necessarily create more value if it isn’t relevant to the shopper. 61% of shoppers say promotions actually influence them before they’ve settled on a brand, so the win is showing up earlier with the right message, not just a lower price later. Our research backs that up too: 90% of shoppers told us they’d switch back to a preferred brand with the right promotion. Better targeting gets you there without giving away margin on shoppers who were buying anyway.
6. How can brands use data and AI to determine who actually needs an offer, rather than giving discounts to customers who would have purchased anyway?
This is where a lot of promotional budget can be spent unnecessarily, subsidizing purchases that were already going to happen. Predictive intent data lets us separate shoppers who are still deciding from ones who are already locked in, so the offer goes where it can actually change behavior. Audiences built that way have driven 25 to 40% lower cost per acquisition in our testing compared to broader targeting. It really comes down to using the data you already have more precisely, not spending more to reach everyone.
7. As retail, media and commerce become increasingly connected, how is this changing the way brands think about the customer journey?
Retail, media and commerce blurring together means the old idea of a linear path to purchase doesn’t really hold up anymore. Brands need audience intelligence that connects purchase history with real-time intent, so they can act on where a shopper actually is today, not where a segment says they should be. That’s part of why we’ve been connecting our own data and media across a broad set of retail partners, so brands aren’t managing that complexity retailer by retailer. The opportunity is to understand the journey in real time, rather than trying to predict it once a year.
8. What is one assumption about AI-driven marketing or promotions that you think marketers should rethink?
Probably that more AI automatically means better marketing. The technology is only as good as the data underneath it, and a lot of the current excitement is around generative AI, which is great for content but wasn’t built to tell you who to target or why. The distinction that actually matters is whether a recommendation traces back to real transaction data. Marketers who keep that distinction clear will be better positioned to get meaningful value from AI.
About Pia Ostos:
Pia Ostos is the EVP and President of Martech at Inmar Intelligence, leading the company’s marketing technology business across data, media, and commerce. Since joining Inmar in 2020, has held multiple leadership roles, including Chief Performance Officer, aligning talent, culture, and brand to support strategic priorities. With 12+ years of brand management experience at Fortune 500 companies including Procter & Gamble, Hasbro, and Wayfair, Pia is known for launching and scaling businesses through innovative strategy, go-to-market execution and a culture of excellence.
