Why 1 in 3 people now call AI a ‘friend’: How agentic AI will transform retail and shopping

Keith Shaw: Most people consider generative AI as a question-and-answer tool, like a search engine. But new research shows more people are treating AI as a good friend and trusted advisor.

We’re going to talk about what kind of impact this will have across many areas on this episode of Today in Tech.

Keith Shaw: Hi everybody, welcome to Today in Tech. I’m Keith Shaw. Joining me on the show today is Brett Leary. He is with Accenture and is the Global Gen AI Lead for the retail industry. It’s a long title at Accenture — welcome to the show.

Brett Leary: Thank you, I appreciate it, Keith. Yes, that title does sound pretty long and fancy. Keith: So tell me about the research you’ve been doing in this space, because it feels like this is more of an emotional attachment conversation.

Usually that’s something psychologists or behavioral researchers might explore, but there’s also business impact here, especially in retail. Right? Brett: Absolutely.

When we started this research months ago, we didn’t expect to find this kind of emotional connection. We jumped on this early because we started to see interesting behavior, even from personal use.

You could go to ChatGPT or Google’s Gemini and ask for a product recommendation, and it would give you useful information to help you decide.

Brett: But then we saw people go further with their prompting — multi-turn conversations, deeper context about the product, and more personalized recommendations based on prior queries or preferences. The emotional bond emerged quickly. One of the findings was that one-third of users consider AI a good friend.

I was surprised by that. I’ve worked in technology for more than 20 years, and in the early days of the internet, I never considered Google a friend. I didn’t consider my toaster a friend — it was a tool. This shift is real.

Were you surprised by how many people treat AI like it has a personality? Keith: Are they projecting personality onto these tools, or are they experiencing them differently than maybe people like you and me?

Brett: I think it’s the latter. These LLM-powered tools — GPT, Gemini, Claude — create a perception of empathy, and that feels real to users. They listen, they respond quickly, and they adapt. Those are human-like traits we connect with.

Prior technologies like Amazon’s Alexa tried, but they often missed the mark. They were either overly eager to answer or interrupted at odd times, which frustrated users. Keith: Right.

Like when I’d ask for the weather and get a three-paragraph answer. Brett: Exactly. But ChatGPT doesn’t seem to trigger that same frustration. Maybe it’s because every interaction starts with positive reinforcement — “That’s a great question,” or “Happy to help.” Whether intentional or not, that builds emotional trust.

There are parallels to earlier digital simulations — The Sims, online avatars — where people formed emotional bonds with software. Technology has long been a tool for connection or escape.

Brett: When natural language processing improved and voice assistants like Alexa emerged, there were attempts to make interactions feel more human, but the technology wasn’t there yet. Today, NLP combined with generative AI has created a conversational tone that appears to understand you as a person.

Sometimes it can feel overly agreeable, but it still draws people in and encourages engagement. That emotional pull is powerful, and it has major implications for how people interact with technology — including in retail. We will definitely get into that.

Keith: Before we get into retail, one part of your research really stood out. You said 93% of active users would ask AI for help with personal development goals. We’re starting to see AI being used for things like life coaching and even therapy.

But that can go in a dangerous direction. There have already been reports of people being encouraged by AI to make harmful decisions. Let’s stay on the positive side for now — what about AI for self-improvement? Is that a good thing, or could people get too dependent on it?

Brett: We definitely see both sides. At Accenture, we focus on enterprise use cases, so we aren’t developing therapy bots, but we’re very aware of that trend. For companies and brands, the takeaway is this: emotional connection increases loyalty.

If customers feel like a tool understands and supports them, they engage more. That’s powerful in retail because emotional connection drives brand preference. It also opens new service opportunities. With AI’s ability to personalize, a grocery retailer could offer nutrition coaching. A fitness brand could offer AI-powered exercise planning.

Companies couldn’t scale these services before — but now they can.

Keith: That makes sense. I think people are already testing that. I once asked AI how to start a workout routine — not because I fully trusted it, but because it was a first step. Before generative AI, I’d probably have to ask a trainer or search for articles.

Now I can get a simple plan to start with. People do the same thing now with job hunting — using AI to improve resumes, prepare for interviews, or explore career goals. That’s what a life coach would normally help with. Brett: Exactly.

And platforms like LinkedIn and Indeed now openly integrate AI to support job seekers. It’s becoming the norm. AI is quietly embedding itself into everyday workflows. People won’t always realize they’re using AI — they’ll just feel like technology is more helpful.

Keith: In your research, did you see companies in the retail space already starting to integrate generative AI like this?

For example, using it as part of their customer experience — maybe like a fitness company that lets you ask questions and get customized suggestions from their product catalog rather than from the whole internet?

Brett: There have definitely been high-level discussions, but most companies are still in the early stages. Every brand is experimenting with proofs of concept, but few have scaled anything yet. Where we’re seeing adoption first is internal use cases — hiring, customer service workflows, content generation.

Companies want to move faster, but they’re cautious about deploying AI in customer-facing roles too quickly. They don’t want brand risk.

Brett: There’s a spectrum of adoption. Some large retailers — and even Amazon — are public and very aggressive with AI. Others have thinner margins, like grocery retailers, so they’re more careful. But everyone is experimenting.

The big challenge now is moving from pilots to production in a way that actually creates business value.

Keith: There’s been criticism in the media that generative AI hasn’t delivered the massive business value it promised. People expected an immediate transformation. Is that fair? Brett: Some of that criticism is fair. Many companies stayed too focused on small pilots — like isolated content tools or narrow use cases.

That won’t move the needle financially. What we’re seeing now is a shift toward applying AI to end-to-end workflows — merchandising, supply chain, customer service, product development. That’s where the value is.

And it’s not just generative AI — it’s gen AI combined with classic machine learning that unlocks real improvement: faster insights, better inventory decisions, reduced operational cost.

Brett: A lot of these functions — HR, finance, legal, customer service — are text-heavy, rules-heavy, and repetitive. Gen AI is extremely good at accelerating them. Companies are now combining generative models with existing data systems to rethink how work gets done. That’s the trend to watch.

Keith: That reminds me of the early days of data warehousing and business intelligence. In retail, the famous story was always that grocery stores discovered shoppers buying diapers were also buying beer. That insight led to smarter product placement.

Now AI is taking that kind of insight and scaling it, right? Brett: Absolutely. That idea still holds true — understand the shopper, anticipate behavior, personalize the experience. What’s different now is the volume of new signals.

Every time someone prompts ChatGPT about needing a refrigerator, sneakers, or a car, that’s a signal. Companies want to harness those early intent signals. They don’t want to wait for a search or a product page visit anymore — they want to know what customers want before they start shopping.

That’s a huge competitive advantage.

Keith: Your research also found generational differences in how people use AI. I’m a Gen Xer, and I tend to use AI as a tool — ask a question, get an answer. But younger users seem more comfortable trusting it as more than just a tool.

What did you see in the data? Brett: You’re right — the emotional connection does skew younger. Gen Z and millennials are more likely to refer to AI as a friend or advisor.

But the interesting finding was this: once people become active users — meaning they use AI at least once a week — the generational differences nearly disappear. Boomers, Gen X, millennials, Gen Z — they all start to show similar levels of emotional connection once they build the habit.

That was surprising. This isn’t just a young person trend — adoption drives trust.

Keith: Let’s shift into product recommendations. I hate the word “influencer,” but here we are. For years, people relied on influencers to tell them what products to buy. Now we’re hearing about AI becoming an influencer. I tested this myself recently when I needed a new refrigerator.

I went to ChatGPT and asked for recommendations. It gave me a list based on size, features, and reliability. It felt helpful instead of salesy. Is AI becoming the new product influencer? Brett: It’s already happening. This is one of the hottest topics in retail and consumer goods.

We’re seeing a shift from search-based product discovery to conversational discovery powered by AI. People trust these tools to narrow down decisions quickly. OpenAI even announced commerce integrations — users can now buy products directly inside ChatGPT. One in ten active users already uses AI for product research.

And that number is rising.

Brett: This changes everything for brands. The big question now is: How do we make sure our products show up in AI recommendations? It used to be about SEO — search engine optimization. Now it’s GEO — generative engine optimization, also called answer engine optimization.

Companies are trying to understand what kinds of content large language models prefer when generating recommendations. Keith: So this is like the new SEO game, but for AI? Brett: Exactly.

LLMs seem to favor content from authoritative sources — well-written articles, product reviews from reputable sites, structured product data, and trusted platforms like Wikipedia and Reddit. Brands now need content strategies optimized for AI, not just for Google search results. The game has changed.

Keith: Here’s what I found interesting in my fridge search. The first brand it recommended was Whirlpool. I told it I didn’t want Whirlpool, and it instantly pivoted to GE and LG based on my preferences.

That felt incredibly personal — much more than the static product grids you get on ecommerce sites. So I asked it: “How do I know these recommendations are trustworthy?” It said it uses data from Consumer Reports, Wirecutter, and expert review sites. That built trust.

There were no ads, no sponsored results. It felt honest. But my concern is — once monetization hits, won’t AI recommendations be corrupted? Brett: That’s a real concern. Right now we’re in a “golden age” of AI recommendations — high utility, no ads, minimal monetization pressure.

But once these AI platforms flip the monetization switch, sponsored answers will start appearing. When that happens, trust becomes the key differentiator. Companies that invest in transparency and explainability will win.

Brett: We’re already seeing the beginnings of this shift. Sponsored answers and answer engine ads are coming. Companies are trying to figure out how to show up in conversational product journeys.

Every major brand we work with is now asking the same question: “How do we stay visible when AI becomes the starting point for shopping?” Keith: And that’s where this generative engine optimization comes in — GEO or AEO. Brett: Yes, exactly.

Companies are now rethinking their content strategy to optimize for AI recommendations. They’re asking: What product information do LLMs find credible? How do we get cited as a source? Should we allow AI models to crawl our product catalogs?

Do we send structured product feeds to LLM companies like OpenAI, Anthropic, and Google? These are conversations happening inside every major retailer and consumer brand right now.

Keith: When I searched for a refrigerator, I started with ChatGPT. Once I narrowed it down, I went straight to the retailer to buy it. I never even visited the brand’s website.

That seems like a big problem for manufacturers like GE or Samsung — if AI becomes the trusted middleman, brands lose direct relationships with customers. Brett: Exactly right. In that journey, the brand loses visibility — and potentially loyalty.

That’s why brands want their product information and messaging represented accurately inside AI-generated answers. If AI is going to be the new product advisor, it becomes a powerful gatekeeper. Brands cannot afford to sit out that conversation.

Brett: This also raises questions about the future of loyalty. If I start every purchase with a neutral AI assistant, I might not care about brand loyalty anymore. If AI handles product discovery and comparison for me, price becomes the main driver — and that’s dangerous for brands.

It’s a race to the bottom.

Keith: But isn’t that already how ecommerce has been trending? Think about Amazon. I don’t always care who I’m buying from as long as I get the best price fast. But I do care with big purchases — home appliances, cars, big electronics. I want trust. I want reliability.

For low-cost items like batteries or socks, I don’t care. Is that what you mean by consideration level? Brett: Exactly. In retail, we call it purchase consideration. Low-consideration products are things like snacks and toiletries — customers are price sensitive. Mid-consideration products might be shoes or small electronics.

High-consideration items are things like appliances, furniture, or vehicles — purchase confidence becomes more than just price.

Brett: The higher the consideration level, the more customers care about things like product quality, service, delivery, return policies, installation, support. That’s where brands can differentiate — through service, not just price.

But if AI flattens out brand storytelling, then the only differentiator left is price — and that’s a problem for profitability. So brands now need to communicate their value story inside AI conversations — performance, warranty, sustainability, service, trust — so they aren’t reduced to a price tag.

Keith: Let’s talk about reviews. Ecommerce has always relied on customer reviews to build trust. But lately, those reviews have felt less reliable — too many fakes, too many bots, too many sponsored reviews. Do you think AI might fix reviews or make them worse? Brett: Both.

Generative AI is already summarizing product reviews on platforms like Amazon. Instead of reading 2,000 reviews about a refrigerator, AI generates a summary of what customers are actually saying. That’s helpful. But the trust problem remains — where is that summary coming from? Who controls it?

Was it influenced by a brand? AI needs to be transparent about where its output comes from.

Brett: This is why explainability has become critical. If AI recommends a product, it needs to show why — what sources it used, what criteria it evaluated. That builds trust. Explainability will become standard in AI-powered commerce. The brands that embrace transparency will win long-term customer trust.

Keith: Here’s another part of my fridge story — after I got AI recommendations, I didn’t buy directly through AI. I knew the retailer I wanted to buy from. It was a local appliance company I’ve used before. They’re reliable, they deliver fast, and they haul away the old appliance.

I trusted the retailer more than the brand or the AI. So I ended up calling them instead of buying online. I never even went to GE’s website. That seems like a problem for brands, right? Brett: It is.

And you’re not alone — loyalty often lives with the retailer, not the manufacturer. That’s why brands are racing to make sure AI reflects accurate product information, availability, pricing, warranty, and brand values. If brands don’t participate, they risk becoming invisible in the AI economy.

Brett: This goes back to content strategy and GEO. Brands now need to think about how they show up in AI-generated answers, not just on Google search results or retail product pages.

They need to ensure that the ground truth about their products — the features, benefits, data sheets, certifications, warranties — is correctly represented in AI models. If they don’t, AI will fill in the gaps with whatever it finds online, which may be outdated or inaccurate.

Brett: Loyalty is also at stake. If every product journey starts with a generic AI query like “What’s the best refrigerator under $1,200?” then pricing pressure increases and brand loyalty decreases. It becomes a commodity race, and that’s bad for margins.

So brands now need to emphasize differentiation — service, reliability, sustainability, personalization. Otherwise, AI will turn everything into a price war. Keith: Right.

And sometimes price doesn’t matter — service does. I went local because I trusted their service. They delivered, installed, and hauled away the old refrigerator. That mattered to me more than saving $50 online.

So as AI simplifies product choice, maybe retailers and brands will have to win on service and experience again. Brett: Exactly. And that’s where strategy is shifting — away from pure product competition and toward value-added services.

If pricing becomes transparent and AI levels the field, differentiation will move to experience.

Brett: We’re already seeing this in loyalty programs and post-purchase services. Delivery quality, installation, repair coverage, financing options, recycling or removal programs — all of this becomes part of the value story.

The clearer brands and retailers can communicate that value — especially inside AI conversations — the better positioned they’ll be to win.

Keith: This reminds me of early voice assistants. When Alexa launched, people wondered: “If I say, ‘Alexa, order more batteries,’ which battery brand does it send?” The concern was that whoever controlled the voice interface would control the sale.

Are we in the same situation again — but now with AI agents? Brett: Absolutely, yes. That’s a great comparison. Just like with voice assistants, brands now need to make sure they’re part of the AI conversation. If they don’t, they risk losing visibility and market share.

But unlike voice assistants, AI agents are smarter, more personalized, and more deeply integrated into the buying journey. So the stakes are much higher this time.

Keith: Speaking of AI agents — let’s go there. The next wave after chatbots is agents: AI that can take actions for you. We’re moving toward “agentic AI,” where AI doesn’t just recommend things — it does things. It books appointments, orders groceries, manages finances.

How do you see that affecting shopping? Brett: This is where things get really interesting. Yes, agentic AI — AI that takes action — is here. Early versions are already live. Google has Project Astra and Project Mariner. OpenAI has agent mode. Chinese platforms are already launching autonomous shopping bots.

And we’re seeing these tools move from recommendations to transactions.

Keith: Let me give you an example. When I shop for jeans on Amazon, it makes me jump through 16 dropdown boxes — size, cut, fit, color, price. It’s tedious. In a world of agents, I imagine saying: “Buy me the exact same jeans I got last time.” Done.

No dropdowns, no filters. Way easier. Brett: That’s exactly where this is going. Instead of navigating websites manually, agents will remember your preferences — your size, fit, style, budget — and shop for you. The more you trust the agent, the more freedom you’ll give it.

That creates a deep lock-in effect. But it also raises big questions — privacy, bias, monetization, brand control.

Keith: But I don’t see that world yet. Right now, agents still need a lot of hand-holding. For an agent to buy jeans for me, I’d have to give it a ton of personal information — sizes, brands, styles, budget.

I don’t think most people are ready to hand that over yet. Do you think we’ll actually get to a world where agents handle shopping for us? Brett: Yes, we will. The short answer is yes. Every major tech company is racing toward that future.

The question isn’t if — it’s when and who wins. Agents are already starting to remember user preferences through conversation history and stored profiles. Combine that with past purchases and location data, and they begin forming a pretty accurate profile of your shopping habits.

Brett: OpenAI has shown early demos where its agents actually browse websites on your behalf — scrolling, clicking, filling out forms. Companies like Etsy and Target have already tested integrations with AI shopping agents. It’s early and clunky, but progress is fast. These are version 0.1 prototypes.

In two to three years, this will feel normal. Keith: That’s both impressive and terrifying. I’m not sure I want something shopping for me behind the scenes.

Brett: That’s the big divide — excitement versus fear. People who value convenience will embrace agents quickly. People who worry about privacy will hesitate. But historically, convenience wins. We saw this with online shopping, smartphones, voice assistants.

Once a few people adopt and gain an advantage — saving time, saving money — everyone else follows.

Brett: Early projects from Google (Project Mariner), OpenAI (agent mode), and Chinese platforms like Alibaba and Meituan are already building full shopping agents. They’re testing things like: “Here’s my weekly grocery budget — order everything my family needs.” It’s not perfect yet, but you can see where this is going.

Keith: I can see retailers like Amazon doing this, but could a new company come along and become the universal shopping agent for everything? Or will agents be tied to specific retailers and brands? Brett: Great question.

There are two competing models forming: Closed agents – Retailers like Amazon or Walmart will build agents you use only within their ecosystem. Open agents – Companies like OpenAI, Anthropic, and Google want you to use their agents everywhere, across the entire internet.

Brett: We’re watching a platform war forming — like iOS vs. Android, but for shopping agents. Whoever owns the agent owns the customer relationship — and potentially the whole shopping journey.

Keith: I saw OpenAI announce something called “agentic commerce,” where you can buy products directly inside ChatGPT. But they also said the merchants will pay fees. So this is how AI companies will make money — by becoming middlemen in every transaction. Brett: Exactly. This is the monetization strategy.

Forget subscriptions. The real money is in transaction fees. AI companies want a piece of ecommerce. Sponsored product placement. Affiliate revenue. Marketplace listing fees. Personalized ads. It’s all coming.

Brett: Perplexity has already tested sponsored answers. They don't look like banner ads — they look helpful, like part of the conversation. That’s what brands need to prepare for: advertising inside conversations. Keith: So we’re in the golden age now — helpful recommendations, high trust.

But once agents become good and monetization kicks in, everything gets flooded with ads. Then the experience gets worse again. Brett: That’s a fair prediction. Unless — and this is important — companies commit to transparency and responsible AI practices along the way.

Keith: Let’s talk about the privacy side of this. For agents to work, they need personal data — shopping history, budget, health, lifestyle. But how much data are people actually going to allow AI to have? I don’t trust Meta, for example.

I don’t think I’d ever use a Facebook agent because I don’t trust what they’d do with that data. Brett: And you are not alone. Trust is now a competitive advantage. People will choose AI tools based on how responsibly they handle data.

Responsible AI isn’t just a compliance checkbox anymore — it’s a brand strategy. At Accenture, we call it Responsible AI: guardrails for privacy, data security, bias prevention, human oversight, and transparency. Companies that don’t earn trust will lose customers to those that do.

Keith: Here’s a funny but kind of scary scenario. Say my agent starts analyzing my grocery purchases and notices I’m buying too many Oreos.

Then maybe it goes, “Hey Keith, looks like you’re gaining weight — I’ll order you jeans one size bigger next time.” Then it tells me to go to the gym. I don’t know if I want an AI judging me.

Brett: That might sound funny, but it’s actually a real concern in AI design — overstepping boundaries. Personalization versus intrusion. Relevance versus creepiness.

In our consumer research, we saw three levels of trust emerging: Recommendation trust – “Give me product suggestions.” Advisor trust – “Give me guidance to help me improve.” Autonomy trust – “Act on my behalf.” Brett: Not everyone wants level 3. Some may not even want level 2.

AI tools will need to adapt to user comfort levels — or risk losing trust entirely.

Keith: And companies can’t bury this stuff in 80 pages of terms and conditions. Nobody reads those. People need transparency written in plain English: “Here’s the data we collect. Here’s how we use it. Here’s how you control it.” Brett: Exactly. And that is becoming standard practice.

The best companies will be upfront. Users should always know: When they’re talking to AI What data is being used Why a recommendation was made How to turn features off How to delete their data Brett: Explainability builds trust. Hiding AI makes people nervous. Revealing how it works earns loyalty.

Keith: And now that AI can act — take actions, make purchases — companies need even stronger guardrails, right? Brett: Yes. If AI is going to act as a personal agent, companies must follow responsible AI practices.

That means building systems that are fair, safe, transparent, and aligned with human values. It also means knowing when to keep a human in the loop — especially for high-impact decisions. Brett: This is also why many companies haven’t fully deployed AI agents yet. They’re afraid of reputational risk.

If an AI agent makes a harmful suggestion or behaves unpredictably, it can damage trust instantly. So companies are focusing on guardrails before they scale — security, bias mitigation, content control, and escalation to humans when needed. Keith: And we’ve all dealt with frustrating chatbots.

Half the time, you don’t even know if you’re talking to a human or an AI. And when something goes wrong, you just keep looping in circles. Brett: Exactly. And that’s why transparency matters.

Companies should clearly say: “You’re talking to an AI assistant” or “Here’s how to reach a human agent.” Users get frustrated when they think they’re talking to a person and later discover it was AI. That breaks trust.

Keith: Let’s go to the future. Big question: Will personal AI shopping agents be the breakout application of generative AI? Five years from now, will most people be using AI agents to shop? Or is something else coming that we’re not even thinking about yet?

Brett: I think AI-powered shopping agents will absolutely be one of the killer applications. But first, AI will redefine search. That’s happening right now. Search is shifting from links to answers to actions.

Instead of searching Google for “best winter boots,” people now ask AI, “Which waterproof boots should I buy for hiking in Vermont?” AI gives a tailored answer instantly. That’s a massive behavioral shift.

Brett: So phase one is AI-powered search and recommendations. Phase two is agentic AI — tools that act on your behalf. Whether agents become universal or fragmented by brand ecosystems remains to be seen. But yes, agents will become common. Keith: But humans still like going to physical stores.

I don’t think those are going away. Brett: Completely agree. In fact, in our research, physical stores are still the number one most trusted source for product decisions — people like touching, feeling, trying things.

What surprised us was that AI has now moved into the number two spot — ahead of friends and family.

Keith: Wow — so people now trust AI more than their friends when it comes to product recommendations? Brett: Yes — for active users. Once someone uses AI at least once a week, AI becomes their second most trusted product advisor, behind only physical stores. That’s huge.

That’s why brands are taking this seriously. And now they’re asking: if AI is influencing purchases in the store, what tools should store associates have? Should AI help them too?

Brett: We’re already working with retailers who are equipping store associates with AI assistants — voice tools, computer vision, product data copilots. Imagine asking a store associate, “Do you have this fridge in stock in stainless steel instead of black?” and they get the answer instantly from AI. That’s coming.

Keith: That would have helped in my situation. When my fridge broke, the first retailer I called said it was out of stock. The second told me I had to wait six weeks. The third finally helped me.

AI could help retailers avoid losing business just because someone couldn’t find inventory information fast enough. Brett: Exactly. Inventory accuracy is a massive opportunity area for AI. Product discovery will start with AI — but fulfillment still happens in the real world.

Retailers who connect AI with inventory and supply chain will win.

Keith: So the in-store experience is still critical. But now, AI is influencing the customer before they ever walk in. By the time someone gets to the store, they already know what they want because AI helped them narrow it down. Brett: Exactly.

The customer is coming in more informed than ever. That means store associates need to keep up. AI won’t replace them — it will equip them. It will help them find answers faster, personalize recommendations, and prevent the kind of friction you experienced during your refrigerator search.

We call this “AI-powered retail,” and it applies across the entire customer journey — before, during, and after the sale.

Keith: This is fascinating stuff. I could talk about this all day — and we probably will in a future episode when we get deeper into agentic AI and retail strategy. Brett, thanks so much for joining us. Brett: Thanks, Keith. I’d be glad to come back.

This space is evolving quickly, and we’re just getting started.

Keith: And that’s going to do it for this week’s show. Be sure to like this video, subscribe to the channel, and share your thoughts in the comments below. Join us every week for new episodes of Today in Tech. I’m Keith Shaw — thanks for watching.