Those are not early-adopter numbers, that’s a channel forming in real time.

Bain estimates that 30-45% of US consumers already use generative AI to research and compare products. I would argue this number is already much higher. McKinsey projects the US agentic commerce market alone could reach $1 trillion by 2030. Another piece of Bain’s research suggests AI agents could account for 15-25% of US e-commerce sales by the end of the decade (with grocery and household essentials leading adoption since those are the categories where convenience matters more than brand loyalty).

The structural problem isn’t awareness. Most CPG executives have heard of agentic commerce. The problem is that the entire branded CPG playbook, from package design to shelf placement to trade promotion to retail media, was built to influence a human being standing in an aisle making an emotional decision. AI shopping agents don’t stand in aisles, they don’t respond to end caps or eye-level placement or the color psychology on your bag. They parse through structured data, they compare verified claims, they optimize for consumer-specified parameters (i.e. price, ingredients, ratings, availability, sustainability certifications, etc.) and they surface the product that best matches the query. Not the product with the biggest media spend. Not the product with the most shelf facings. The product that has the best data.

That’s a different competitive game than the one most brands are playing today. For example, when Amazon Rufus recommends products, independent analyses suggest it favors Amazon’s own brands at rates significantly disproportionate to their market share. That tells you the architecture of this new shelf - the platform that controls the agent controls the recommendation; and the recommendation is the new shelf placement.

Moreover, the infrastructure buildout is accelerating faster than most CPG companies are tracking. Google launched the Universal Commerce Protocol at NRF in January 2026 (co-developed with Walmart, Target, Shopify, Etsy and Wayfair) creating an open standard that lets AI agents discover products, negotiate purchases and complete checkout without a human ever touching a screen. OpenAI and Stripe have their own competing protocol, the Agentic Commerce Protocol, already live inside ChatGPT’s instant checkout. Amazon Rufus handles over 250 million daily queries (roughly 14% of all Amazon searches) and converts those users at 60% higher rates than traditional search. Perplexity launched Comet, an AI agent that shops across retailers on your behalf, and Amazon immediately sued to block it. A federal court issued this temporary block in March but was struck down by the Ninth Circuit court just days later. The legal fight over who controls agentic commerce is being litigated right now; and the outcome will shape how products get discovered, compared and purchased for at least the next decade.

While Google’s UCP and OpenAI’s ACP are the early attempts to create a standardized protocol for any AI agent to discover and transact with any merchant, “open” in platform economics has a history that usually ends with the platform extracting more value than it distributes. The brands that built their digital strategies around Facebook organic reach in 2014, then Amazon search ranking in 2018, then retail media in 2022, know exactly how this cycle works. The channel starts open, the economics look favorable and then the platform tightens the algorithm and charges rent.

The difference this time is that the intermediary isn’t just controlling the shelf. It’s replacing the shopper. When an AI agent makes a purchasing decision based on structured data and preference algorithms, the consumer relationship (the very thing that brand equity is supposed to represent) runs through a layer of code that the brand doesn’t own, can’t see and increasingly can’t influence through traditional marketing. Right now, most brands aren’t visible to that code. Product data lives in retailer-managed systems, uploaded through portals designed for human category managers, stored in formats optimized for planogram software - not for agentic parsing. The brand’s own website often hasn’t even been meaningfully updated since the direct-to-consumer push of 2019.

An AI agent trying to compare a mid-market protein bar against competitors has to scrape inconsistent data from five different retailer sites, none of which match the brand’s own nutrition claims format. Meanwhile, the private label alternative has clean, structured, machine-readable data because the retailer built it that way from the start. The agent doesn’t know or care which product is the “brand leader”, it surfaces the product that has the better data.

Most of the conversation around AI in CPG right now centers on internal deployment (demand sensing, supply chain optimization, pricing models, etc.). Those matter, but they don’t touch the question agentic commerce is forcing - Is your product being found by the AI agent that’s making the buying decision on your consumer’s behalf?

Structured product data that agents can parse without scraping. Standardized claims with third-party verification an algorithm can read. Commerce infrastructure that lets your product be discovered, compared and purchased inside an AI conversation without the consumer ever visiting your website or the retailer’s. The gap between where that data needs to be and where it actually sits today, for most brands, is enormous.

The companies treating agentic commerce readiness as a 2028 initiative are making the same mistake as the brands that treated e-commerce as a 2015 initiative in 2010 - risking arriving only after the shelf is already built, the protocols are already set and the algorithms have already learned which brands to recommend (and which to skip). You can’t shortcut your way to AI discoverability, tou can’t acquire it in a deal unless you’re specifically buying data infrastructure and digital commerce capability and every adoption curve in the data suggests critical mass is closer to 2027 than 2030.

The shelf is going invisible, and agents don’t read packaging.

By mid-2025, a quarter of buy now pay later users had already started financing grocery purchases (up from 14% the year before). Gen Z adoption was at a third. Since then, every indicator of how deep and how frequent the behavior has become has gotten worse. Klarna is live at Walmart and Target, in-store and online. Affirm runs at Costco. Afterpay operates through mobile wallets at major chains across the country. The infrastructure for financing your weekly grocery run didn’t exist five years ago. Now it’s at every major register in America.

This isn’t a payments story. It’s a consumer relationship story - and most of the CPG industry doesn’t have the infrastructure to see it happening.

The stress underneath is broader than any single payment method. The personal savings rate is hovering around 4%. Consumer confidence dropped to 84.5 in January, the lowest reading since 2014. Layer on $1,300 per household in annual tariff-driven cost increases and the picture is a K-shaped recovery that scan data flattens into a single line. Upper-income households are absorbing the price increases (perhaps annoyed, but without changing behavior). Lower-income households are either trading down to private label and discovering they’re getting 90% of the product at 60% of the price (which turns a temporary switch into a permanent one), or financing branded purchases through installment plans. Both show up as transactions. One is durable volume while the other is volume on a timer. The aggregate number that lands in the category review mixes the two cohorts into a single trend line that looks like a modest, manageable decline while hiding that the brand is increasingly dependent on a narrower, wealthier base and a lower-income cohort that’s one credit squeeze away from deteriorating.

The problem for branded CPG isn’t that consumers are using installment plans. It’s that scan data (the foundation of every category review, every pricing model and every volume forecast in the industry) can’t distinguish between a transaction funded by cash flow and a transaction funded by four payments. Nielsen tracks that someone bought the product. It doesn’t track that they split the payment because their grocery bill outran their checking account balance.

So when a CPG company reads scan data and concludes “volumes are stabilizing” or “the consumer is holding,” they may be right about the transaction and wrong about the consumer. A portion of the volume that looks like loyalty is actually inertia financed by debt. And debt-financed volume doesn’t behave like income-financed volume. It doesn’t erode gradually, it falls off when credit tightens or defaults spike or the consumer hits their installment ceiling and has to make actual trade-off decisions at shelf.

This is a first-party data problem and it exposes the gap at the center of most CPG operating models: The majority of branded manufacturers are fully intermediated through retailers, working off syndicated data that arrives 60 to 90 days after the transaction, with zero visibility into the financial health of their consumer base. They don’t know if their repeat purchaser is a financially secure household that chose their brand or a stretched household that hasn’t gotten around to switching yet.

That’s always been a strategic disadvantage. In an environment where a growing share of grocery spending is financed rather than funded, it becomes an analytical blind spot with direct P&L risk.

Consider what a company with deep consumer relationships would do differently. First-party data and direct consumer channels (those with a real sensing capability, not just a CRM list) let you see purchasing pattern changes before they show up in syndicated scans. You would be able to identify when your consumer base is under financial stress by watching order frequency, basket composition, and product-level substitution behavior in real time, while allowing you to respond with pack size adjustments, promotion timing and mix shifts that protect both the consumer relationship and the margin. You could answer the question that scan data can’t: is this consumer loyal, or leveraged? Are they choosing your brand because it earns the premium, or are they still on the shelf because Klarna made the decision painless enough to not revisit? That’s not a nuance, that’s the difference between a volume base you can build on and a volume base sitting on a credit bubble.

If you’re a branded manufacturer reading this and thinking “we can’t answer that question”, you aren’t alone - Most can’t. But you can start closing the gap faster than you’d think.

1. The first move isn’t building a DTC channel or investing in connected packaging. It’s overlaying what you already have. Take your syndicated scan data and cross-reference it against publicly available consumer financial stress indicators at the ZIP code level (delinquency rates, savings rates, BNPL penetration where the data exists, etc.). You won’t get individual-consumer visibility, but you’ll get a heat map of which geographies your volume is most exposed. If your strongest scan data markets overlap with the highest financial stress indicators, that’s your vulnerability map; It changes how you allocate trade spend, where you test pack size adjustments and which markets get promotional support versus price holds.

2. The second move is retailer data partnerships. Most major retailers have loyalty card data that reveals purchase frequency, basket composition shifts and trade-down behavior at the household level. This data is far more granular than anything syndicated panels provide. The retailers won’t hand it over for free, but the joint business planning conversations are already happening. The question is whether you’re using those conversations to ask about volume trends or about consumer financial health. Same meeting, different question, completely different strategic value.

3. The third move is the one that takes longer but matters most - Building your own first-party consumer sensing capability. That means something different for every company: it could be a DTC channel, a loyalty program, connected packaging with QR-driven engagement, or social listening infrastructure that tracks real-time sentiment and substitution intent. The mechanism matters less than the principle - you need a signal that tells you what your consumer is doing and feeling before the scanner tells you what they bought.

None of this requires a two-year digital transformation. The ZIP-level overlay is a thirty-day project. The retailer data conversation is already on the calendar. The first-party build is a longer arc, but you can start scoping it now with a clear mandate: we need to know whether our volume base is loyal or leveraged before the next earnings cycle.

Looking forward, I’d expect every CPG company heading into Q1 earnings season (i.e. PepsiCo on April 16, Nestlé on the 23rd and P&G on April 24 and the entire sector reporting through May) to talk about volume trends, pricing strategy and consumer resilience; But the number to listen for isn’t in the earnings release. It’s the one none of them can report - what share of their volume is financed rather than funded?

The companies that figure out how to see that number, through direct relationships, first-party data, or partnerships that give them visibility past the transaction, will be the ones that see the floor before it drops. The rest are running a forecast model with a growing blind spot that appears to get bigger every quarter that installment financing spreads deeper into the grocery aisle.

When a growing share of your consumers are financing their groceries, “the consumer is holding” isn’t a data-backed conclusion, it’s a hope, and hope is usually a lagging indicator.

A store brand. Sold in a warehouse. With virtually no traditional media spend.

That’s not a private label success story. That’s a structural break in how the CPG value chain works. For decades, the model was clear: brands innovated and built consumer relationships, retailers provided distribution and shelf space. The brand was the product. The retailer was the infrastructure. Kirkland Signature flips that entirely — the retailer is the brand, and the brand is the product. And at $90 billion, it’s not an experiment. It’s the new reality.

The scope is what should concern CPG executives most. Kirkland isn’t winning in one or two categories. It spans vodka, diapers, olive oil, batteries, laundry detergent, coffee, vitamins - every aisle, every occasion. Kirkland diapers are made by Kimberly-Clark, the same manufacturer behind Huggies. The olive oil meets the same sourcing standards as premium Italian brands. The quality gap that used to justify national brand pricing has effectively closed, and consumers know it.

And the numbers say they’re choosing accordingly. Costco topped Caliber’s 2026 U.S. Trust & Like Poll, dethroning Amazon, with a score of 79, up year-over-year at a time when tariff anxiety crushed consumer confidence and Target fell three points. Circana’s data shows U.S. private label sales crossed $330 billion total, with Gen Z adoption on pace to surpass boomers by mid-2026. The generation that was supposed to be brand-obsessed is choosing the warehouse club label. At this point, Kirkland isn’t private label. It’s the label.

The question I keep coming back to is what it actually means when a retailer’s house brand out-revenues P&G. Brand equity still gets talked about in most boardrooms like it’s a permanent asset; something accumulated over decades that compounds on its own. But it doesn’t. It erodes the moment you stop earning it. And the information asymmetry that used to protect it collapsed the moment someone could Google “who makes Kirkland diapers” in the aisle and find out it’s Kimberly-Clark. When Kirkland offers comparable quality at a lower price, under a name that 147 million cardholders trust enough to renew at a 92.3% rate, the old moat doesn’t drain slowly. It just stops being a moat.

I don’t think the answer for national brands is trying to out-Kirkland Kirkland. You’re not going to beat a $90 billion private label on price or breadth. But Kirkland’s dominance has a boundary: it wins in categories where the quality gap has closed and the purchase decision comes down to trust and value. In categories where genuine differentiation exists (specialty nutrition, clinical skincare, performance formulations, etc.), the brand still owns the relationship. The answer is figuring out what a dedicated brand can offer that a retailer’s label fundamentally can’t. Genuine product innovation that pushes categories forward, not line extensions that rearrange what already exists.

The tools exist. Real-time consumer sensing and scraped demand signals can spot unmet needs before a retailer’s buying team reacts. Generative AI is compressing product development timelines, reformulation at scale, rapid sensory optimization, computational flavor and fragrance design. Pipelines that used to take years can move in months; First-party data and DTC channels can build a consumer relationship that doesn’t depend on shelf position at all. And this isn’t theoretical. The two companies still ahead of Kirkland in the revenue rankings are the ones already moving this way: Nestlé compressed its product ideation cycle from six months to six weeks, and PepsiCo’s PepGenX platform achieved similar gains across its development pipeline. It’s not a coincidence that the CPG companies Kirkland hasn’t caught yet are the ones investing most aggressively in innovation speed.

But here’s the part most companies haven’t solved (and why generative AI hasn’t yet the silver bullet), that kind of innovation velocity requires R&D, marketing, supply chain, and commercial teams to operate as one system, not four functions that hand work to each other over the wall. The technology to move faster exists. The cross-functional orchestration to actually use it - Insight to formulation to shelf in a single, coordinated motion is what most CPG organizations still aren’t wired for. That’s the real gap. Not the tools. The operating model.

Subscribe now

McCormick announced that it’s combining with Unilever’s entire food business in a deal valued at $44.8 billion.

That means the company that already owns Frank’s RedHot, French’s, Cholula, OLD BAY, and Lawry’s will now also own Hellmann’s, Knorr, and Marmite. The combined entity will generate $20 billion in revenue and project $600 million in annual cost synergies. Unilever shareholders will own 55% of the new company, McCormick shareholders 35%, with Unilever retaining a 9.9% stake..

The deal structure tells you everything about why this is happening. McCormick is paying $15.7 billion in cash for a portfolio where Knorr and Hellmann’s alone represent roughly 70% of sales. These aren’t distressed brands — they’re category leaders that simply don’t fit the story Unilever wants to tell anymore. Unilever’s pivot is toward personal care, where growth rates are faster and margins are richer. Their food business was performing fine. It just wasn’t performing at the pace that justified the capital and attention it was consuming relative to their beauty and wellness portfolio.

What caught my attention is how cleanly this maps to what every other CPG giant is doing simultaneously. P&G is exiting entire product categories. Reckitt divested its home care portfolio — Air Wick, Woolite, Easy-Off — for $4.8 billion to concentrate on eleven “powerbrands.” Nestlé initiated the sale of its €5 billion global water division, separating Perrier and San Pellegrino from core operations. Kimberly-Clark’s $48.7 billion acquisition of Kenvue just cleared shareholder approval, creating a $32 billion consumer health company. European CPG deal value nearly tripled to $56 billion in 2025.

This isn’t a trend. It’s a phase change. The cost takeout playbook that defined CPG strategy for five years has run its course, and the companies that squeezed every efficiency from their existing portfolios are now asking a harder, more honest question: which brands should we even own?

The technology piece is what makes this wave different from every previous restructuring cycle. AI-powered demand forecasting and real-time POS analytics are making it possible to evaluate brand performance with a level of granularity that didn’t exist three years ago. When you can see that clearly into category-level contribution, the underperformers — or the mismatches — become impossible to ignore. McCormick’s CEO Brendan Foley called the Unilever food business “one we have long admired.” That’s not flattery. That’s a company that’s been watching the data and waiting for the portfolio thesis to align.

The question I keep coming back to is what this means for the middle of the market. These divested brands aren’t failures — they’re orphans of strategic focus. And for PE-backed platforms, mid-market operators, and focused acquirers willing to invest in them, this reshuffling represents a once-in-a-cycle opportunity to pick up proven consumer awareness at a fraction of what it would cost to build from scratch. The industry isn’t shrinking. It’s sorting itself. And the companies that emerge with high-conviction, focused portfolios and the data infrastructure to run them — those are the ones that’ll set the pace for the next decade.