Conagra Hands AI the Cost Side of Snacks — Mondelez and Hershey Are Already Winning the Shelf Side

Project Catalyst is Conagra's response to a six-year volume slide and three straight years of negative organic growth. Announced December 19, 2025, and amplified at the February 17, 2026 CAGNY conference, the program aims to re-engineer \"big and core\" CPG processes — finance, tax, supply chain, demand planning, marketing — using AI, data and automation. Management has not disclosed a single dollar figure for Catalyst savings, but it has set the productivity hurdle: above 4% of cost of goods sold in FY27, with a long-term 4% annual productivity target layered on top of an underlying ~5% inflation backdrop. The company also flagged about $40M of FY27 capital expenditure earmarked for Catalyst and a separate $125M supply-chain resilience investment.

Mondelez International has reframed AI as a discovery-and-margin tool, not a back-office one. The Oreo maker committed roughly $40M to a generative-AI marketing platform built with Publicis and Accenture that management says is cutting content production costs by 30%-50%. More strategically, Mondelez hired a global lead for agentic commerce in April 2026 and rebuilt Oreo's data infrastructure after finding the brand appeared in only 10% of AI cookie-recommendation queries — that figure has since climbed to 70% as Mondelez unblocked AI crawlers and structured product data for LLM retrieval. The company also overhauled its $3.5B digital commerce strategy to optimize for AI-search-driven traffic.

Hershey, for its part, made AI the centerpiece of its March 31, 2026 Investor Day at the NYSE under the banner \"Lead Next Generation Snacking.\" Management tied AI directly to revenue and margin levers: a five-year innovation pipeline weighted toward premium and better-for-you snacks, a more automated and AI-enabled supply chain, and a long-term adjusted EPS growth target of 35% through 2030. Q2 2026 results, reported July 30, delivered on the early read — net sales of $2.79B (+6.6%), adjusted EPS of $1.90 (32.9% above consensus). Crucially, the call framed agentic-AI visibility as the new shelf share — a metric that did not exist two years ago and that consumer-goods CFOs are now being asked to forecast.

The $1.4B figure is real and material — it is Conagra's FY26 net cash flow from operating activities, down from $1.7B a year earlier. Free cash flow tracked lower, to $979M from $1.3B, as capex rose and dividends consumed another ~$670M. The cash engine is what funds Conagra's $7.1B debt-load paydown (net leverage 3.83x at year-end, down from prior year) and the $40M-$165M of strategic investments it is layering on for FY27. The dividend cut to $0.70 annualized from $1.40 frees roughly $1B of incremental cash over three years — most of which management says will go to deleveraging toward a 3.0x target.

Conagra runs 42 manufacturing plants, 25 distribution centers and ~17,400 employees that produce more than 5,000 SKUs across four segments (Grocery & Snacks, Refrigerated & Frozen, International, Foodservice). Its biggest snack brands — Slim Jim, Duke's, Angie's BOOMCHICKAPOP, Vlasic popcorn, Healthy Choice — are sold predominantly through Walmart, Kroger, Costco and Albertsons. That downstream concentration is now the AI risk: 81% of Conagra sales come from #1 or #2 brands in their categories, but \"category leadership\" no longer maps cleanly to LLM recommendation share. FTI Consulting's July 2026 \"New Shelf Space\" survey found that traffic from generative-AI platforms to merchant websites surged 4,700% year-over-year, and that 40%-50% of consumers now use AI for product research. Mondelez's own retail partners told Digiday they expect 30% of site traffic to be agentic by 2028.

Upstream, Conagra sources from a sprawling co-manufacturer and ingredient-supplier base (its careers postings still advertise \"manufacturing partners and co-packers\" as a planning responsibility). The supply-chain resilience investment — $125M incremental in FY27 — is explicitly tied to lower costs via inventory-days reduction and product-mix reassessment. That suggests Project Catalyst's upstream payoff will be measured in working-capital turns and freight optimization, not in capacity expansion — and reinforces that the AI bet is structurally defensive, not offensive.

The private-label wave of the past decade took roughly ten years to lift private-label dollar share from ~17% to ~23%. The agentic-commerce shift is doing something similar but on a months-not-years timeline, because the bottleneck is no longer shelf placement but structured data. FTI Consulting's report introduces the term \"LLM Shelf Share\" — how an AI model represents a brand versus competitors — and ranks it as critical as physical or digital shelf placement. A brand that dominates Google page one but is invisible on Claude, ChatGPT or Gemini forfeits the next purchase funnel before a consumer ever types a search.

That is the real margin re-architecture the brief gestures at. Three transmission channels are working at once: (1) assortment rationalization — AI agents pick the best two or three SKUs in a category, so long-tail branded items lose distribution and the surviving items must compete on price; (2) trade-promotion compression — AI agents comparison-shop in real time, eroding the negotiating advantage of large brands with big trade budgets; (3) marketing-cost compression — Mondelez's 30%-50% content-cost cut is the proof point, and once one peer demonstrates that ceiling, every other brand's marketing P&L gets re-marked. The combined effect on incumbent CPG margins is negative if you are not on the right side of the data plumbing.

Short-term (days to quarters): three near-term catalysts will determine whether Project Catalyst is a credible turnaround or a re-arranged deck chair. First, Conagra's Q1 FY27 earnings on September 30, 2026 — the first print under new CEO John Brase (effective June 1, 2026) and the first quarter to absorb the dividend reset and the brand-building reinvestment (+$40M, +14%). Second, Brase's first major strategic reset announcement, expected alongside FY27 guidance updates — management already eliminated the COO role in late July, a signal that organizational simplification is the first Catalyst deliverable. Third, Walmart's GroceryShop 2026 disclosures (September) on how Sparky surfaces mid-tier brands, which will be the first large-scale test of whether Conagra's #1/#2 brand share translates to AI shelf share.

Long-term (one to three years): the structural question is whether AI cost-out can keep Conagra's adjusted operating margin above 11% while peers like Mondelez and Hershey compound margin via AI-driven mix-shift and content-cost deflation. McKinsey's $160B-$270B annual CPG profit-pool estimate is the ceiling; the question is whether it accrues to brands (Mondelez, Hershey are spending to capture it), retailers (Walmart's Sparky ad inventory is now priced into the Walmart Connect business), or platform owners (OpenAI, Anthropic, Google). Conagra's positioning — internal process automation only — means it captures the productivity slice but not the discovery slice, and that is a margin ceiling rather than a floor.