Retail AI Report: Where the Money Is Moving

This week the frontier labs traded blows on price and capability, but the more telling story sits underneath the model announcements. Agentic AI is quietly graduating from a developer novelty into an always-on tool for the back-office functions that make up most of retail headcount. At the same time, retailers are locking AI into their core budgets, and the data suggests the value is migrating from customer-facing chatbots toward the margin-sensitive work of forecasting, pricing, and fulfillment. 

OpenAI ships GPT-5.6

Artificial Intelligence

OpenAI broadly released GPT-5.6 in three tiers this week: Sol for heavy reasoning and coding, Terra for balanced everyday work at roughly half the cost of GPT-5.5, and Luna for fast, cheap tasks. The launch arrived alongside a new agent called ChatGPT Work and followed a staggered rollout tied to a U.S. government pre-release safety review. CEO Sam Altman said Sol is 54% more token-efficient on agentic coding tasks.

Why it matters: GPT-5.6 is the model most large retailers will standardize on for internal productivity and agentic workflows, and Terra's halved cost directly improves the economics of high-volume tasks like product-copy generation and customer service. When cost per task drops by half, workloads that were too expensive to automate last quarter become viable this quarter. ChatGPT Work's ability to complete multi-hour tasks end-to-end pushes the conversation from AI assistants that help staff toward AI workers that own a process in merchandising, marketing, and operations.

Source: OpenAI

Anthropic's Cowork goes cross-device, and the usage data reframes the value

Artificial Intelligence

Anthropic moved its Claude Cowork agent to the cloud, turning a desktop-only tool into a cross-device platform where a task can start on a laptop, run autonomously in the background, and be reviewed from a phone. More revealing than the feature was the data behind it: across 1.2 million anonymized sessions from more than 600,000 organizations, the majority of Cowork activity was knowledge work rather than software development.

Why it matters: For two years the enterprise AI pitch centered on developer productivity, yet the people actually running these agents are analysts, planners, and operators. Those are precisely the merchandising, planning, and back-office roles that dominate retail payrolls. Leaders sizing their AI opportunity should stop benchmarking against engineering efficiency and start measuring the hours spent on reporting, reconciliation, and vendor communication, because that is where the near-term gains are concentrated.

Source: VentureBeat

Grok 4.5 escalates the price-and-efficiency war

Artificial Intelligence

SpaceXAI launched Grok 4.5, its first model since going public and its first co-trained with coding startup Cursor. Priced at $2 per million input tokens and $6 per million output tokens, the company claims roughly twice the token efficiency of leading rivals and a top ranking on a legal-agent benchmark, though its own numbers show it trailing frontier leaders like Fable 5. The model is positioned as a workhorse for coding, agentic tasks, and knowledge work.

Why it matters: A third serious low-cost frontier model deepens the price war that keeps pushing down the compute cost behind retail applications, from pricing engines to customer chat. For procurement, this is leverage: multiple vendors now offer near-parity performance at falling prices, which argues for model-agnostic architecture rather than single-vendor lock-in. Grok's strength in spreadsheet and Office automation also signals that agents capable of producing finished business deliverables are now cheaply available to retail teams, not just prototypes.

Source: TechCrunch

AI becomes a core retail budget line, with technical debt as the ceiling

Retail

A KPMG survey of 250 retail executives found that 52% now spend $50 million or more annually on digital technology, with 28% allocating between $100 million and $250 million. AI and automation, including generative and agentic systems, ranked among the top areas for increased investment. Nearly half of respondents named technical debt as a barrier to going further.

Why it matters: These figures confirm AI has moved from pilot budgets to core spending at scale, which raises the competitive pressure on laggards to fund equivalent programs or cede ground. The prominence of technical debt is the more actionable signal: the constraint on AI returns is no longer model quality but the state of a retailer's data and infrastructure. CTOs and CIOs should read this as a mandate to modernize the plumbing first, because agents cannot orchestrate what they cannot cleanly access.

Source: Retail Dive

SPS Commerce bets on shared data standards for the agentic supply chain

Retail

SPS Commerce unveiled AI-enabled fulfillment capabilities including PDF order automation, deeper SAP S/4HANA and Shopify integration, and shared performance dashboards, positioning its network as the connective tissue for what it calls an agentic supply chain ecosystem. The company also became a founding member of the Commerce Operations Foundation and backed the launch of the Order Network eXchange (onX), a shared operational language for orders, inventory, and fulfillment data.

Why it matters: Supply chain automation lives or dies on whether clean, standardized data can flow between retailers, suppliers, and logistics systems, and most retailers still lack that foundation. An industry-wide standard like onX could finally close the gap between selling channels and fulfillment execution that quietly caps how much can be automated. Operations leaders should watch adoption closely, because standards win by network effect, and the retailers who wire in early will find their trading partners easier and cheaper to automate against.

Source: FreightWaves

Gen-AI in retail approaches $1.55 billion, and the value is moving to the back office

Retail

A new market report projects generative AI in retail stores will grow from $1.35 billion in 2025 to $1.55 billion in 2026, reaching $2.62 billion by 2030. The growth is attributed to e-commerce expansion, richer customer data, and demand for personalization. The applications drawing investment include demand forecasting, inventory management, dynamic pricing, visual merchandising, and support chatbots.

Why it matters: The forecast gives executives a benchmark for sizing their own investments against the market, but the more useful signal is where the money is flowing. Spend is shifting away from customer-facing chatbots toward margin-sensitive functions like demand forecasting, dynamic pricing, and assortment. Those are the levers that move gross margin rather than deflect service tickets, and retailers still anchoring their AI strategy to chatbots risk investing in the least valuable corner of the market.

Source: GlobeNewswire (ResearchAndMarkets)

What to Watch

Watch how quickly the halved-cost model tiers from OpenAI and SpaceXAI translate into re-scoped automation projects, since falling token prices tend to expand the list of tasks worth automating within a single planning cycle. Keep an eye on onX adoption among major retailers and suppliers, as early network momentum will decide whether it becomes an industry standard or another well-intentioned consortium. And with public AI backlash rising, expect vendor features aimed at trust and mindful use to shape how comfortably both employees and customers accept the agents now moving into daily retail workflows.