#pymntsAI

The Universal Commerce Protocol (UCP), which was introduced in January, is now letting U.S. shoppers buy items from Etsy and Wayfair, from within Google’s AI Mode in Search and the company’s Gemini app, Vidhya Srinivasan, vice president and general manager, Ads and Commerce at Google, wrote in a Wednesday (Feb. 11) blog post.

This UCP-powered checkout will soon be extended to Shopify, Target and Walmart, according to the post.

In addition to these merchants, hundreds of tech companies, payments partners and retailers have contacted Google with interest in integrating the standard, according to the post.

“In 2026, agentic commerce is no longer just a concept, it’s reality,” Srinivasan wrote in the post. “It will transform how we shop, from discovery to decision, while helping brands differentiate themselves.”

When Google announced UCP on Jan. 11, the company said the agentic commerce standard is designed to “work across the entire shopping journey.”

“UCP establishes a common language for agents and systems to operate together across consumer surfaces, businesses and payment providers,” the company said at the time in a press release. “So instead of requiring unique connections for every individual agent, UCP enables all agents to interact easily.”

The release added that UCP is designed to work across verticals and is compatible with industry protocols such as Agent2Agent (A2A), Agent Payments Protocol (AP2) and Model Context Protocol (MCP).

PYMNTS reported Jan. 12 that the launch of UCP puts Google squarely in the standards race and that rivals such as OpenAI, Amazon and Microsoft are developing their own agentic commerce systems.

The competition will not center on features alone, but on which protocols gain adoption and become embedded in everyday buying behavior, according to the report.

Wayfair said Jan. 12 that it participated in the development of UCP for agentic commerce and would soon enable shoppers to check out directly from Wayfair without leaving Google during their research.

“Wayfair is investing in AI-powered discovery wherever our customers are — whether that is on our own app or across external AI platforms,” Wayfair Chief Technology Officer Fiona Tan said in a press release.

A study by MIT Sloan describes what it calls a “productivity paradox” in AI adoption. Drawing on firm-level data, researchers found that early adopters frequently see limited or uneven performance improvements when AI tools are layered onto fragmented workflows rather than embedded within redesigned operating models.

“AI isn’t plug-and-play,” said University of Toronto professor Kristina McElheran, a digital fellow at the MIT Initiative on the Digital Economy and one of the lead authors of the study.

In many cases, companies invest heavily in algorithms, automation systems and predictive tools without reworking decision rights, retraining employees or integrating data flows across production lines.

The result is a widening gap between AI spending and realized value. While some leading firms ultimately unlock gains, others stall despite deploying advanced systems.

AI Layered on Legacy Systems

According to the study, early AI deployments in manufacturing tend to be additive rather than transformative. Companies introduce predictive maintenance models, computer vision inspection tools or demand forecasting algorithms, but leave underlying processes intact. This creates friction between automated recommendations and human workflows, limiting measurable productivity improvements.

Manufacturing environments are particularly complex. Production lines depend on tightly sequenced tasks, supplier coordination and legacy industrial control systems. When AI is introduced without harmonizing these systems, it can increase coordination costs in the short term. Workers must interpret AI outputs, reconcile them with existing protocols and adjust routines that were optimized for a pre-AI environment.

The study suggested that productivity gains emerge only when firms pair AI with organizational redesign. That includes reassigning decision authority closer to data sources, standardizing data architectures and investing in workforce retraining.

Companies that treat AI as a tool upgrade see marginal benefits. Those that treat it as an operating model shift capture more durable returns. “Once firms work through the adjustment costs, they tend to experience stronger growth,” McElheran said. “But that initial dip the downward slope of the J-curve is very real.”

Hardware, Infrastructure and Strategic Control

The paradox also intersects with the infrastructure buildout supporting AI.

As reported by The Astute Group, the emerging alliance between OpenAI and Foxconn signals a strategic shift in AI hardware manufacturing. The partnership reflects growing recognition that AI competitiveness depends not only on software models but also on control over advanced manufacturing capacity.

As PYMNTS reported, Dassault Systèmes and Nvidia have entered a long-term strategic alliance to develop a unified industrial framework for mission-critical AI applications across multiple sectors. The two tech giants are building a shared industrial AI platform that aims to produce validated digital twins to boost speed, accuracy and sustainability across engineering, manufacturing, biology and materials science.

In this environment, short-term productivity slowdowns may reflect transitional costs as firms recalibrate production systems to accommodate more data-intensive operations. Reconfiguring plants for sensor integration, edge computing and AI-driven quality control requires capital expenditures and temporary disruptions.

Where ROI Becomes Advantage

According to Microsoft, the highest ROI appears in predictive maintenance, quality inspection, energy optimization and supply chain orchestration.

The company cited a study estimating that manufacturers adopting a unified data platform and scaling AI across operations could see up to 457% projected ROI over three years. The analysis ties returns to reduced downtime, improved yield, lower defect rates and better inventory management when AI is integrated across both IT and operational technology systems rather than deployed in isolated pilots.

Microsoft points to predictive maintenance and quality inspection as high-impact use cases. AI models trained on sensor and production data can identify equipment anomalies before failure, reducing unplanned downtime and maintenance costs. Computer vision systems deployed on production lines can detect defects earlier in the process, helping manufacturers reduce scrap, rework and warranty exposure while improving throughput consistency.

For all PYMNTS digital transformation coverage, subscribe to the daily Digital Transformation Newsletter.

The cryptocurrency exchange says its Agentic Wallets, announced Wednesday (Feb. 11), let users quickly equip agents with autonomous spending, earning and trading capabilities.

“AI agents are everywhere – answering questions, summarizing documents, and assisting with tasks,” Coinbase wrote on its blog.

“But today’s agents hit a wall when they need to actually do something that requires money. They can recommend a trade, but they can’t execute it. They can identify an API they need, but they can’t pay for it. They’re stuck waiting for human approval at every financial decision point.”

Coinbase says this new offering builds on its AgentKit, created to build wallets into agents. Agentic Wallets, the company added, were designed to “give any agent a wallet.” The project is based around x402, the payments protocol for autonomous AI use cases.

“Already battle-tested with over 50M transactions, x402 enables machine-to-machine payments, API paywalls, and programmatic resource access without human intervention,” Coinbase said.

The wallets also offer users protections, such as session caps that let them determine the maximum an agent can spend per session, as well as controls on individual transaction sizes.

The new offering follows last year’s launch of “Payments MCP,” a Coinbase tool that supports agentic commerce by giving AI agents access to on-chain financial tools such as wallets, on-ramps and stablecoin payments.

PYMNTS wrote last year about the importance of digital identity measures as AI agents take on more tasks in the Web3 space.

In this environment, “identity is portable, composable and verifiable,” that report said. AI agents will need to navigate this new landscape “with clarity and consistency, whether they are interacting with humans, smart contracts or each other.”

Much like websites have domain names and SSL certificates to signal trust, AI agents will need identity protocols to demonstrate their legitimacy and intent. This could be especially crucial in high-stakes areas like finance and payments, where trust is key and failure can be costly.

“Every crypto protocol needs to consider how this new technology will fit into their operations,” Harrison Seletsky, director at SPACE ID, told PYMNTS in an interview. “Verifiable on-chain identities will simplify AI-to-human and AI-to-AI interactions, making them safer by giving AI agents a humanly recognizable name, thereby cleaning out bots and bad actors from misrepresenting themselves.”

The new arrangement will embed Udemy’s content directly into OpenAI’s interface via an app in ChatGPT, the companies announced Wednesday (Feb. 11). This integration combines Udemy’s catalogue of more than 290,000 courses from 90,000 instructors, with ChatGPT’s conversational AI capabilities, the announcement added.

“For approximately 800 million weekly active users of ChatGPT, one of the most common use cases is learning,” Udemy said in the release. “With Udemy’s app in ChatGPT, users will be able to access cutting-edge technical and soft skills content on Udemy directly from the AI tool, eliminating traditional barriers to content discovery and engagement.”

Aside from the course content itself, the companies say the platform offers things like interactive assessments, practical labs and competency validation.

“What sets this apart from traditional AI tutoring is our focus on building real expertise through trusted content, practical assessments and validated skill development rather than just providing answers,” said Hugo Sarrazin, Udemy’s president and CEO.

The partnership comes amid increasing AI adoption among both consumers and businesses, as PYMNTS Intelligence research has found.

Findings from “From Assistive to Agentic AI: Consumers Wade Into Autonomous Commerce,” shows that consumers are increasingly comfortable letting artificial intelligence (AI) “sense, decide and act on their behalf,” PYMNTS wrote Wednesday.

The research shows that 71% of consumers are interested in using agentic AI for health and wellness management, 70% for travel planning, 69% for tasks like grocery shopping and meal planning, and 66% for managing their bills.

“The headline finding is not adoption alone,” that report added. “It is a conditional adoption. Consumers want agentic AI to reduce friction in daily life, manage complexity and handle recurring decisions. At the same time, they insist on controls, reversibility and recognizable payment credentials once money is involved.”

Research from another PYMNTS Intelligence report, “CFOs Push AI Forward but Keep a Hand on the Wheel,” showed similar caution in the business world.

As covered here Wednesday, agentic AI promises to shift much of the chief financial officer’s function away from time-consuming, manual numbers crunching in favor of big-picture tasks such as scenario modeling and strategic analysis.

“But allowing an algorithm, not your finger on a mouse, to send millions of dollars to a supplier in an instant is a high-stakes, high-risk endeavor,” PYMNTS wrote. “So is allowing an agent to pull the trigger on shifting big dollars from one budget bucket to the next. That’s why CFOs aren’t yet all-in on turning over the reins.”