Agentic artificial intelligence is embedding itself into everyday shopping and enterprise workflows, turning simple prompts into fully executed tasks and streamlining online interactions that improve customer-facing and supply chain interactions.
The promise that beckons is hands-off commerce, whether you’re a consumer or business, but the reality remains mixed. In B2C, retailers are rolling out AI agents across customer interactions. In B2B, adoption is more measured as firms weigh security and governance. Agentic AI is no longer just a promise; it’s an advancing reality, although one still shaped by caution.
Walmart’s “super agents” strategy, for example, now powers customer, supplier, associate and developer interactions, aiming to make AI the primary interface for commerce and operations.
“Agents can help automate and simplify pretty much everything that we do,” Walmart Chief Technology Officer Suresh Kumar said at a July event, adding the timing reflects how “customers are ready,” using AI in nearly everything.
Sparky, the retailer’s customer-facing assistant, handles tasks from product recommendations to budget planning.
During its second-quarter 2025 earnings call in August 2024, Walmart revealed that generative AI helped enrich or improve over 850 million catalog data points, a task that would have required 100 times the workforce without AI.
On Thursday (Aug. 21), during the company’s second-quarter 2026 earnings call, CEO Doug McMillion said Sparky will develop agentic capabilities over time.
“As we improve and scale Sparky, we’ll make it even smarter and more personalized,” he said. “It’ll be the primary digital vehicle for discovery, shopping and for managing everything from reorders to returns … The other super agents we’re building include one for associates that’ll bring everything into one place, from scheduling to sales data. [There’s also] one for our suppliers, sellers and advertisers that they will use to manage things like onboarding, orders and campaigns.”
Rival Amazon is also moving forward on agentic AI. The company said agentic AI customers, or autonomous software agents acting on behalf of users, are a key driver of future growth. These AI agents are treated as enterprise customers, using AWS tools to automate retail transactions.
Amazon CEO Andy Jassy said during a second-quarter earnings call July 31 that the company is preparing for enterprise AI agents to shape demand across the platform.
B2B Transformation via Agentic AI
In a nod to the shifts occurring in commercial settings, Transcard integrated agentic AI into its SMART Exchange platform in April to automate vendor onboarding and know your business (KYB) processes. AI digitizes vendor management with high-touch, customized interactions while reducing manual steps.
In another example, Infosys BPM said in May that it rolled out AI agents within its Accounts Payable on Cloud solution, which is powered by its Topaz generative AI and Azure’s AI stack. The solution autonomously processes invoices with more accuracy and efficiency.
Mastercard launched its Agent Pay initiative in May, bringing agentic payments to the fore. The offering introduces agentic tokens that enable AI agents to transact on behalf of users with trust and security. Mastercard is partnering with Microsoft, IBM and others to scale agentic commerce into enterprise operations.
Visa’s Intelligent Commerce APIs embed agent capabilities directly into transaction flows, ensuring AI agents can shop and pay autonomously.
In B2B, deploying autonomous AI agents raises complex security and governance challenges that extend beyond traditional automation. Risk, trust and oversight are critical, especially in high-stakes areas like finance and procurement.
B2B ecosystems are “built on deep trust and low risk tolerance,” where autonomous agentic systems must be carefully controlled to avoid catastrophes such as misrouted payments or broken agreements, PYMNTS wrote June 10.
Trustly Chief Legal and Compliance Officer Kathryn McCall framed the shift as a “governance revolution,” urging adoption of “bounded autonomy,” including sandboxed environments, scoped permissions, audit logs, human-in-the-loop controls, and forensic replay to ensure every action can be traced and reversed.
AI agents should be treated as non-human actors with unique system identities, complete with explainability, oversight and kill switches, she said.
Across retail and business, agentic AI is evolving from proof-of-concept to foundational infrastructure. As AI evolves, it will increasingly define how commerce is initiated, executed and completed.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.