Program report · filed November 11, 2026
AI in SalesMeasured report
Dell Product Lead Maps Four Fixes Retailers Need Before AI Agents Shop
Dell's Saurabh Pitkar says retailers need supply chain flywheels, headless APIs, agent monitoring and B2B guardrails as agentic commerce heads toward 2027.
By Daniel Okafor3 min read693 words
Program notes
- Dell's unified commerce platform powers $150 billion-plus in global transactions
- Pitkar names four retailer fixes: supply chain flywheel, headless commerce, omnichannel monitoring, B2B identity and guardrails
- B2B agentic commerce market is 3-5x the consumer market, per Pitkar
- Protocols in play include OpenAI's ACP, Google's Universal Commerce Protocol and AP2, Meta-Sierra's personal agent protocol, and Visa's payments protocol
- Voice agents can stay on support calls 30-40 minutes; fraud can scale across 20 agents at once

Dell's commerce platform processes more than $150 billion in global transactions, and the man who helped build it says retailers must make four structural fixes before AI agents replace human shoppers on their sites. Saurabh Pitkar, Director of Product Management for Agentic Commerce and Agentic AI at Dell, laid out the requirements in an interview with CX Today's Sean Nolan.
Pitkar's remit covers all third-party and external commerce integrations that sell Dell products through LLMs, marketplaces, social apps, and emerging channels such as Meta's Muse AI. His core claim: agents compress product discovery from hours to minutes, and minutes to seconds, because they fire 20 parallel searches and map intent to specifications algorithmically.
Where do consumers already gain?
Pitkar identified two proven consumer benefits. First, low-clarity intent: researching a DSLR camera once required opening 10 to 20 tabs across product pages and reviews, then making a judgment call. An agent now collapses that session. Second, high-clarity repeat buys: groceries, consumables, and seasonal purchases where the shopper trusts the agent to repurchase the exact same item without deviation.
Both points are asserted from Pitkar's own usage, not from published Dell conversion data. No sample size or methodology accompanied the claims.
What does this change for retailers?
Pitkar framed four investment areas, calling them his personal views rather than Dell policy.
- Supply chain flywheel. Structure product data, connect live inventory from distributors, factories, and warehouses in real time, and let pricing teams act on that feed. The goal: influence margin on external channels instead of dragging users to your website.
- Headless commerce. Expose operations like add-to-cart as APIs, not UX buttons. An agent should not have to identify and click a button in a browser. Following protocols such as OpenAI's Agentic Commerce Protocol (ACP), Google's Universal Commerce Protocol, or MCP makes capabilities machine-readable and scales the architecture.
- Omnichannel monitoring. Traffic, time-on-site, and journey data disappear when agents replace human visitors. Retailers need proxy metrics to infer agent intent indirectly.
- B2B identity and guardrails. In agent-to-agent deals, establishing both parties' identity and transaction limits is critical. Pitkar's example: a seller agent negotiating a computer order might exceed its allowed discount threshold — a guardrail violation.
How does marketing split its budget?
Nolan and Pitkar agreed that decades-old tactics need rethinking. Humans respond to nudges — "only two items left in stock," or a 30 percent price-drop banner. Agents do not.
"AI agents are pretty much looking at facts and consistency," Pitkar said. They query APIs, know category-level performance trends, and read exact stock counts — one, five, or ten — without reacting to urgency messaging.
Marketing dollars will split across both personas in the near and medium term, Pitkar said. Down-funnel metrics — revenue, conversion — still apply. Up-funnel attribution has, in his words, gone out the window, because agents do not disclose user intent or origin.
What happens toward 2027?
Protocol convergence tops Pitkar's opportunity list. The field already includes ACP from OpenAI, Google's Universal Commerce Protocol, a personal agent protocol from Meta and Sierra, Google's AP2, and Visa's agentic payments protocol. The internet itself was built on multiple protocols, he noted, but businesses need enough convergence to pick investments and transact securely.
His second prediction: commercial agentic commerce outgrows consumer. The B2B market is three to five times the consumer market, at minimum, and multiple sales motions could see step-change efficiency gains.
Challenges cluster on the business side. Fraud scales exponentially — an attacker who finds a loophole can deploy 20 agents on fraudulent transactions simultaneously. Liability remains unresolved: who owns damages and returns when agents transact? Anomaly detection and tight channel monitoring are Pitkar's stated defenses.
He also flagged operational stress tests. A voice agent can hold a customer-support line for 30 or 40 minutes without complaint, something no human caller sustains, and agentic search traffic arrives as rapid-fire queries that can spike infrastructure.
Nolan closed by naming liability, fraud, and trust as the questions business leaders must answer before agents carry cards, money, and private data at scale.
via cxtoday.com (Source)
More from Daniel Okafor
Show full bio
Correspondent covering regulation, policy and their impact on Sales technique and methodology, sales enablement at Quota Brief.
127 articles
Also rated
- AI Agents Are Now the Shoppers, and Brands Are Rewriting the Pitch
- Brands Now Sell to Two Audiences: Humans and AI Agents
- US Brands Rewrite Sales Pitches as AI Agents Start Shopping
- Perplexity's Computer Claims 3.5x Revenue on $30,000 of Meta Ad Spend
- Agentic AI Is Collapsing the Sales-Marketing Boundary