Salesforce is pushing artificial intelligence deeper into the retail transaction, expanding technology that allows AI agents to help shoppers find and compare products, answer questions and move through checkout while drawing on the commerce systems behind the purchase.
The company’s Dreamforce conference, held in San Francisco from September 15 to 17, provided a closer look at how Salesforce sees AI moving beyond product discovery and customer service. Its latest technology connects agents with commerce, customer data and inventory systems, allowing them to take actions on behalf of shoppers and businesses.
One of the clearest retail examples is Carter, a shopper agent designed to guide customers through product discovery, comparison and purchase. Its introduction comes as AI is already becoming part of shopping behaviour in Canada: Salesforce research previously reported by Retail Insider found that 39 per cent of Canadian consumers had used an AI tool for shopping during the previous year.
Carter Moves AI Closer to the Transaction
Carter can answer product questions, compare merchandise and support checkout within a conversation. Retailers can customize the agent, including giving it their own name, while authenticated shopper context can draw on information such as previous orders and preferences. Other functionality includes cross-selling and upselling.
US sporting goods retailer Hibbett provides an early example. Salesforce says the retailer launched its Hibbett AI implementation in six weeks and that the agent now handles 90 per cent of its core shopper journeys.
The connection with checkout moves the technology beyond the product research and recommendation functions that have characterized much of consumer-facing generative AI. Connecting the agent with commerce systems allows it to participate in the purchase instead of simply helping a customer decide what to buy.
Connecting AI With Physical Stores
The commerce technology extends into physical retail through Salesforce’s Shop the Store functionality, which allows shoppers to select a location and search merchandise available at that store. Salesforce is also working to unify inventory across B2C Commerce and its point-of-sale technology, providing a common inventory record incorporating sales, returns and transfers.
An AI shopping agent could therefore draw on information spanning digital commerce and physical stores. A customer could ask about a product, determine whether it is available at a particular location and continue toward an online purchase, store pickup or visit to the store using the same underlying inventory information.
For omnichannel retailers, the technology connects agentic commerce more closely with their store networks instead of treating it strictly as an e-commerce function.
When Shopping Starts Outside the Retailer’s Website
Salesforce is preparing its commerce technology for shopping journeys that begin outside a retailer’s website or app. Through Headless Commerce and its broader Headless 360 architecture, the company is making business capabilities available to authorized AI agents while retaining established permissions, workflows and governance.
Salesforce has announced commerce integrations involving external AI environments including ChatGPT, Google’s Gemini and Google Search’s AI Mode. An integration with Google’s Universal Commerce Protocol is expected to support product discovery and commerce through Google AI experiences while allowing participating retailers to remain the merchant of record.
This work began before Dreamforce. Salesforce announced a major expansion of Agentforce Commerce in July, including the general availability of Shopper Agent, Buyer Agent and Merchant Agent. The Dreamforce announcements provided a broader view of how those commerce capabilities fit within the company’s AI strategy.
The development has implications for digital customer acquisition. If more product discovery occurs inside AI interfaces, retailers will need product information, inventory and other commerce data that can be accessed and interpreted by AI systems as well as conventional search engines and their own digital storefronts.
Canadian Shoppers Are Already Using AI
Salesforce research previously discussed with Retail Insider found that 89 per cent of Canadian commerce professionals believe AI and large language models will become essential to product discovery within the next year. Thirty-four per cent said they were already using agentic AI, while another 33 per cent planned to do so within six months.
Independent research from Retail Council of Canada and Leger found that 11 per cent of Canadian shoppers had used AI assistants for shopping research during the previous 30 days. Among those users, 46 per cent said AI helped them find products faster, 43 per cent said it made comparisons easier and 40 per cent said it saved time.
The findings show AI gaining a place alongside search engines, retailer websites, marketplaces, social platforms and stores as consumers research purchases.
Retail’s Data Problem
Only 30 per cent of Canadian retailers surveyed by Salesforce said their customer data was fully unified across sales, marketing, commerce and service, creating a potential obstacle as retailers give AI agents access to more parts of the transaction.
An agent helping someone research running shoes can operate with relatively limited information. An agent being asked whether a particular shoe is available in the shopper’s size at a specific store, whether a loyalty offer applies and whether it can complete the purchase requires dependable inventory, pricing, promotions, customer identity and transactional data.
Salesforce’s broader research suggests the issue extends beyond Canada. In a separate study of organizations deploying AI agents, only 31 per cent had fully unified their data before deployment. Salesforce found that organizations that unified their data first reported reaching meaningful returns from agents sooner than those that had not.
The research also identified clean and accessible data, clearly defined agent responsibilities and predetermined paths for escalating issues to humans among the factors associated with successful deployments.
“Having a sound data strategy is really going to be important,” Caila Schwartz, Director of Industry Insights at Salesforce, previously told Retail Insider.
The Infrastructure Behind the Agents
At Dreamforce, Salesforce unveiled AIforce, an interface layer designed to connect AI with Salesforce data, workflows, business logic, permissions and governance. Headless 360 similarly makes Salesforce business capabilities available to authorized AI agents, with Headless Commerce extending the approach to commerce functions.
Salesforce also introduced Koa, its first CRM reasoning model, developed using NVIDIA’s Nemotron architecture and post-trained using synthetic enterprise scenarios informed by Salesforce’s CRM experience. The company says Koa matches or exceeds leading models on its CRM benchmark while producing three times fewer errors, based on its own testing.
These technologies are designed to give AI agents controlled access to the same systems businesses use to manage customers, operations and transactions.
Holiday 2026 Could Provide an Early Test
Salesforce predicts that 20 per cent of global e-commerce traffic during the 2026 holiday season will originate from AI agents. The figure includes consumer-facing shopping agents, autonomous back-end systems and automated competitive scrapers, so it does not mean one in five consumers will shop through an AI conversation.
The company estimates that AI influenced 20 per cent of global online sales during the 2025 holiday season, representing US$262 billion in spending. The figures are based on Salesforce’s own commerce data and analysis.
Physical stores will remain a major part of the holiday shopping environment. Salesforce research previously reported by Retail Insider found that 83 per cent of Canadian consumers expected to shop in stores during the 2026 holiday season, while Retail Council of Canada and Leger research has also found that stores continue to account for the majority of Canadian purchases.
That combination helps explain the relevance of connecting AI with store-level inventory. A shopping journey that begins with an agent could lead to an online order, store pickup or a visit to a location where the merchandise has already been identified as available.
What Agentic Commerce Means for Canadian Retail
The Dreamforce announcements show Salesforce extending AI agents further into product discovery, customer data, inventory and transactions while opening more of its commerce infrastructure to AI interfaces outside conventional applications.
Canadian consumers are already experimenting with AI as a shopping tool, and Salesforce’s research suggests many Canadian commerce organizations are moving toward agentic deployments. The company’s Canadian data also shows a gap between those ambitions and the underlying infrastructure, with only 30 per cent of surveyed retailers reporting fully unified customer data.
As agents take on more of the shopping journey, retailers will need accurate inventory, pricing, promotions and customer information available across channels. Salesforce is building technology to give AI greater access to those systems, while the quality and integration of retailers’ underlying data will determine what those agents can reliably do.
