AI Agents Are Coming for the Instruction Manual

AI Agents Are Coming for the Instruction Manual

When a SharkNinja customer scans a QR code on a product box, there is no confusing manual or phone tree waiting for them. There’s an artificial intelligence (AI) agent that knows exactly which product they have and walks them through setup, step by step, according to a May 19 blog post by Salesforce.

The maker of Ninja kitchen appliances and Shark vacuum cleaners built the agent to replace the moment most customers dread: fumbling through a dense instruction manual trying to figure out where to start. The unboxing agent answers follow-up questions in context and surfaces product videos when a visual would help. A human is not involved unless the customer asks for one.

“The QR code on the box is the new instruction manual,” said Carolin Duerkop, technology transformation partner at SharkNinja, said in the blog post. “Scan it, and you’re in a conversation with someone who knows exactly which product you have and what you’re trying to do.”

That shift points to something larger happening in enterprise AI. Most deployments have focused on answering customer questions faster and reducing pressure on call centers. The technology has been moving from chatbots that retrieve information to agents that guide customers through tasks, PYMNTS reported in April. SharkNinja’s agent is an early example of what that looks like in practice.

SharkNinja launches about 25 new products each year, according to the blog post. The unboxing agent is designed to scale across the company’s catalog. Getting it there required solving a content problem first: The company’s product manuals are dense, image-heavy PDFs that are not built for AI consumption.

The team used AI to generate draft setup steps from each document, then had content authors review and refine before publishing.

Advertisement: Scroll to Continue

Stan Konopka, vice president of digital technology at SharkNinja, told Salesforce the underlying lesson was simple: “The technology’s there, but it’s only as good as the data and the knowledge and the product catalogue,” he said.

The unboxing agent was not SharkNinja’s first. The company built a digital concierge earlier, a single agent on its website handling troubleshooting, order tracking and product recommendations. SharkNinja had been preparing its website to include AI-powered features for months before going live, with the goal of letting its service teams focus on higher-value customer interactions, Digital Commerce 360 reported in October. That agent handled more than 250,000 conversations, Salesforce said in its blog post, giving the team the foundation to extend into something more ambitious.

SharkNinja’s chief information officer, Velia Carboni, has been direct about what the company is building toward. “We want a lifetime relationship with that one customer, not just a one-time transaction,” she told Chief Marketer in a January 2025 interview.

The unboxing agent fits that logic. A customer who has a good first experience with a product is more likely to return. One who gets lost in a setup manual may not.

To make sure the agent was ready for customers, SharkNinja developed what it calls an “attack the bot” program. Call center representatives spend time each day trying to break the agent, testing edge cases no standard QA script would surface. The team also runs weekly conversation engineering sessions with customer experience colleagues to refine how the agent communicates.

“Get subject matter experts who know their stuff to test the AI,” Konopka said in Salesforce’s post. “That’s where the magic happens.”

The approach reflects a broader lesson emerging across enterprise AI deployments. Companies including ClassPass have found that moving from basic support automation to deeper AI integration requires restructuring how work is organized, not just adding new tools, PYMNTS reported. In its February earnings call, Booking Holdings reported that AI helped cut customer service costs by 10% even as volumes rose 10%, illustrating the efficiency gains available to companies that get the implementation right.