How Nortek is using AI to make more time for customers

Turning reactive support into room for customer work

Keeley Morris is Nortek HVAC Global Europe’s Commercial Operations Manager, overseeing a broad range of functions, from inside sales and customer service to design and quoting. She first looked at Zendesk in 2022, initially seeking live messaging for the company’s new e-commerce website. “I was simply looking for a live messaging functionality. It was very much trying to get self-service on the website, get your own price, find your own part, track your own delivery.”

The goal extended beyond adding a digital channel. Keeley wanted the inside team to spend less time on repetitive questions whose answers already existed in product information, and more time supporting customers, handling escalations, and following up on quotes. As she put it, “It’s about getting my team, the inside team, to try and help drive sales.”

To build support for the change, Keeley asked employees which tasks felt repetitive or frustrating. The team helped identify common requests and develop FAQs and knowledge content, giving the people who would use the system a role in shaping what it could take on.

Bringing customer conversations together

Before Zendesk, customer history was split across sales and technical Outlook inboxes, phone calls, and other channels. Keeley recalled, “We had a sales inbox, we had a technical inbox, and nothing was visual together.” This fragmented view limited agents’ visibility into the full interaction history and made it harder to coordinate customer issues across departments.

Zendesk brought phone, email, and messaging interactions into a shared workspace, giving agents a consolidated view of customer history across the sales and technical teams. The team could see who was contacting them and what had already been discussed. Nortek’s first rollout also connected product and technical information from its website, giving AI a knowledge base for answering product questions. The rollout did not yet connect Zendesk to the ERP, however, so agents still had to look up order details in NetSuite separately. Bringing that order information into the workspace became a next step in the roadmap.

Making years of product knowledge useful

Nortek had a decade of sales brochures and installation manuals in PDF format. The content was valuable, but technical tables were formatted differently across documents, sometimes with merged cells. When the Zendesk AI read the PDFs as they were, those variations made it harder to interpret the information consistently.

The team converted ten years of brochures and manuals into web pages and cleaned up the tables. In Keeley’s tests, the change made the answers accurate: “As soon as we did that, the accuracy was 100% whenever we asked the AI a question.” It also exposed inconsistencies between documents, such as a product brochure and an installation manual describing different information. Those discrepancies were in Nortek’s source material too, so the team could correct them rather than relying on an answer that might repeat outdated or conflicting details.

In a one-month review of technical questions, Keeley found that Zendesk AI could answer 60% of the questions the team examined. The result showed the potential of making trusted technical knowledge easier to access, while the content cleanup helped address the gaps that could undermine answer quality.

Extending assistance to common parts enquiries

Keeley estimates that common parts enquiries make up about 60% of the team’s request volume. Nortek is developing email auto-assists for these enquiries. The Zendesk AI reads each message and suggests relevant parts, prices, and lead times for an employee to review.

The team has updated its procedures to support the workflow. Keeley describes the auto-assist as “like magic.” For now, employees approve each suggestion before it is sent, keeping the team in control as Nortek builds confidence in the recommendations.

The next stage would connect Zendesk with Nortek’s NetSuite ERP and CRM data. Today, agents still have to open NetSuite separately to find order information. Bringing that information into the Zendesk workspace would give agents customer and order context together, especially for requests like dispatch updates. It would also provide a foundation for automating more of the request from the first email through to a complete answer.

A lesson in prioritizing and building trust

Zendesk’s range of capabilities initially made it tempting to try to launch everything at once. Looking back, Keeley says prioritizing the most repetitive, time-consuming tasks and preparing the underlying content were important parts of the rollout.

She also spent longer testing AI before launch because she was concerned it might invent answers. Her experience shifted that view: the AI works from the information it is given, and inaccurate source information can produce an inaccurate answer. Improving the knowledge base was therefore part of making the AI useful, and helped Nortek identify product information that needed attention. Looking back, she says her biggest lesson was to “perhaps trust the AI more, because it would have saved me a lot of testing time.”

By starting with shared customer history, useful technical content, and a set of practical support needs, Nortek has built a foundation for further automation. The early results show how that foundation can help the team keep pace with demand while creating more time for complex service and proactive customer engagement.