Designing a support system that empowers users, anticipates needs and evolves with your business
We brought the public facing and customer service sides of the house together to create a world class support experience.
Collaborators
32 engineers, 5 product managers, 2 directors & 1 vp
Industry
E-commerce
Designer Setup
5 ux designers + 1 content designer
Timeline
1 year
Why
Customer service defines your brand’s reputation–it’s the most personal connection between your business and its customers.
Great customer service boosts your business.
It feels personal
It remembers who the customer is and what interactions they’ve had in the past
It doesn’t burden the customer
Goal
To track the entire user journey to ensure every merchant feels recognized and supported by Shopify. Whether exploring the Help Center or speaking with an advisor, we want to ensure that no one started from square one–always keeping merchants one step closer to having their problem solved.
Challenge
Shopify’s vast feature set, each uniquely adaptable to a merchant's store, required deep domain expertise to address the wide variety of unique queries effectively. On top of this, integrating AI and LLM tools to empower users presented its own complexities—a Pandora’s box of opportunities and risks as the technology continued to evolve, demanding careful balance between automation and human insight.
Outcome
The outcome of our efforts was a seamlessly connected platform where the Help Center and Contact Center Tools continuously informed each other. The Help Center’s AI knew what interactions merchants had with support previously and what stage and needs their store would have, while support advisors gained insight into the steps merchants had already taken before reaching out. This integration empowered merchants to self-resolve issues more frequently, improved customer service ratings, reduced support contact volume, and enabled advisors to respond more quickly and effectively to each merchant request.
How we started
We started by observing both support advisors and merchants using our platforms, gaining firsthand insight into their workflows and challenges. To complement this, we gathered quantitative data to identify the most common merchant queries in the Help Center and the ones advisors found hardest to solve.
From there, we quickly developed journey maps to illustrate these experiences, providing a clear picture of the user journey. These maps became essential tools for aligning the team, setting objectives, and defining a roadmap to tackle the most impactful problems.
Understanding the journey allowed us to focus on metrics that delivered significant value to Shopify.
Reducing
Training costs
Advisor contact rate
Average Advisor response time
Average handle time
Advisor onboarding time
Advisor churn
Raising
Customer service ratings
Resolution rates
Workshops facilitated
Contextual inquiries
Observing merchants and advisors in their real environments, uncovering pain points and workflow inefficiencies that informed more effective, user-centered design solutions.
Jobs to be done
Identifying the core motivations and needs of both merchants and advisors, ensuring our solutions targeted the most impactful problems and aligned with user goals.
Our research efforts were crucial to the project’s success, providing a solid foundation for feature proposals and design decisions grounded in user insights. By involving teammates from every discipline in these workshops, we not only enhanced their understanding of design thinking but also encouraged cross-functional contributions. This collaboration ensured diverse perspectives were included and helped the team better understand the rationale behind our outputs and recommendations.
Challenges to overcome
Implementing LLMs (Large Language Models)
Many users approached our AI tools with skepticism, shaped by previous experiences with ineffective chatbots. Rebuilding trust required us to design a system that consistently delivered value, demonstrating its usefulness from the first interaction.
We had to distinguish between tasks that simply needed to be done—such as repetitive, time-consuming actions—and moments where we could empower business owners to grow their expertise. Tedious tasks were automated to save time, while more complex or creative challenges became opportunities for digital assistance to educate users, helping them become more skilled and confident entrepreneurs.
Advisors were already juggling multiple windows and dozens of tabs just to manage their daily tasks, leaving little room for additional tools or information. We needed to differentiate between features that would truly empower advisors and accelerate their work versus those that would add to their cognitive load. The solution was to consolidate these scattered resources into a single, unified interface—a "single pane of glass"—that served up exactly what advisors needed, precisely when they needed it, enabling them to work more efficiently without overwhelming them.
Solutions
North Star
Keep the burden off the merchant
We do more so that they can do less.
Fastest resolution
Self resolution is often faster than speaking with an advisor.
Right resources the first time
Right articles, people, solutions.
Help Center Search
We moved beyond search results, to provide not only answers to their initial question, but also to the questions they didn’t know to ask.
Traditional long-form help articles often overwhelm merchants with information they don't need to solve their immediate problem. It’s time-consuming to sift through it all, especially when many issues have multiple root causes.
By combining AI-powered inputs with insights into a merchant’s past actions, we identified the most likely root causes upfront. This approach allowed us to deliver a streamlined solution in a concise, easy-to-digest bulleted list.
Help Center Advisor Loops
Merchants often entered the Help Center seeking answers, with many attempting to resolve their issues on their own. However, when they ultimately needed an advisor, the process broke down—merchants had to recount their entire journey, leaving them frustrated and feeling like they were starting from scratch.
To fix this, we connected the dots. We logged merchants’ actions in the Help Center and provided that information to advisors before the chat began. This allowed advisors to start the interaction with context, confirm the merchant’s issue, and often have a solution ready to propose.
Return Visits
On return visits to the Help Center, merchants shouldn’t have to remember where they left off or search for an answer they received weeks or months ago.
We made it easy for merchants to pick up right where they left off. By cataloging all their searches and the answers they received, we eliminated the need to start from square one. Merchants can also review past interactions—whether they or a colleague from their company reached out to Support—allowing them to continue unresolved issues seamlessly and stay informed on previous resolutions.
Impacts
+4%
In customer service rating
+10%
Self resolution by merchants in the Help Center
-6%
In Advisor contact rate
Research Assistant
We removed noise and provided a single pane of glass to help advisors focus on delivering answers instead of scrambling for resources.
Previously, advisors had to manage multiple windows and tabs to find the resources needed to resolve a merchant’s issue, slowing them down and adding unnecessary stress.
We streamlined this process by pulling key information—like the merchant’s activity, past tickets, and similar resolved cases—into a single, personalized dashboard. This gave advisors quick, clear insights tailored to the inquiry.
Advisors stayed in control, using the AI’s suggestions to craft responses with a human touch. This balance of efficiency and empathy allowed them to resolve issues faster while maintaining a personal connection with merchants.
Advisor Solutions
Differing Opinions
This wasn’t the first solution we explored, but it proved to be the best.
Some early explorations carried high hopes for what new LLM technology could achieve. One idea was to have the AI read the transcript and automatically draft a response for the advisor to review and send.
However, the results fell short. The AI-generated responses were easy to spot as robotic, leaving merchants—who had already waited in a queue—feeling deceived and frustrated that they hadn’t reached a real human.
Additionally, it didn’t save advisors much time. They still had to review and edit the responses, verify citations, and address inaccuracies when the summaries missed key details. The technology simply wasn’t ready to step into the role we envisioned, leading us to pursue other solutions.
Keeping Current
Products change quickly and the information about them needs to keep up.
Support Advisors engage with Shopify's product information more frequently—and under higher pressure—than anyone else in the company. They are often the first to identify when information is incorrect or insufficient for its intended purpose.
To address this, we empowered advisors to flag subpar information for review by our content teams. When time allowed, advisors could also provide additional context on what was wrong.
On the backend, we paired this system with a process that tracked solutions advisors implemented during interactions that weren’t yet documented. These solutions were cross-referenced with existing Help Center articles, allowing us to suggest updates to our content team, ensuring our resources stayed accurate and relevant.
Voice & Tone
We collaborate with many offshore BPOs to support millions of merchants, but differences in training often result in inconsistent voice and tone across interactions.
To address this, we built an LLM-powered tool that helps advisors refine their messaging while preserving Shopify’s brand voice. The tool automatically corrects grammar and spelling errors, trims overly flowery language, and provides guidance for handling sensitive scenarios with extra care. Importantly, it ensures advisors maintain a natural, human touch in their responses without robbing them of the opportunity to craft their own messages. This approach balances consistency with personalization, allowing advisors to communicate authentically while aligning with Shopify’s standards.
Impacts
-3 weeks
Required training time for new advisors to be ready to speak with merchants on their own
-35 seconds
Reduction in Average response time from advisors to customers
+2%
First contact resolution rate