•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
17
•
•
18
•
•
19
•
•
20
•
•
21
•
•
22
•
•
23
•
•
24
•
•
25
•
•
•
•
•
•
•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
17
•
•
18
•
•
19
•
•
20
•
•
21
•
•
22
•
•
23
•
•
24
•
•
25
•
•
•
•
•
•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
17
•
•
18
•
•
19
•
•
20
•
•
21
•
•
22
•
•
23
•
•
24
•
•
25
•
•
26
•
•
27
•
•
28
•
•
29
•
•
30
•
•
•
•
•
•
•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
•
•
•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
•
•
•
•
•
•
1
•
•
2
•
•
3
•
•
4
•
•
5
•
•
6
•
•
7
•
•
8
•
•
9
•
•
10
•
•
11
•
•
12
•
•
13
•
•
14
•
•
15
•
•
16
•
•
•
•
Gorgias Conversational Help Center: From search to resolution
Gorgias Conversational Help Center: From search to resolution
Gorgias Conversational Help Center: From search to resolution
Gorgias Conversational Help Center: From search to resolution
Gorgias empowers over 15,000+ ecommerce brands to centralize and automate customer support across chat, email, and social media. Yet, one channel remains unchanged: the Help Center. Shoppers today are still required to search, skim articles, and self-diagnose, often leaving without clear answers and forced to escalate.
In early 2025, I led design and research for a vision and pilot that layers a conversational AI experience on top of the existing Help Center. Instead of browsing through articles, shoppers can ask a question in natural language and receive an accurate, contextual answer, powered by our AI Agent, ultimately guiding them all the way to resolution.
Gorgias empowers over 15,000+ ecommerce brands to centralize and automate customer support across chat, email, and social media. Yet, one channel remains unchanged: the Help Center. Shoppers today are still required to search, skim articles, and self-diagnose, often leaving without clear answers and forced to escalate.
In early 2025, I led design and research for a vision and pilot that layers a conversational AI experience on top of the existing Help Center. Instead of browsing through articles, shoppers can ask a question in natural language and receive an accurate, contextual answer, powered by our AI Agent, ultimately guiding them all the way to resolution.
Gorgias empowers over 15,000+ ecommerce brands to centralize and automate customer support across chat, email, and social media. Yet, one channel remains unchanged: the Help Center. Shoppers today are still required to search, skim articles, and self-diagnose, often leaving without clear answers and forced to escalate.
In early 2025, I led design and research for a vision and pilot that layers a conversational AI experience on top of the existing Help Center. Instead of browsing through articles, shoppers can ask a question in natural language and receive an accurate, contextual answer, powered by our AI Agent, ultimately guiding them all the way to resolution.
Impact
Impact
Advanced company strategy: Project was a cornerstone of a Shopper Experience Vision I worked on with a senior designer in tandem, reshaping the end-to-end shopper journey that prioritizes conversation at every touchpoint
Validated approach to adoption: Pilot confirmed that conversational UI significantly reduced effort and increased resolution rates, while revealing that a small subset of shoppers still leveraged manual navigation.
Built scalable infrastructure: The patterns established here became the blueprint for Gorgias’s evolving AI UI platform
Advanced company strategy: Project was a cornerstone of a Shopper Experience Vision I worked on with a senior designer in tandem, reshaping the end-to-end shopper journey that prioritizes conversation at every touchpoint
Validated approach to adoption: Pilot confirmed that conversational UI significantly reduced effort and increased resolution rates, while revealing that a small subset of shoppers still leveraged manual navigation.
Built scalable infrastructure: The patterns established here became the blueprint for Gorgias’s evolving AI UI platform
Advanced company strategy: Project was a cornerstone of a Shopper Experience Vision I worked on with a senior designer in tandem, reshaping the end-to-end shopper journey that prioritizes conversation at every touchpoint
Validated approach to adoption: Pilot confirmed that conversational UI significantly reduced effort and increased resolution rates, while revealing that a small subset of shoppers still leveraged manual navigation.
Built scalable infrastructure: The patterns established here became the blueprint for Gorgias’s evolving AI UI platform
Role
Role
Role
Product Design & Research, Gorgias (B2B SaaS)
Product Design & Research, Gorgias (B2B SaaS)
Product Design & Research, Gorgias (B2B SaaS)
Product Design & Research, Gorgias (B2B SaaS)
Duration
Duration
4 months, early 2025
4 months, early 2025
4 months, early 2025
Collaborators
Collaborators
Vina R., Product Manager
Natasha P., Senior Product Designer
3 Engineers
Vina R., Product Manager
Natasha P., Senior Product Designer
3 Engineers
Vina R., Product Manager
Natasha P., Senior Product Designer
3 Engineers
Impact
Advanced company strategy: Project was a cornerstone of a Shopper Experience Vision I worked on with a senior designer in tandem, reshaping the end-to-end shopper journey that prioritizes conversation at every touchpoint
Validated approach to adoption: Pilot confirmed that conversational UI significantly reduced effort and increased resolution rates, while revealing that a small subset of shoppers still leveraged manual navigation.
Built scalable infrastructure: The patterns established here became the blueprint for Gorgias’s evolving AI UI platform
Self-serve is no longer enough
Self-serve is no longer enough
Help Centers have traditionally been a brand’s trust anchor — the primary place where shoppers try to answer product questions and resolve issues on their own. In fact, 81% of shoppers prefer self-service before contacting support.
But across industries, Help Centers haven’t evolved.
They remain dense and text-heavy, structured around articles, not outcomes, and frustrating to navigate.
Meanwhile, shoppers’ expectations are rapidly changing.
As AI tools become part of daily life, shoppers increasingly expect support that mirrors these experiences: instant, conversational, and context-aware. Traditional Help Centers no longer fit the expectation of “just ask and get an answer.”
Opportunity
This shift in behavior creates a strategic opportunity for Gorgias to:
expand automation in a way that drives revenue and support efficiency, while modernizing the Help Center for an AI-first world that matches evolving shopper expectations.
To unlock this opportunity, we first needed to understand what was preventing shoppers from resolving their own issues.
Help Centers have traditionally been a brand’s trust anchor — the primary place where shoppers try to answer product questions and resolve issues on their own. In fact, 81% of shoppers prefer self-service before contacting support.
But across industries, Help Centers haven’t evolved.
They remain dense and text-heavy, structured around articles, not outcomes, and frustrating to navigate.
Meanwhile, shoppers’ expectations are rapidly changing.
As AI tools become part of daily life, shoppers increasingly expect support that mirrors these experiences: instant, conversational, and context-aware. Traditional Help Centers no longer fit the expectation of “just ask and get an answer.”
Opportunity
This shift in behavior creates a strategic opportunity for Gorgias to:
expand automation in a way that drives revenue and support efficiency, while modernizing the Help Center for an AI-first world that matches evolving shopper expectations.
To unlock this opportunity, we first needed to understand what was preventing shoppers from resolving their own issues.
Help Centers have traditionally been a brand’s trust anchor — the primary place where shoppers try to answer product questions and resolve issues on their own. In fact, 81% of shoppers prefer self-service before contacting support.
But across industries, Help Centers haven’t evolved.
They remain dense and text-heavy, structured around articles, not outcomes, and frustrating to navigate.
Meanwhile, shoppers’ expectations are rapidly changing.
As AI tools become part of daily life, shoppers increasingly expect support that mirrors these experiences: instant, conversational, and context-aware. Traditional Help Centers no longer fit the expectation of “just ask and get an answer.”
Opportunity
This shift in behavior creates a strategic opportunity for Gorgias to:
expand automation in a way that drives revenue and support efficiency, while modernizing the Help Center for an AI-first world that matches evolving shopper expectations.
To unlock this opportunity, we first needed to understand what was preventing shoppers from resolving their own issues.
The problem
An information maze
When shoppers land on the Help Center, they are met with competing entry points: search bars, categories, and contact links all fighting for attention. The interface reflects the company’s internal structure, not the shopper’s mental model. And because it was originally designed for desktop, it becomes even more overwhelming when compressed onto mobile.
In fact, more than 60% of shoppers immediately go to the search bar, making it the primary starting point for most visitors
Despite its usability challenges, the Help Center is one of our most valuable automation channels, thanks to Flows (branched workflows for top questions) and Order Management features. Shoppers clearly prefer self-service when it delivers quick, relevant answers.
When shoppers land on the Help Center, they are met with competing entry points: search bars, categories, and contact links all fighting for attention. The interface reflects the company’s internal structure, not the shopper’s mental model. And because it was originally designed for desktop, it becomes even more overwhelming when compressed onto mobile.
In fact, more than 60% of shoppers immediately go to the search bar, making it the primary starting point for most visitors
Despite its usability challenges, the Help Center is one of our most valuable automation channels, thanks to Flows (branched workflows for top questions) and Order Management features. Shoppers clearly prefer self-service when it delivers quick, relevant answers.
When shoppers land on the Help Center, they are met with competing entry points: search bars, categories, and contact links all fighting for attention. The interface reflects the company’s internal structure, not the shopper’s mental model. And because it was originally designed for desktop, it becomes even more overwhelming when compressed onto mobile.
In fact, more than 60% of shoppers immediately go to the search bar, making it the primary starting point for most visitors
Despite its usability challenges, the Help Center is one of our most valuable automation channels, thanks to Flows (branched workflows for top questions) and Order Management features. Shoppers clearly prefer self-service when it delivers quick, relevant answers.
High reliance on shopper effort
Despite being the primary entry point for most shoppers, search often fails to deliver relevant answers. Common reasons include: Difficulty understanding natural language and varied phrasings, lack of personalization, weak semantic matching, and incomplete, outdated, or poorly tagged content, leaving gaps in coverage.
These challenges force shoppers to repeat searches, skim irrelevant content, and expend unnecessary effort, contributing to frustration and lower resolution rates.
A 90-day sample of an enterprise customer revealed that 52% of 36,674 searches returned zero results, a clear signal that self-service is failing at scale.
Despite being the primary entry point for most shoppers, search often fails to deliver relevant answers. Common reasons include: Difficulty understanding natural language and varied phrasings, lack of personalization, weak semantic matching, and incomplete, outdated, or poorly tagged content, leaving gaps in coverage.
These challenges force shoppers to repeat searches, skim irrelevant content, and expend unnecessary effort, contributing to frustration and lower resolution rates.
A 90-day sample of an enterprise customer revealed that 52% of 36,674 searches returned zero results, a clear signal that self-service is failing at scale.
Despite being the primary entry point for most shoppers, search often fails to deliver relevant answers. Common reasons include: Difficulty understanding natural language and varied phrasings, lack of personalization, weak semantic matching, and incomplete, outdated, or poorly tagged content, leaving gaps in coverage.
These challenges force shoppers to repeat searches, skim irrelevant content, and expend unnecessary effort, contributing to frustration and lower resolution rates.
A 90-day sample of an enterprise customer revealed that 52% of 36,674 searches returned zero results, a clear signal that self-service is failing at scale.
Dead-ends and blind escalation
Even when shoppers find articles, they often don’t fully resolve their issue. Help Centers are intentionally generic, designed to serve a broad audience, which means unique problems frequently go unaddressed.
This generic nature also shapes perception: shoppers may view the Help Center as impersonal, unhelpful, or a barrier rather than a trusted resource. There is no guidance for related suggestions, and no smooth path forward or alternative solutions.
When shoppers eventually contact support, all context from prior searches is lost, forcing users to repeat information and slowing resolution.
Even when shoppers find articles, they often don’t fully resolve their issue. Help Centers are intentionally generic, designed to serve a broad audience, which means unique problems frequently go unaddressed.
This generic nature also shapes perception: shoppers may view the Help Center as impersonal, unhelpful, or a barrier rather than a trusted resource. There is no guidance for related suggestions, and no smooth path forward or alternative solutions.
When shoppers eventually contact support, all context from prior searches is lost, forcing users to repeat information and slowing resolution.
Even when shoppers find articles, they often don’t fully resolve their issue. Help Centers are intentionally generic, designed to serve a broad audience, which means unique problems frequently go unaddressed.
This generic nature also shapes perception: shoppers may view the Help Center as impersonal, unhelpful, or a barrier rather than a trusted resource. There is no guidance for related suggestions, and no smooth path forward or alternative solutions.
When shoppers eventually contact support, all context from prior searches is lost, forcing users to repeat information and slowing resolution.
From hypotheses to a Vision-first approach
To explore faster, more effective self-serve support, we initially used a Thoughtful Execution Tree to generate hypotheses from our research for an MVP. Ideas included:
Embedding the Help Center into Chat
Adding AI features such as Summaries, Smart Feedback, Nudges, or an AI Mode in the search bar
For each, we defined abstract impact metrics to measure estimated effects on shoppers, the business, and merchants based on their core goals (automation, time to resolution). While these concepts could help shoppers find answers faster, they felt like patches on the existing model rather than proactive, future-facing solutions.
To explore faster, more effective self-serve support, we initially used a Thoughtful Execution Tree to generate hypotheses from our research for an MVP. Ideas included:
Embedding the Help Center into Chat
Adding AI features such as Summaries, Smart Feedback, Nudges, or an AI Mode in the search bar
For each, we defined abstract impact metrics to measure estimated effects on shoppers, the business, and merchants based on their core goals (automation, time to resolution). While these concepts could help shoppers find answers faster, they felt like patches on the existing model rather than proactive, future-facing solutions.
To explore faster, more effective self-serve support, we initially used a Thoughtful Execution Tree to generate hypotheses from our research for an MVP. Ideas included:
Embedding the Help Center into Chat
Adding AI features such as Summaries, Smart Feedback, Nudges, or an AI Mode in the search bar
For each, we defined abstract impact metrics to measure estimated effects on shoppers, the business, and merchants based on their core goals (automation, time to resolution). While these concepts could help shoppers find answers faster, they felt like patches on the existing model rather than proactive, future-facing solutions.
From hypotheses to a Vision-first approach
Leadership push and vision alignment
To explore faster, more effective self-serve support, we initially used a Thoughtful Execution Tree to generate hypotheses from our research for an MVP. Ideas included:
Embedding the Help Center into Chat
Adding AI features such as Summaries, Smart Feedback, Nudges, or an AI Mode in the search bar
For each, we defined abstract impact metrics to measure estimated effects on shoppers, the business, and merchants based on their core goals (automation, time to resolution). While these concepts could help shoppers find answers faster, they felt like patches on the existing model rather than proactive, future-facing solutions.
During ideation, new leadership questioned how the project would fit into the end-to-end shopping experience. The Help Center was just one touchpoint — shoppers also interact with browsing and purchasing flows. We needed to understand how this project could influence the entire journey.
I also recognized that current AI Agent capabilities and UI patterns limited our ability to deliver a solution that would meet shopper expectations and differentiate us from competitors.
We used vision as an alignment tool, framing discussions around a long-term shopper experience to guide pilot scope and product-market fit, while demonstrating the impact designers can have on strategy.
Defining the end-to-end shopper vision
Working closely with a senior designer in the AI tribe, we aligned on the shopper journey stages before and after the Help Center — from discovery to purchase.
From this work, we defined a future-facing vision: a mobile-first, AI-driven platform that supports the entire shopping journey within a single chat window, allowing shoppers to browse, get answers, and complete purchases seamlessly.
Once the vision was clear, we scaled the project back to focus solely on the Help Center, ensuring all stakeholders were aligned on the core direction. The Help Center became one touchpoint in this larger, cohesive shopper journey, informed by the long-term vision.
Leadership push and vision alignment
During ideation, new leadership questioned how the project would fit into the end-to-end shopping experience. The Help Center was just one touchpoint — shoppers also interact with browsing and purchasing flows. We needed to understand how this project could influence the entire journey.
I also recognized that current AI Agent capabilities and UI patterns limited our ability to deliver a solution that would meet shopper expectations and differentiate us from competitors.
We used vision as an alignment tool, framing discussions around a long-term shopper experience to guide pilot scope and product-market fit, while demonstrating the impact designers can have on strategy.
Defining the end-to-end shopper vision
Working closely with a senior designer in the AI tribe, we aligned on the shopper journey stages before and after the Help Center — from discovery to purchase.
From this work, we defined a future-facing vision: a mobile-first, AI-driven platform that supports the entire shopping journey within a single chat window, allowing shoppers to browse, get answers, and complete purchases seamlessly.
Once the vision was clear, we scaled the project back to focus solely on the Help Center, ensuring all stakeholders were aligned on the core direction. The Help Center became one touchpoint in this larger, cohesive shopper journey, informed by the long-term vision.
During ideation, new leadership questioned how the project would fit into the end-to-end shopping experience. The Help Center was just one touchpoint — shoppers also interact with browsing and purchasing flows. We needed to understand how this project could influence the entire journey.
I also recognized that current AI Agent capabilities and UI patterns limited our ability to deliver a solution that would meet shopper expectations and differentiate us from competitors.
We used vision as an alignment tool, framing discussions around a long-term shopper experience to guide pilot scope and product-market fit, while demonstrating the impact designers can have on strategy.
Defining the end-to-end shopper vision
Working closely with a senior designer in the AI tribe, we aligned on the shopper journey stages before and after the Help Center — from discovery to purchase.
From this work, we defined a future-facing vision: a mobile-first, AI-driven platform that supports the entire shopping journey within a single chat window, allowing shoppers to browse, get answers, and complete purchases seamlessly.
Once the vision was clear, we scaled the project back to focus solely on the Help Center, ensuring all stakeholders were aligned on the core direction. The Help Center became one touchpoint in this larger, cohesive shopper journey, informed by the long-term vision.
The Vision: Redefining the Help Center experience
The future Help Center is no longer just a repository of answers. By becoming a conversational companion, it engages shoppers, anticipates their needs, and delivers instant, personalized support, while seamlessly escalating to human agents when necessary. This shift transforms self-service from passive search into interactive guidance, boosting trust, satisfaction, and automation.
The future Help Center is no longer just a repository of answers. By becoming a conversational companion, it engages shoppers, anticipates their needs, and delivers instant, personalized support, while seamlessly escalating to human agents when necessary. This shift transforms self-service from passive search into interactive guidance, boosting trust, satisfaction, and automation.
The future Help Center is no longer just a repository of answers. By becoming a conversational companion, it engages shoppers, anticipates their needs, and delivers instant, personalized support, while seamlessly escalating to human agents when necessary. This shift transforms self-service from passive search into interactive guidance, boosting trust, satisfaction, and automation.
Turning queries into dialogue
Most shoppers head straight to the search bar, guessing keywords and clicking through articles until something helps. Now, imagine a simple prompt in a familiar interface: “How can I help you?” Keyword guesswork becomes a real conversation with AI Agent, which understands intent, anticipates needs, and surfaces confidence levels so users know when information is verified versus inferred.
Most shoppers head straight to the search bar, guessing keywords and clicking through articles until something helps. Now, imagine a simple prompt in a familiar interface: “How can I help you?” Keyword guesswork becomes a real conversation with AI Agent, which understands intent, anticipates needs, and surfaces confidence levels so users know when information is verified versus inferred.
Most shoppers head straight to the search bar, guessing keywords and clicking through articles until something helps. Now, imagine a simple prompt in a familiar interface: “How can I help you?” Keyword guesswork becomes a real conversation with AI Agent, which understands intent, anticipates needs, and surfaces confidence levels so users know when information is verified versus inferred.
Guided, end-to-end resolution
Article reading becomes direct resolution without speaking to a person. AI Agent can confirm a damaged item, analyze the issue, and process a replacement, delivering end-to-end resolution in a single conversation.
By drawing on Help Center articles, workflows, real-time order data, and internal policies, AI Agent reduces handoffs, partial fixes, and frustration. When it cannot fully resolve an issue, it offers alternatives or next steps, maintaining user trust.
Article reading becomes direct resolution without speaking to a person. AI Agent can confirm a damaged item, analyze the issue, and process a replacement, delivering end-to-end resolution in a single conversation.
By drawing on Help Center articles, workflows, real-time order data, and internal policies, AI Agent reduces handoffs, partial fixes, and frustration. When it cannot fully resolve an issue, it offers alternatives or next steps, maintaining user trust.
Article reading becomes direct resolution without speaking to a person. AI Agent can confirm a damaged item, analyze the issue, and process a replacement, delivering end-to-end resolution in a single conversation.
By drawing on Help Center articles, workflows, real-time order data, and internal policies, AI Agent reduces handoffs, partial fixes, and frustration. When it cannot fully resolve an issue, it offers alternatives or next steps, maintaining user trust.
Building trust through transparency
AI Agent is transparent about sources and clearly identifies itself, so shoppers understand they are interacting with AI. Smart follow-ups keep the conversation moving, making it easy for users to respond without extra effort. Feedback loops let users provide input, helping the AI improve over time while merchants can monitor and retrain it based on gaps or patterns.
AI Agent is transparent about sources and clearly identifies itself, so shoppers understand they are interacting with AI. Smart follow-ups keep the conversation moving, making it easy for users to respond without extra effort. Feedback loops let users provide input, helping the AI improve over time while merchants can monitor and retrain it based on gaps or patterns.
AI Agent is transparent about sources and clearly identifies itself, so shoppers understand they are interacting with AI. Smart follow-ups keep the conversation moving, making it easy for users to respond without extra effort. Feedback loops let users provide input, helping the AI improve over time while merchants can monitor and retrain it based on gaps or patterns.
Seamless human handoff
As a last resort, a human agent can seamlessly step into the conversation within the same interface, already briefed, so shoppers never have to repeat themselves. Recovery paths are smooth and intuitive, keeping experiences frictionless while ensuring trust and reliability.
As a last resort, a human agent can seamlessly step into the conversation within the same interface, already briefed, so shoppers never have to repeat themselves. Recovery paths are smooth and intuitive, keeping experiences frictionless while ensuring trust and reliability.
The engine powering AI Agent
Under the hood, AI Agent runs on our existing platform, enhanced with layered intelligence and deep integrations that make every interaction smarter and more connected. This foundation powers today’s experience in chat and email, while positioning the product to evolve toward a unified, intelligent, end-to-end shopper experience.
To reduce hallucinations and build trust in AI responses, AI Agent leverages a Retrieval-Augmented Generation (RAG) layer, drawing from:
Help Center guidance and articles
Real-time Shopify order data
Internal policies and workflows
External documentation as needed
By combining these sources, AI Agent delivers contextual, reliable answers, while remaining flexible to expand into broader experiences across the Gorgias ecosystem.
Under the hood, AI Agent runs on our existing platform, enhanced with layered intelligence and deep integrations that make every interaction smarter and more connected. This foundation powers today’s experience in chat and email, while positioning the product to evolve toward a unified, intelligent, end-to-end shopper experience.
To reduce hallucinations and build trust in AI responses, AI Agent leverages a Retrieval-Augmented Generation (RAG) layer, drawing from:
Help Center guidance and articles
Real-time Shopify order data
Internal policies and workflows
External documentation as needed
By combining these sources, AI Agent delivers contextual, reliable answers, while remaining flexible to expand into broader experiences across the Gorgias ecosystem.
The Reality: Piloting a lean proof of concept
Building the full vision would have required months of engineering to restructure the backend infrastructure and introduce new UI components, which wasn’t feasible. Instead, we created a high-control, low-agency pilot: small, testable, and easy to observe, using the existing Chat infrastructure, which already has AI Agent integrated.
This allowed us to measure risk, validate the desirability of a conversational-first experience, and learn quickly without over-investing, especially since no competitor offered a similar experience at the time.
Building the full vision would have required months of engineering to restructure the backend infrastructure and introduce new UI components, which wasn’t feasible. Instead, we created a high-control, low-agency pilot: small, testable, and easy to observe, using the existing Chat infrastructure, which already has AI Agent integrated.
This allowed us to measure risk, validate the desirability of a conversational-first experience, and learn quickly without over-investing, especially since no competitor offered a similar experience at the time.
Approach
We anchored the pilot using our existing AI-powered Chat, simulating conversational support while maintaining control over AI Agent responses. This created a realistic testing environment without the cost of building a new platform. We dog-fooded participants’ existing Chats with their Help Center’s top-asked questions, ensuring it could handle most queries.
The pilot included:
• A/B test: Ran for 1 month, comparing the existing Help Center to the conversational version. Measured shopper engagement, resolution rates, and drop-off behavior
• UI tweaks: Reused the Chat interface while toning down the “chat” look so shoppers perceived it as self-serve rather than live support
• Fallback browsing: Instead of hiding articles completely, we tucked them behind a simple text button, allowing us to observe whether shoppers preferred the old browsing path or used it as a backup to AI
• Fixed prompt suggestions: Merchants defined a short set of common starter questions. While not dynamically generated, this allowed us to observe baseline engagement
Approach
We anchored the pilot using our existing AI-powered Chat, simulating conversational support while maintaining control over AI Agent responses. This created a realistic testing environment without the cost of building a new platform. We dog-fooded participants’ existing Chats with their Help Center’s top-asked questions, ensuring it could handle most queries.
The pilot included:
• A/B test: Ran for 1 month, comparing the existing Help Center to the conversational version. Measured shopper engagement, resolution rates, and drop-off behavior
• UI tweaks: Reused the Chat interface while toning down the “chat” look so shoppers perceived it as self-serve rather than live support
• Fallback browsing: Instead of hiding articles completely, we tucked them behind a simple text button, allowing us to observe whether shoppers preferred the old browsing path or used it as a backup to AI
• Fixed prompt suggestions: Merchants defined a short set of common starter questions. While not dynamically generated, this allowed us to observe baseline engagement
Approach
We anchored the pilot using our existing AI-powered Chat, simulating conversational support while maintaining control over AI Agent responses. This created a realistic testing environment without the cost of building a new platform. We dog-fooded participants’ existing Chats with their Help Center’s top-asked questions, ensuring it could handle most queries.
The pilot included:
• A/B test: Ran for 1 month, comparing the existing Help Center to the conversational version. Measured shopper engagement, resolution rates, and drop-off behavior
• UI tweaks: Reused the Chat interface while toning down the “chat” look so shoppers perceived it as self-serve rather than live support
• Fallback browsing: Instead of hiding articles completely, we tucked them behind a simple text button, allowing us to observe whether shoppers preferred the old browsing path or used it as a backup to AI
• Fixed prompt suggestions: Merchants defined a short set of common starter questions. While not dynamically generated, this allowed us to observe baseline engagement
Learnings & Impact
As a result of the pilot, an MVP was planned for release in H2. While I left Gorgias shortly after, the new infrastructure and UI were designed to support conversational AI experiences across the product, backed by new design system components tailored for this interaction pattern.
Impact of vision work
Long-Term Thinking: Establishing a future-facing vision demonstrated to stakeholders that design was contributing strategically, not just incrementally.
Alignment Tool: By framing discussions around the end-to-end shopper experience, we helped leadership make informed decisions about pilot scope and product-market fit. This also highlighted opportunities for design to lead strategy, which hadn’t been formally emphasized before due to time constraints.
Changes in shopper behavior
Early testing revealed meaningful shifts in behavior:
Shoppers began entering longer, natural-language questions instead of short keyword searches
This led to fewer drop-offs and higher resolution rates
Merchants acknowledged potential cost implications but were supportive of optional automated interactions in their Help Centers
As a result of the pilot, an MVP was planned for release in H2. While I left Gorgias shortly after, the new infrastructure and UI were designed to support conversational AI experiences across the product, backed by new design system components tailored for this interaction pattern.
Impact of vision work
Long-Term Thinking: Establishing a future-facing vision demonstrated to stakeholders that design was contributing strategically, not just incrementally.
Alignment Tool: By framing discussions around the end-to-end shopper experience, we helped leadership make informed decisions about pilot scope and product-market fit. This also highlighted opportunities for design to lead strategy, which hadn’t been formally emphasized before due to time constraints.
Changes in shopper behavior
Early testing revealed meaningful shifts in behavior:
Shoppers began entering longer, natural-language questions instead of short keyword searches
This led to fewer drop-offs and higher resolution rates
Merchants acknowledged potential cost implications but were supportive of optional automated interactions in their Help Centers
As a result of the pilot, an MVP was planned for release in H2. While I left Gorgias shortly after, the new infrastructure and UI were designed to support conversational AI experiences across the product, backed by new design system components tailored for this interaction pattern.
Impact of vision work
Long-Term Thinking: Establishing a future-facing vision demonstrated to stakeholders that design was contributing strategically, not just incrementally.
Alignment Tool: By framing discussions around the end-to-end shopper experience, we helped leadership make informed decisions about pilot scope and product-market fit. This also highlighted opportunities for design to lead strategy, which hadn’t been formally emphasized before due to time constraints.
Changes in shopper behavior
Early testing revealed meaningful shifts in behavior:
Shoppers began entering longer, natural-language questions instead of short keyword searches
This led to fewer drop-offs and higher resolution rates
Merchants acknowledged potential cost implications but were supportive of optional automated interactions in their Help Centers
How we used AI-powered prompts to double chat engagement and drive a 22% increase in GMV.
How we used AI-powered prompts to double chat engagement and drive a 22% increase in GMV.
How we used AI-powered prompts to double chat engagement and drive a 22% increase in GMV.
The engine powering AI Agent
Under the hood, AI Agent runs on our existing platform, enhanced with layered intelligence and deep integrations that make every interaction smarter and more connected. This foundation powers today’s experience in chat and email, while positioning the product to evolve toward a unified, intelligent, end-to-end shopper experience.
To reduce hallucinations and build trust in AI responses, AI Agent leverages a Retrieval-Augmented Generation (RAG) layer, drawing from:
Help Center guidance and articles
Real-time Shopify order data
Internal policies and workflows
External documentation as needed
By combining these sources, AI Agent delivers contextual, reliable answers, while remaining flexible to expand into broader experiences across the Gorgias ecosystem.