When Rhoback introduced an AI Agent to its customer experience team, it did more than automate routine tickets. Implementation revealed an opportunity to improve documentation, collaborate cross-functionally, and establish a clear brand tone of voice. 

Samantha Gagliardi, Associate Director of Customer Experience at Rhoback, explains the entire process in the first episode of our AI in CX webinar series.

Key takeaways:

• Implement quickly and iterate. Rhoback’s initial rollout process took two weeks, right before BFCM. Samantha moved quickly, starting with basic FAQs and then continuously optimizing.  
• Train AI like a three-year-old. Although it is empathetic, an AI Agent does not inherently know what is right or wrong. Invest in writing clear Guidance, testing responses, and ensuring document accuracy. 
• Approach your AI’s tone of voice like a character study. Your AI Agent is an extension of your brand, and its personality should reflect that. Rhoback conducted a complete analysis of its agent’s tone, age, energy, and vocabulary. 
• Embrace AI as a tool to reveal inconsistencies. If your AI Agent is giving inaccurate information, it’s exposing gaps in your knowledge sources. Uses these early test responses to audit product pages, help center content, Guidance, and policies.
• Check in regularly and keep humans in control. Introduce weekly reviews or QA rituals to refine AI’s accuracy, tone, and efficiency. Communicate AI insights cross-functionally to build trust and work towards shared goals.

Top learnings from Rhoback’s AI rollout  

1. You can start before you “feel ready”

With any new tool, the pre-implementation phase can take some time. Creating proper documentation, training internal teams, and integrating with your tech stack are all important steps that happen before you go live. 

But sometimes it’s okay just to launch a tool and optimize as you go. 

Rhoback launched its AI agent two weeks before BFCM to automate routine tickets during the busy season. 

Why it worked:

• Samantha had audited all of Rhoback’s SOPs, training materials, and FAQs a few months before implementation. 
• They started by automating high-volume questions such as returns, exchanges, and order tracking.
• They followed a structured AI implementation checklist. 

2. Audit your knowledge sources before you automate

Before turning on Rhoback’s AI Agent, Samantha’s team reviewed every FAQ, policy, and help article that human agents are trained on. This helped establish clear CX expectations that they could program into an AI Agent. 

Samantha also reviewed the most frequently asked questions and the ideal responses to each. Which ones needed an empathetic human touch and which ones required fast, accurate information?  

“AI tells you immediately when your data isn’t clean. If a product detail page says one thing and the help center says another, it shows up right away.” 

Rhoback’s pre-implementation audit checklist:

• Review customer FAQs and the appropriate responses for each. 
• Update outdated PDPs, Help Centre articles, policies, and other relevant documentation.
• Establish workflows with Ecommerce and Product teams to align Macros, Guidance, and Help Center articles with product descriptions and website copy. 

Read more: How to Optimize Your Help Center for AI Agent

3. Train your AI Agent in small, clear steps

It’s often said that you should train your AI Agent like a brand-new employee. 

Samantha took it one step further and recommended treating AI like a toddler, with clear, patient, repetitive instructions. 

“The AI does not have a sense of good and bad. It’s going to say whatever you train it, so you need to break it down like you’re talking to a three-year-old that doesn’t know any different. Your directions should be so detailed that there is no room for error.”

Practical tips:

• Use AI to build your AI Guidance, focusing on clear, detailed, simple instructions. 
• Test each Guidance before adding new ones.
• Treat the training process like an ongoing feedback loop, not a one-time upload.

Read more: How to Write Guidance with the “When, If, Then” Framework

4. Prioritize Tone of Voice to make AI feel natural

For Rhoback, an on-brand Tone of Voice was a non-negotiable. Samantha built a character study that shaped Rhoback’s AI Agent’s custom brand voice.

“I built out the character of Rhoback, how it talks, what age it feels like, what its personality is. If it does not sound like us, it is not worth implementing.”

Key questions to shape your AI Agent’s tone of voice:

• How does the AI Agent speak? Friendly, funny, empathetic, etc…?
• Does your AI Agent use emojis? How often?
• Are there any terms or phrases the AI Agent should always or never say?

5. Use AI to surface knowledge gaps or inconsistencies

Once Samantha started testing the AI Agent, it quickly revealed misalignment between Rhoback’s teams. With such an extensive product catalog, AI showed that product details did not always match the Help Center or CX documentation. 

This made a case for stronger collaboration amongst the CX, Product, and Ecommerce teams to work towards their shared goal of prioritizing the customer. 

“It opened up conversations we were not having before. We all want the customer to be happy, from the moment they click on an ad to the moment they purchase to the moment they receive their order. AI Agent allowed us to see the areas we need to improve upon.” 

Tips to improve internal alignment:

• Create regular syncs between CX, Product, Ecommerce, and Marketing teams.
• Share AI summaries, QA insights, and trends to highlight recurring customer pain points.
• Build a collaborative workflow for updating documents that gives each team visibility. 

6. Build trust (with your team and customers) through transparency 

Despite the benefits of AI for CX, there’s still trepidation. Agents are concerned that AI would replace them, while customers worry they won’t be able to reach a human. Both are valid concerns, but clearly communicating internally and externally can mitigate skepticism. 

At Rhoback, Samantha built internal trust by looping in key stakeholders throughout the testing process. “I showed my team that it is not replacing them. It’s meant to be a support that helps them be even more successful with what they’re already doing," Samantha explains.

On the customer side, Samantha trained their AI Agent to tell customers in the first message that it is an AI customer service assistant that will try to help them or pass them along to a human if it can’t. 

How Rhoback built AI confidence:

• Positioned AI as a personal assistant for agents, not a replacement.
• Let agents, other departments, and leadership test and shape the AI Agent experience early.
• Told customers up front when automation was being used and made the path to a human clear and easy.

Read more: How CX Leaders are Actually Using AI: 6 Must-Know Lessons

Putting these into practice: Rhoback’s framework for an aligned AI implementation 

Here is Rhoback’s approach distilled into a simple framework you can apply.

1. Audit your content: Ensure your FAQs, product data, policies, and all documentation are accurate.
2. Start small: Automate one repetitive workflow, such as returns or tracking.
3. Train iteratively: Add Guidance in small, testable batches.
4. Prioritize tone: Make sure every AI reply sounds like your brand.
5. Align teams: Use AI data to resolve cross-departmental inconsistencies and establish clearer communication lines.
6. Be transparent: Tell both agents and customers how AI fits into the process.
7. Refine regularly: Review, measure, and adjust on an ongoing basis.
   

Watch the full conversation with Samantha to learn how AI can act as a catalyst for better internal alignment. 

📌 Join us for episode 2 of AI in CX: Building a Conversational Commerce Strategy that Converts with Cornbread Hemp on December 16.

‍{{lead-magnet-1}}

TL;DR:

• Most food & beverage support tickets during BFCM are predictable. Subscription cancellations, WISMO, and product questions make up the bulk—so prep answers ahead of time.
• Proactive CX site updates can drastically cut down repetitive tickets. Add ingredient lists, cooking instructions, and clear refund policies to product pages and FAQs.
• FAQ pages should go deep, not just broad. Answer hyper-specific questions like “Will this break my fast?” to help customers self-serve without hesitation.‍
• Transparency about stock reduces confusion and cart abandonment. Show inventory levels, set up waitlists, and clearly state cancellation windows.

In 2024, Shopify merchants drove $11.5 billion in sales over Black Friday Cyber Monday. Now, BFCM is quickly approaching, with some brands and major retailers already hosting sales.

If you’re feeling late to prepare for the season or want to maximize the number of sales you’ll make, we’ll cover how food and beverage CX teams can serve up better self-serve resources for this year’s BFCM. 

Learn how to answer and deflect customers’ top questions before they’re escalated to your support team.

💡 Your guide to everything peak season → The Gorgias BFCM Hub

Handling BFCM as a food & beverage brand

During busy seasons like BFCM and beyond, staying on top of routine customer asks can be an extreme challenge. 

“Every founder thinks BFCM is the highest peak feeling of nervousness,” says Ron Shah, CEO and Co-founder of supplement brand Obvi. 

“It’s a tough week. So anything that makes our team’s life easier instantly means we can focus more on things that need the time,” he continues. 

Anticipating contact reasons and preparing methods (like automated responses, macros, and enabling an AI Agent) is something that can help. Below, find the top contact reasons for food and beverage companies in 2025. 

Top contact reasons in the food & beverage industry 

According to Gorgias proprietary data, the top reason customers reach out to brands in the food and beverage industry is to cancel a subscription (13%) followed by order status questions (9.1%).

7 ways to improve your self-serve resources before BFCM

1. Add informative blurbs on product pages 
2. Craft additional help center and FAQ articles 
3. Automate responses with AI or Macros 
4. Get specific about product availability
5. Provide order cancellation and refund policies upfront
6. Add how-to information
7. Build resources to help with buying decisions 

1) Add informative blurbs on product pages

Because product detail queries represent 5.7% of contact reasons for the food and beverage industry, the more information you provide on your product pages, the better. 

Include things like calorie content, nutritional information, and all ingredients.  

For example, ready-to-heat meal company The Dinner Ladies includes a dropdown menu on each product page for further reading. Categories include serving instructions, a full ingredient list, allergens, nutritional information, and even a handy “size guide” that shows how many people the meal serves. 

2) Craft additional Help Center and FAQ articles

FAQ pages make up the information hub of your website. They exist to provide customers with a way to get their questions answered without reaching out to you.   

This includes information like how food should be stored, how long its shelf life is, delivery range, and serving instructions. FAQs can even direct customers toward finding out where their order is and what its status is. 

In the context of BFCM, FAQs are all about deflecting repetitive questions away from your team and assisting shoppers in finding what they need faster. 

That’s the strategy for German supplement brand mybacs. 

“Our focus is to improve automations to make it easier for customers to self-handle their requests. This goes hand in hand with making our FAQs more comprehensive to give customers all the information they need,” says Alexander Grassmann, its Co-Founder & COO.

As you contemplate what to add to your FAQ page, remember that more information is usually better. That’s the approach Everyday Dose takes, answering even hyper-specific questions like, “Will it break my fast?” or “Do I have to use milk?”

While the FAQs you choose to add will be specific to your products, peruse the top-notch food and bev FAQ pages below. 

Time for some FAQ inspo:

• Juice Plus+ FAQ
• Everyday Dose FAQ
• Chamberlain Coffee FAQ
• Jurassic Fruit FAQ
• Rebel Cheese FAQ
• The Dinner Ladies FAQ

3) Automate responses with AI or macros

AI Agents and AI-powered Shopping Assistants are easy to set up and are extremely effective in handling customer interactions––especially during BFCM.  

“I told our team we were going to onboard Gorgias AI Agent for BFCM, so a good portion of tickets would be handled automatically,” says Ron Shah, CEO and Co-founder at Obvi. “There was a huge sigh of relief knowing that customers were going to be taken care of.” 

And, they’re getting smarter. AI Agent’s CSAT is just 0.6 points shy of human agents’ average CSAT score. 

Here are the specific responses and use cases we recommend automating: 

• WISMO (where is my order) inquiries 
• Product related questions 
• Returns 
• Order issues
• Cancellations 
• Discounts, including BFCM related 
• Customer feedback
• Account management
• Collaboration requests 
• Rerouting complex queries

Get your checklist here: How to prep for peak season: BFCM automation checklist

4) Get specific about product availability

With high price reductions often comes faster-than-usual sell out times. By offering transparency around item quantities, you can avoid frustrated or upset customers. 

For example, you could show how many items are left under a certain threshold (e.g. “Only 10 items left”), or, like Rebel Cheese does, mention whether items have sold out in the past.  

You could also set up presales, give people the option to add themselves to a waitlist, and provide early access to VIP shoppers. 

5) Provide order cancellation and refund policies upfront 

Give shoppers a heads up whether they’ll be able to cancel an order once placed, and what your refund policies are. 

For example, cookware brand Misen follows its order confirmation email with a “change or cancel within one hour” email that provides a handy link to do so. 

Your refund policies and order cancellations should live within an FAQ and in the footer of your website. 

6) Add how-to information 

Include how-to information on your website within your FAQs, on your blog, or as a standalone webpage. That might be sharing how to use a product, how to cook with it, or how to prepare it. This can prevent customers from asking questions like, “how do you use this?” or “how do I cook this?” or “what can I use this with?” etc. 

For example, Purity Coffee created a full brewing guide with illustrations:

Similarly, for its unique preseasoned carbon steel pan, Misen lists out care instructions: 

And for those who want to understand the level of prep and cooking time involved, The Dinner Ladies feature cooking instructions on each product page. 

7) Build resources to help with buying decisions 

Interactive quizzes, buying guides, and gift guides can help ensure shoppers choose the right items for them––without contacting you first. 

For example, Trade Coffee Co created a quiz to help first timers find their perfect coffee match: 

Set your team up for BFCM success with Gorgias 

The more information you can share with customers upfront, the better. That will leave your team time to tackle the heady stuff. 

If you’re looking for an AI-assist this season, check out Gorgias’s suite of products like AI Agent and Shopping Assistant. 

{{lead-magnet-2}}

TL;DR:

• You don’t need to rebuild your Help Center to make it work with AI—you just need to structure it smarter.
• AI Agent reads your content in three layers: Help Center, Guidance, and Actions, following an “if / when / then” logic to find and share accurate answers.
• Most AI escalations happen because Help Docs are vague or incomplete. Start by improving your top 10 ticket topics—like order status, returns, and refunds.
• Make your articles scannable, define clear conditions, link next steps, and keep your tone consistent. These small tweaks help AI Agent resolve more tickets on its own—and free up your team to focus on what matters most.

As holiday season support volumes spike and teams lean on AI to keep up, one frustration keeps surfacing, our Help Center has the answers—so why can’t AI find them?

The truth is, AI can’t help customers if it can’t understand your Help Center. Most large language models (LLMs), including Gorgias AI Agent, don’t ignore your existing docs, they just struggle to find clear, structured answers inside them.

The good news is you don’t need to rebuild your Help Center or overhaul your content. You simply need to format it in a way that’s easy for both people and AI to read.

We’ll break down how AI Agent reads your Help Center, finds answers, and why small formatting changes can help it respond faster and more accurately, so your team spends less time on escalations.

{{lead-magnet-1}}

How AI Agent uses your Help Center content

Before you start rewriting your Help Center, it helps to understand how AI Agent actually reads and uses it.

Think of it like a three-step process that mirrors how a trained support rep thinks through a ticket.

1. Read Help Center docs

Your Help Center is AI Agent’s brain. AI Agent uses your Help Center to pull facts, policies, and instructions it needs to respond to customers accurately. If your articles are clearly structured and easy to scan, AI Agent can find what it needs fast. If not, it hesitates or escalates.

2. Follow Guidance instructions

Think of Guidance as AI Agent’s decision layer. What should AI Agent do when someone asks for a refund? What about when they ask for a discount? Guidance helps AI Agent provide accurate answers or hand over to a human by following an “if/when/then” framework.

3. Respond and perform

Finally, AI Agent uses a combination of your help docs and Guidance to respond to customers, and if enabled, perform an Action on their behalf—whether that’s changing a shipping address or canceling an order altogether.

Here’s what that looks like in practice:

This structure removes guesswork for both your AI and your customers. The clearer your docs are about when something applies and what happens next, the more accurate and human your automated responses will feel.

A Help Center written for both people and AI Agent:

• Saves your team time
• Reduces escalations
• Helps every customer get the right answer the first time

What causes AI Agent to escalate tickets, and how to fix it

Our data shows that most AI escalations happen for a simple reason––your Help Center doesn’t clearly answer the question your customer is asking.

That’s not a failure of AI. It’s a content issue. When articles are vague, outdated, or missing key details, AI Agent can’t confidently respond, so it passes the ticket to a human.

Here are the top 10 topics that trigger escalations most often:

Each of these topics needs a dedicated, clearly structured Help Doc that uses keywords customers are likely to search and spells out specific conditions. 

Here’s how to strengthen each one:

• Order status: Include expected delivery timelines, tracking link FAQs, and a clear section for “what to do if tracking isn’t updating.”
• Return request: Spell out eligibility requirements, time limits, and how to print or request a return label.
• Order cancellation: Define cut-off times for canceling and link to your “returns” doc for shipped orders.
• Product quality issues: Explain what qualifies as a defect, how to submit photos, and whether replacements or refunds apply.
• Missing item: Clarify how to report missing items and what verification steps your team takes before reshipping.
• Subscription cancellation: Add “if/then” logic for different cases: if paused vs. canceled, if prepaid vs. monthly.
• Order refund: Outline refund timelines, where customers can see status updates, and any exceptions (e.g., partial refunds).
• Product details: Cover sizing, materials, compatibility, or FAQs that drive most product-related questions.
• Return status: State how long returns take to process and where to check progress once a label is scanned.
• Order delivered but not received: Provide step-by-step guidance for checking with neighbors, filing claims, or requesting replacements.

Start by improving these 10 articles first. Together, they account for nearly half of all AI Agent escalations. The clearer your Help Center is on these topics, the fewer tickets your team will ever see, and the faster your AI will resolve the rest.

How to format your Help Center docs for LLMs

Once you know how AI Agent reads your content, the next step is formatting your help docs so it can easily understand and use them. 

The goal isn’t to rewrite everything, it’s to make your articles more structured, scannable, and logic-friendly. 

Here’s how.

1. Use structured, scannable sections

Both humans and large language models read hierarchically. If your article runs together in one long block of text, key answers get buried.

Break articles into clear sections and subheadings (H2s, H3s) for each scenario or condition. Use short paragraphs, bullets, and numbered lists to keep things readable.

Example:

How to Track Your Order

• Step 1: Find your tracking number in your confirmation email.
• Step 2: Click the tracking link to see your delivery status.
• Step 3: If tracking hasn’t updated in 3 days, contact support.

A structured layout helps both AI and shoppers find the right step faster, without confusion or escalation.

2. Write for “if/when/then” logic

AI Agent learns best when your Help Docs clearly define what happens under specific conditions. Think of it like writing directions for a flowchart.

Example:

• “If your order hasn’t arrived within 10 days, contact support for a replacement.”
• “If your order has shipped, you can find the tracking link in your order confirmation email.”

This logic helps AI know what to do and how to explain the answer clearly to the customer.

3. Clarify similar terms and synonyms

Customers don’t always use the same words you do, and neither do LLMs. If your docs treat “cancel,” “stop,” and “pause” as interchangeable, AI Agent might return the wrong answer.

Define each term clearly in your Help Center and add small keyword variations (“cancel subscription,” “end plan,” “pause delivery”) so the AI can recognize related requests.

4. Link to next steps

AI Agent follows links just like a human agent. If your doc ends abruptly, it can’t guide the customer any further.

Always finish articles with an explicit next step, like linking to:

• A form
• Another article
• A support action page

Example: “If your return meets our policy, request your return label here.”

That extra step keeps the conversation moving and prevents unnecessary escalations.

5. Keep tone consistent

AI tools prioritize structure and wording when learning from your Help Center—not emotional tone. 

Phrases like “Don’t worry!” or “We’ve got you!” add noise without clarity.

Instead, use simple, action-driven sentences that tell the customer exactly what to do:

• “Click here to request a refund.”
• “Fill out the warranty form to get a replacement.”

A consistent tone keeps your Help Center professional, helps AI deliver reliable responses, and creates a smoother experience for customers.

LLM-friendly Help Centers in action

You don’t need hundreds of articles or complex workflows to make your Help Center AI-ready. But you do need clarity, structure, and consistency. These Gorgias customers show how it’s done.

Little Words Project: Simple formatting that boosts instant answers

Little Words Project keeps things refreshingly straightforward. Their Help Center uses short paragraphs, descriptive headers, and tightly scoped articles that focus on a single intent, like returns, shipping, or product care. 

That makes it easy for AI Agent to scan the page, pull out the right facts, and return accurate answers on the first try.

Their tone stays friendly and on-brand, but the structure is what shines. Every article flows from question → answer → next step. It’s a minimalist approach, and it works. Both for customers and the AI reading alongside them.

Dr. Bronner’s: Making tools work for the team

Customer education is at the heart of Dr. Bronner’s mission. Their customers often ask detailed questions about product ingredients, packaging, and certifications. With Gorgias, Emily and her team were able to build a robust Help Center that helped to proactively give this information.

The Help Center doesn't just provide information. The integration of interactive Flows, Order Management, and a Contact Form automation allowed Dr. Bronner’s to handle routine inquiries—such as order statuses—quickly and efficiently. These kinds of interactive elements are all possible out-of-the-box, no IT support needed.

Read more: How Dr. Bronner's saved $100k/year by switching from Salesforce, then automated 50% of interactions with Gorgias 

Ekster: Building efficiency through automation and clarity

When Ekster switched to Gorgias, the team wanted to make their Help Center work smarter. By writing clear, structured articles for common questions like order tracking, returns, and product details, they gave both customers and AI Agent the information needed to resolve issues instantly.

"Our previous Help Center solution was the worst. I hated it. Then I saw Gorgias’s Help Center features, and how the Article Recommendations could answer shoppers’ questions instantly, and I loved it. I thought: this is just what we need." —Shauna Cleary, Head of Ecommerce at Ekster

The results followed fast. With well-organized Help Center content and automation built around it, Ekster was able to scale support without expanding the team.

“With all the automations we’ve set up in Gorgias, and because our team in Buenos Aires has ramped up, we didn’t have to rehire any extra agents.” —Shauna Cleary, Head of Ecommerce at Ekster

Learn more: How Ekster used automation to cover the workload of 4 agents 

Rowan: Clean structure that keeps customers (and AI) on track

Rowan’s Help Center is a great example of how clear structure can do the heavy lifting. Their FAQs are grouped into simple categories like piercing, shipping, returns, and aftercare, so readers and AI Agent can jump straight to the right topic without digging. 

For LLMs, that kind of consistency reduces guesswork. For customers, it creates a smooth, reassuring self-service experience. 

TUSHY: Balancing brand voice with automation

TUSHY proves you can maintain personality and structure. Their Help Center articles use clear headings, direct language, and brand-consistent tone. It makes it easy for AI Agent to give accurate, on-brand responses.

“Too often, a great interaction is diminished when a customer feels reduced to just another transaction. With AI, we let the tech handle the selling, unabashedly, if needed, so our future customers can ask anything, even the questions they might be too shy to bring up with a human. In the end, everybody wins!" —Ren Fuller-Wasserman, Senior Director of Customer Experience at TUSHY

Quick checklist to audit your Help Center for AI

Ready to put your Help Center to the test? Use this five-point checklist to make sure your content is easy for both customers and AI to navigate.

1. Are your articles scannable with clear headings?

Break up long text blocks and use descriptive headers (H2s, H3s) so readers and AI Agent can instantly find the right section.

2. Do you define conditions with “if/when/then” phrasing?

Spell out what happens in each scenario. This logic helps AI Agent decide the right next step without second-guessing.

3. Do you cover your top escalation topics?

Make sure your Help Center includes complete, structured articles for high-volume issues like order status, returns, and refunds.

4. Does each article end with a clear next step or link?

Close every piece with a call to action, like a form, related article, or support link, so neither AI nor customers hit a dead end.

5. Is your language simple, action-based, and consistent?

Use direct, predictable phrasing. Avoid filler like “Don’t worry!” and focus on steps customers can actually take.

By tweaking structure instead of your content, it’s easier to turn your Help Center into a self-service powerhouse for both customers and your AI Agent.

Make your Help Center work smarter

Your Help Center already holds the answers your customers need. Now it’s time to make sure AI can find them. A few small tweaks to structure and phrasing can turn your existing content into a powerful, AI-ready knowledge base.

If you’re not sure where to start, review your Help Center with your Gorgias rep or CX team. They can help you identify quick wins and show you how AI Agent pulls information from your articles.

Remember: AI Agent gets smarter with every structured doc you publish.

Ready to optimize your Help Center for faster, more accurate support? Book a demo today.

{{lead-magnet-2}}