Table of Contents
- Key takeaways
- What are chatbot case studies?
- The three SiteGPT chatbot case studies at a glance
- Case study 1: The SaaS Jobs, an AI career guide on a job board
- The problem
- The setup
- Where to read more
- Case study 2: CBS Bahamas, a retail chatbot with Zendesk escalation
- The problem
- The setup
- Where to read more
- Case study 3: a lighting and electrical group with five language stores
- The problem
- The setup
- Where to read more
- What do these AI chatbot deployments have in common?
- How do you measure AI chatbot deployment results?
- The metrics that matter
- Why adoption is the hidden variable
- A worked example of chatbot ROI
- How to write an AI chatbot implementation case study
- Which chatbot case study matches your situation?
- Which SiteGPT plan fits each setup?
- Frequently asked questions
- Basics
- What is a chatbot case study?
- What makes a chatbot case study credible?
- Where can you find more chatbot examples?
- Results and ROI
- What results can an AI chatbot deployment deliver?
- How long does it take to see chatbot results?
- How do you calculate chatbot ROI?
- SiteGPT setups
- Which SiteGPT plan do these case studies use?
- Can SiteGPT hand a conversation to a person?
- Does SiteGPT answer in more than one language?
- Sources
Created by
Do not index
Created time
Mar 10, 2026 10:09 AM
These chatbot case studies cover three businesses that run SiteGPT today: a SaaS job board, a retailer in The Bahamas, and a lighting and electrical group with five language stores, plus a method for measuring the results of your own AI chatbot deployment.
TL;DR
- The SaaS Jobs runs a SiteGPT career guide on a job board whose listings change daily. The SiteGPT widget is live on the site.
- CBS Bahamas runs a SiteGPT retail chatbot with an escalation path to Zendesk, according to its Vice President of Retail and Sales on the SiteGPT home page.
- A lighting and electrical group runs five SiteGPT chatbots, one per language store, each trained on that store's product pages.
- All three train the chatbot on content the business had already published, so setup needed no new training data.
- To judge any chatbot case study, look for a baseline, a named metric, a time window, and a source you can open. Then run the same math on your own numbers.
Key takeaways
Case study | Business type | What the chatbot does | SiteGPT feature it depends on |
Job board for SaaS roles | Guides job seekers through live listings and career questions | Daily sitemap scan (Scale and above) | |
Retailer | Answers product questions and escalates to the team in Zendesk | Zendesk integration (Scale and above) | |
Ecommerce, five language stores | Answers technical pre-sale questions on product pages | Several chatbots, each trained on its own store (Enterprise for 4 or more) |
What are chatbot case studies?
Chatbot case studies are written accounts of how a business deployed a chatbot, what problem it solved, and what changed after launch. A useful one names the business, the setup, and the result, and lets you check at least part of it.
This page collects SiteGPT's own customer stories. SiteGPT is the subject here, so it is not ranked against other tools. Each story has a full write-up, linked below, and this page summarizes each in a few lines.
For chatbot examples from other vendors and large brands, three other guides cover that ground:
- Companies using chatbots for customer service collects public deployments from large brands.
- Customer service chatbot examples shows what good answers look like, conversation by conversation.
- Ecommerce chatbot examples breaks down store patterns, from pre-sale questions to order status.
The three SiteGPT chatbot case studies at a glance
The three SiteGPT chatbot case studies differ by industry, content type, and how the team hands off to a person. The table shows what can be checked on a live page today.
Customer | Problem | Setup | What you can check today |
The SaaS Jobs | Job seekers had questions a search bar could not answer, and listings change daily | One SiteGPT chatbot trained on the job board's pages | The SiteGPT widget loads on The SaaS Jobs website |
CBS Bahamas | Shoppers asked product questions around the clock | A SiteGPT chatbot with escalation to Zendesk | A named testimonial on the SiteGPT home page |
Lighting and electrical group | Repeat technical pre-sale questions arrived by phone and email in five languages | Five SiteGPT chatbots, one per language store | The customer is not named publicly |
Case study 1: The SaaS Jobs, an AI career guide on a job board
The SaaS Jobs case study shows a chatbot working on content that changes every day. The SaaS Jobs is a job board for roles at SaaS companies, and its listings are added and removed constantly.
The problem
Job seekers wanted answers a search box cannot give. They asked which roles fit their background, which skills to build, and how to move into SaaS. Employers had their own questions about posting jobs.
The setup
The SaaS Jobs runs one SiteGPT chatbot for both groups. It answers from the job board's own pages, including live listings. The SiteGPT widget script loads on the site today.
A job board needs its chatbot to notice new and expired listings. On SiteGPT, the daily sitemap scan finds new or removed URLs, and it is included on Scale and Enterprise. Starter has manual refresh only, and Growth refreshes monthly.
Where to read more
The full story, with the conversation types and the setup steps, is in the career guidance chatbot case study.
Case study 2: CBS Bahamas, a retail chatbot with Zendesk escalation
The CBS Bahamas case study shows a retail chatbot that answers product questions and passes the hard ones to people. CBS Bahamas is a retailer in The Bahamas.
The problem
Retail shoppers ask product questions at any hour, and many of them need a precise answer before they buy. Staffing live chat around the clock is expensive.
The setup
CBS Bahamas runs a SiteGPT chatbot trained on its own website content, with an escalation path to its team in Zendesk. Brent Burrows II, Vice President of Retail and Sales at CBS Bahamas, describes the setup on the SiteGPT home page:
"We've got the bot dialled in - we're using GPT-4, have an avenue for escalations to Zendesk, and so far I have no complaints."
A setup like this needs two things on SiteGPT today. The Zendesk integration is on Scale and above. The built-in human handoff, which lets a visitor ask for a person, works on every plan.
Where to read more
The full retail story, including the product questions the chatbot handles, is in the AI chatbot for retail case study. For other ways to run a chatbot inside Zendesk, see the best AI chatbots for Zendesk.
Case study 3: a lighting and electrical group with five language stores
The lighting group case study shows how one business runs several chatbots, one per market. The customer sells lighting and electrical components through five online stores, each in its own language.
The problem
Lighting and electrical parts are technical purchases. Shoppers ask about compatibility, specifications, and installation before they buy. Those questions arrived by phone and email in five languages, even when the answer was already on the product page.
The setup
The group runs five SiteGPT chatbots, one per store. Each chatbot is trained on its own store's product pages and answers in that store's language. The team reads chat histories and fixes wrong answers in the source content.
Five chatbots is more than the 3 that Scale includes. A setup this size sits on SiteGPT Enterprise, which has custom pricing. A business with one catalog in several languages can often run a single chatbot, because SiteGPT answers in 95+ languages.
Where to read more
The full story, including the daily review routine, is in the ecommerce chatbot case study. For more on running one bot across markets, see the guide to multilingual customer service.
What do these AI chatbot deployments have in common?
These three AI chatbot deployments share one habit in every case: the chatbot answers from content the business had already published. The table shows which other habits each story describes.
Habit | The SaaS Jobs | CBS Bahamas | Lighting group |
Train on content that already exists | Job listings and site pages | The store's website content | Each store's product pages |
Keep that content current | Listings change daily | Not described publicly | Fixes made in the source pages |
Give visitors a route to a person | Not described publicly | Escalation to Zendesk | Not described publicly |
Review real conversations | Chat history shows what job seekers ask | Not described publicly | Daily review of chat histories |
The first habit keeps setup short. Nobody wrote new training data. Each business pointed the chatbot at pages it had already published.
Reviewing conversations is how a team finds gaps. A wrong answer usually traces back to a missing or outdated page. Fixing that page fixes every later answer on the topic. The full list of habits is in chatbot best practices.
How do you measure AI chatbot deployment results?
You measure AI chatbot deployment results by comparing the same metric before and after launch, over the same length of time, on the same channel. Without a baseline, a case study reports activity, not results.
The metrics that matter
Metric | How to calculate it | Where the number comes from |
Resolution rate | Conversations that end without a handoff or a repeat contact, divided by all conversations | Chatbot analytics plus your helpdesk |
Contact volume on the old channel | Phone calls, emails, or tickets per month, before and after launch | Phone system or helpdesk reports |
Escalation rate | Conversations handed to a person, divided by all conversations | Chatbot analytics |
Cost per resolved question | Monthly chatbot cost divided by questions resolved without a person | Billing plus chatbot analytics |
Conversion or sign-up rate | Visitors who chatted and then bought or signed up, against those who did not | Web analytics |
Why adoption is the hidden variable
Adoption decides how much of the result you will see. Gartner surveyed 3,566 B2B and B2C customers in February and March 2026. 49% said they would have used a company chatbot if one were offered. Only 7% used a chatbot or digital assistant in their most recent service interaction.
Gartner tells service leaders to account for that 42-point gap when they estimate chatbot adoption and ROI. Only 27% of customers would try a chatbot again after a negative experience, so a few bad answers early on lower the result for months.
A worked example of chatbot ROI
This example uses hypothetical numbers to show the math. Swap in your own figures, or use the chatbot ROI calculator.
A store gets 1,000 customer questions a month. Each takes an agent 6 minutes, and an agent costs $30 an hour.
Share resolved by the chatbot | Agent hours saved per month | Saving per year | Growth, billed yearly ($948) | Scale, billed yearly ($3,108) |
20% (200 questions) | 20 hours | $7,200 | $6,252 net, 660% ROI | $4,092 net, 132% ROI |
40% (400 questions) | 40 hours | $14,400 | $13,452 net, 1,419% ROI | $11,292 net, 363% ROI |
The saving is hours saved times $30 times 12. Net is the saving minus the plan's yearly cost, and ROI is net divided by that cost. Check that the plan's message quota covers your volume too. Growth includes up to 10k GPT-4.1-mini messages a month, and Scale up to 40k.
Billing yearly also lowers the cost side. Starter costs $708 a year billed monthly or $468 billed yearly, a $240 (34%) saving. Growth costs $1,548 or $948, a $600 (39%) saving. Scale costs $5,148 or $3,108, a $2,040 (40%) saving.
How to write an AI chatbot implementation case study
An AI chatbot implementation case study is only as strong as the numbers recorded before launch. Use these steps to document your own deployment so the result holds up when someone checks it.
- Record the baseline. Write down monthly contact volume, average handling time, and cost per contact for the 90 days before launch.
- Write the scope down. List the question types the chatbot will answer and the ones that go to a person.
- Describe the setup. Note the content sources, the refresh schedule, and the handoff route.
- Measure at 30 and 90 days. Use the same metrics, the same channels, and the same length of period as the baseline.
- Separate cause from timing. Note anything else that changed in the same period, such as a price change, a new product, or a seasonal peak.
- Get sign-off on names and quotes. Publish a name, a title, or a quote only with the customer's written approval.
For the build itself, the step-by-step guide to how to create an AI chatbot covers training, testing, and launch.
Which chatbot case study matches your situation?
The right chatbot case study to read depends on how often your content changes, where your team already works, and how many markets you serve.
If you need... | Read... | Why |
A chatbot for content that changes every day | Its listings are added and removed constantly | |
A retail chatbot that hands off to agents in Zendesk | Its escalations go to Zendesk | |
Separate chatbots for several stores or languages | It runs one chatbot per language store | |
Examples from large brands and other vendors | It covers deployments beyond SiteGPT | |
B2B use cases such as lead capture and onboarding | It focuses on sales and account questions | |
A cost estimate before any build | It turns your ticket volume into hours and dollars |
Which SiteGPT plan fits each setup?
Each SiteGPT plan fits a different size of deployment. The table maps the setups above to the lowest plan that includes the feature they rely on.
Setup | Feature it needs | Lowest plan | Price |
One chatbot on a small, stable site | Manual refresh, 1 chatbot, 1,000 pages | $59 a month billed monthly, or $39 a month billed yearly | |
Two chatbots, Slack or Crisp, monthly refresh | Integrations, 2 chatbots, 10,000 pages | $129 a month billed monthly, or $79 a month billed yearly | |
Daily listing changes, like The SaaS Jobs | Daily sitemap scan, weekly refresh | $429 a month billed monthly, or $259 a month billed yearly | |
Escalation to Zendesk, like CBS Bahamas | Zendesk AI Agent | $429 a month billed monthly, or $259 a month billed yearly | |
Five store chatbots, like the lighting group | More than 3 chatbots | Custom |
SiteGPT has real limits to plan around. Starter has no chat integrations and no automatic refresh. The self-serve plans stop at 3 chatbots. There is no Intercom or HubSpot integration today.
Every self-serve plan starts with a 7-day free trial, and the trial requires a credit card.
Frequently asked questions
Basics
What is a chatbot case study?
A chatbot case study is a written account of one business's chatbot deployment. It describes the problem, the setup, and the result, ideally with a baseline and a source the reader can check.
What makes a chatbot case study credible?
A credible chatbot case study names the business, states a baseline, measures the same metric after launch over a set period, and links to a source outside the vendor's own marketing. Treat any figure without a baseline or a source as a claim.
Where can you find more chatbot examples?
More examples are in companies using chatbots for customer service, customer service chatbot examples, and B2B chatbot examples.
Results and ROI
What results can an AI chatbot deployment deliver?
An AI chatbot deployment can lower contact volume on phone and email, answer questions outside working hours, and show which questions customers ask most. The size of each result depends on how many visitors use the chatbot and how often it answers correctly.
How long does it take to see chatbot results?
Most teams can compare results at 30 and 90 days after launch. The first 30 days show adoption and the most common questions. By 90 days there is enough data to compare against the baseline.
How do you calculate chatbot ROI?
Multiply the agent hours the chatbot saves by the hourly agent cost, subtract the chatbot's yearly cost, and divide by that cost. At 400 questions a month, 6 minutes each, and $30 an hour, the saving is $14,400 a year. Against SiteGPT Growth billed yearly at $948, that is a 1,419% ROI.
SiteGPT setups
Which SiteGPT plan do these case studies use?
The features they depend on set the plan. Daily sitemap scanning and the Zendesk integration are on Scale and above. More than 3 chatbots needs Enterprise.
Can SiteGPT hand a conversation to a person?
Yes. SiteGPT's human support works on every plan. By default the AI escalates when the visitor asks for a person, when the situation is urgent, or when the visitor is frustrated after an unhelpful answer.
Does SiteGPT answer in more than one language?
Yes. SiteGPT supports 95+ languages. A business with different product ranges per market can still run one chatbot per store, as the lighting group does.
Sources
Checked on 16 October 2026.
Written by
