Artificial intelligence chatbots are transforming customer service and employee support. Powered by natural language processing (NLP), chatbots can understand questions posed in human language and instantly deliver answers from a knowledge base.
This tutorial will show you how to easily create your own AI-powered chatbot in just 30 minutes! Here's what we will be doing:
- Set up a SharePoint document library to hold our training data
- Automate the process of turning documents into chatbot learning
- Launch a custom chatbot trained on our data and embed it in a public website
With simple drag-and-drop workflows and no coding required, you can put the power of artificial intelligence to work for your business. Let's get started!
1 - The Power of AI Chatbots
AI chatbots offer many benefits:
- Instant answers for customers - Chatbots can respond immediately 24/7, improving customer satisfaction. They can handle common inquiries, freeing staff for more complex issues.
- Reduced support costs - Automated chatbots lower labor costs for support and customer service. They scale easily without added staff.
- Accessible knowledge - A chatbot makes your knowledge base available instantly to customers and employees. They can find answers at any time of day.
- Marketing assistant - Chatbots shine for lead gen and marketing. They can ask qualifying questions and share product details.
- Research helper - Train a chatbot with your industry research and data to deliver insights.
- HR portal - Let employees query a HR document library to get quick policy answers.
The possibilities are endless when you build a chatbot tailored to your business!
Want to test out our new Chatbot?
Click the red bubble at the bottom right (on this page) and chat away. We're still "training" it, but you will hopefully get the idea. It's been trained on our training catalogue from the Collab365 Academy. We're currently testing out GPT-3.5 to see if it provides adequate answers. There's an option to upgrade to GPT 4, which we'll more than likely move to in the near future.
2 - Demo of Our Chatbot
Let's walk through a specific example to understand how our chatbot will work.
We will train the chatbot using text documents that describe workshops offered by the Collab365 Academy. Each document will contain details like:
We will upload these description documents to our SharePoint training library.
The Make workflow will extract the text and send it to Chatbase to train our bot.
Now the chatbot will be able to answer questions about the workshops intelligently, like:
"What are the best workshops for beginners?"
what are the best workshops for beginners
The chatbot can scan the documents and recommend the right intro level workshop.
Or we can ask:
"What 3 advanced workshops will help me get a high paying cloud job?"
what are differences with platinum and community
You have the full power of GPT so you can get clever.
"I don't speak very good English, as I am french. Can you let me know about Connor Deasey, please? Answer in French"
you can also get it to respond in a different language
Pretty cool, huh! All you need to do is train it on your data and give it a good starting prompt (in the Chatbase dashboard).
This demonstrates how our chatbot can go beyond just keyword matching to understand context and intent based on all the training data.
3 - Tools We'll Use
To set-up our new AI-Powered Chatbot. we'll leverage 3 platforms to create our chatbot, namely Make (the glue between SharePoint and Chatbase), SharePoint (to store the knowledge for our Chatbot in a document library) and Chatbase (to allow your customers, partners or employees to ask native-language questions).
a scenario to publish to chatbase in make
What is Make?
Make is a no-code automation platform that allows anyone to build workflows and integrate apps visually without coding.
Key features:
- Intuitive drag-and-drop interface to connect apps and data sources into workflows called scenarios.
- Build scenarios with unlimited steps and complexity to automate end-to-end processes.
- Real-time data flows let you visually map data between apps.
- Access thousands of pre-built integrations and public APIs with the HTTP module.
- No-code toolkit provides advanced functionality like conditional logic, error handling, data formatting.
- Scenarios go live instantly with no servers to manage.
- Granular permissions allow collaboration across teams.
- Analyze execution history to optimize workflows.
Benefits:
- Speed - go from idea to automation in minutes/hours instead of weeks.
- Power - build complex workflows as if you were a developer.
- Flexibility - connect endless app combinations tailored to your needs.
- Scalability - easily maintain and enhance workflows over time.
- Collaboration - design, build, and improve as a team.
Make allows anyone to digitally transform their business by automating manual processes, connecting disparate systems, and streamlining operations without coding. It brings the power of developers to every employee.
For this use-case, Make allows us to automate workflows to ingest and process data without coding. We'll use it to:
- Watch for new SharePoint documents
- Extract text from documents
- Send data to Chatbase to train the AI
With simple drag and drop modules, we can build the automation in minutes.
SharePoint
SharePoint will be our document library to hold the content we want our chatbot trained on. We can add Word files, PDFs, and text documents full of FAQs, product info, manuals, etc. Please note: if you do add documents such as PDF's and Docx's you will need to use one of Makes modules to convert them to text, costing one more "operation". For this tutorial, I have used a text files and also Chatbases ability to add urls from our website.
Once we set this up, as we add more documents (to SharePoint), Make will automatically pull the text and feed it to the chatbot.
What is Chatbase?
Chatbase is an AI chatbot builder that makes it easy to create customized chatbots powered by your own data. Simply upload documents, paste text, or connect a website, and Chatbase will train an AI chatbot to answer questions about that content.
Some key features:
- Supports 95 languages - train the chatbot in any language.
- Uses GPT-3.5 and GPT-4 models for advanced conversational ability.
- Embed chatting widget on your website or use the API anywhere.
- Give your chatbot a name, personality, and custom instructions on how to respond.
- Make chatbots public or private - easily share or collaborate.
- Your data is securely stored on Google Cloud and AWS servers.
- Intuitive editor to train, test, monitor, and enhance your chatbot.
The benefit of Chatbase is getting a tailored AI chatbot powered by your own data in minutes without any coding. It handles the complex model training for you and makes it easy to create intelligent conversational agents.
Whether you want to improve customer service, share knowledge, or create an AI assistant, Chatbase allows anyone to leverage the power of chatbots quickly.
Now let's see how to put these three platforms together...
4 - Step-by-Step Tutorial
Follow these steps to create your own AI chatbot in 30 minutes:
Step 1 - Create Accounts
First, we'll sign up for free accounts with Make and Chatbase:
Step 2 - Set Up SharePoint
Next, create a SharePoint Online document library to hold our training data:
- Login to your SharePoint Online site
- Create a new document library named "Chatbot Training"
- We'll come back soon to add documents
Step 3 - Configure Chatbase
Head over to Chatbase to set up our chatbot:
- Click Create Bot
- Give it a datasource to start with, either text, Q&A, url or a PDF
- Click Create Chatbot and then let it train.
- Give it a name
- Take a look in settings and pay particular attention to the "Temperature" and "Base Prompt". It's really important to finetune the Base Prompt because this will be used for all of your chat messages. It's your opportunity to give the Bot a command on every single message.
- Here's an example of our Base Prompt:
I want you to act as a Collab365 Customer Services person. Your name is "Collab365 Assistant". You will provide me with answers from the given info. Only suggest Collab365 Training (our workshops, challenges, success paths and conferences). If the answer is not included, say exactly "Hmm, I am not sure." and stop after that. Refuse to answer any question not about the information we supply. Never break character or suggest answers that aren't offered by Collab365. Suggest url's for more information where appropriate.
There's plenty more to learn about with Chatbase, but I will leave you to investigate.
To test this really quickly, let's add some initial training phrases:
- Go to the Train tab
- Add 5-10 common questions and the ideal responses
- This gives the bot some starting data
- Now click the Chatbot tab and ask it some questions
Step 4 - Build the Make Scenario
Now we'll automate pulling text from SharePoint into Chatbase:
- In Make, create a new Scenario
- Add a SharePoint module, connect your site, and configure it to watch the new document library
- Add a HTTP module to get the file contents. Connect it to the SharePoint module.
- Add a JSON module to format the text
- Add the Chatbase module, connect your account, and configure it to call the train endpoint
the settings for the make scenario
That's it! The workflow will now send any new document text to the Chatbase API to train the bot.
Step 5 - Upload Training Data
Let's add some content to SharePoint for the bot to learn from:
documents to train chatbase in sharepoint
- Create several text documents with Q&As, product descriptions, help articles etc.
- Upload them to the SharePoint document library
- The Make workflow will automatically run on a schedule to send them to the Chatbase API
Keep adding more documents whenever you want the chatbot to learn more.
Step 6 - Embed the Chatbot
chatbase embedded in a website
Finally, get the code to embed your chatbot:
- In Chatbase, go to "Embed on website"
- Copy the Embed code
- Paste it into an area on your website page that allows you to add scripts
You now have your own AI chatbot online and working on your website.
TIP: You could also add the chatbot to a Microsoft Teams Channel using another Make Scenario.
5 - Improvements and Tips
Here are some additional steps to further improve your chatbot:
- Use Word/PDFs - Add modules to Make workflow to extract text from these formats
- Scrape websites - Supply Make a list of URLs to automatically scrape and feed to the chatbot
- Monitor usage - Check Chatbase conversations to see which questions are asked and how well the bot is answering. Focus training on weak spots.
- Add to Teams - Embed the bot in Teams channels so employees can easily query the knowledge base.
The great part about this no-code approach is that you can continuously improve the bot by adding more data for it to learn from.
Conclusion
In just 30 minutes, we built an AI-powered chatbot using SharePoint, Make, and Chatbase without writing any code.
The key steps were:
- Creating accounts
- Setting up SharePoint document library
- Configuring the initial Chatbase chatbot
- Building the automation workflow in Make
- Populating SharePoint with training data
- Embedding our finished chatbot
This guide showed how easy it is to put the power of artificial intelligence to work for your business with a custom chatbot trained on your own data.
You can now go even further by adding more documents, integrating with websites and Teams, and monitoring performance. The possibilities are endless!
Want to ask questions or discuss this?
Grab a free community account and leave a comment on this post.