“How to Run a Chatbot Program?” – This is not something you can learn in a day. However, we intend to make it easier by sharing a few resources and knowledge which we managed to collate through consistent effort. It would require one to go through the following steps:

1. Finalize your use case
2. Create a user-centric design
3. Write the chatbot script
4. Create flows on Glific
5. Test with your internal team
6. Run pilots
7. Scale

Finalize your use case: There are multiple ways in which NGOs use chatbots. This blog will give you an idea of different use cases and how different organizations are utilizing them.

While you might resonate with multiple use cases in the blog and want to try them all, it is important for you to start with one use case—either a pressing one or a low-hanging fruit—and then further expand to more use cases. We suggest starting with something where you have a strong on-field presence and access to users, so that you are able to try, test, and iterate at a quick pace.

Create a user-centric design: Before you start drafting messages for your chatbot, it is important for you to first design how your chatbot will look, feel, and operate for your users. To do this, it is important to start by defining your users from a chatbot perspective, as it will bring in more nuances to help you while creating, evaluating, and running a chatbot.

To begin, here are a few questions that we urge you to think about and document for yourselves as a team. Once you have answered them, make decisions which will help you to design a user- centered Chatbot.

Write a script for Conversational flows: Draft the messages you want to send to your users, ensuring the tone, language, vocabulary, and length are suitable for your users. Here is a blog outlining a few things you need to keep in mind while writing your first chatbot script.

Every organization has a different approach to maintaining their content:

• Spreadsheets are commonly used. Sample sheet (to be updated).
• Some use Figma or Miro for easier visualization

Create flows on Glific: Glific offers multiple features that we suggest you explore to make the best use of the platform in solving your problems. To begin, here is a video on how to create a basic chatbot flow on the Glific platform.

Before sharing the flows with your team or end-users for testing, here are 15 quick things you need to cross-check in your flows to ensure everything is in place.

Test with your internal team: Before you share your bot with end-users, it is crucial to test it with your internal team. They will provide the best user perspective and honest opinions. Make changes in your flow as per the feedback (if required) before sending it to the end-users

Run Pilots: Before scaling your chatbot to all your target users, it is crucial to run a few pilots to gauge how it is received by your end-users. Evaluate, learn, and iterate based on the pilot feedback before expanding its reach to a larger audience. 

Few points to keep in mind while conducting a Pilot:

1. Define objectives & success metrics:
   • As a team, define what you aim to achieve through this pilot and the criteria for considering it successful. 
   • It is equally important to go ahead with the assumptions you are going ahead with Eg- All users have access to smartphones and know how to use it.
   • Avoid setting high engagement goals for the pilot. The primary objective should be learning rather than success. Initially, targeting a 20% engagement rate can be reasonable.
2. Selecting the Pilot group:
   • Choose a sample size representative of your larger group.
   • Select individuals who are part of your existing programs rather than a completely new group (individuals you can reach out to for qualitative feedback).
   • We recommend keeping the first pilot group to no more than 50 people.
3. Fail fast, Learn fast:
   • Run pilots for shorter durations (2-4 weeks) to quickly learn and iterate.
4. Monitor & Evaluate:
   • This is crucial for running pilots. Collect both qualitative and quantitative feedback (design your flows keeping this in mind) on your chatbot program to understand its success and user feedback. Reflecting as a team using data will help in identifying gaps and addressing them.
5. Iterate:
   • We recommend you to run at least 2-3 short pilots, implementing changes based on learnings from previous pilots, before scaling the chatbot

Scale: Once you have successfully run different pilots, strategically plan how to scale your chatbot. While there are many factors to consider during scaling, here are a few that might help:

1. Onboarding users: Onboarding users and collecting their basic details is one of the first and most important steps in your chatbot journey. In this detailed blog, we covered five common onboarding methods used by Glific NGOs to simplify how users join the chatbot. We’ve also included the pros and cons of each method to help you make a decision based on your program’s design. Along with the onboarding methods, we also included two methods to streamline the collection of user details (during the registration process) in the blog.  
2. Planning budget- There are 4 players in the larger costing effort (taxes as applicable everywhere)

• Glific – You can find Glific’s Monthly subscription charge here–  this is the only cost you would be paying Glific
• Whatsapp – Pay per use- has a unit of measure called a “Conversation” –  it is any number of messages being initiated in a 24 hour window. Whatsapp charges for the first message around @ INR 90p (0.0107 dollar) or INR 12p(0.0014 dollar) as per  type of message if started by the bot and at INR 34p per message (Dated 17th Oct,2024) . Know more
• Gupshup – Pay per use on number of messages exchanged-per message 0.001dollar. Charges up to USD 75 a month (converts to about INR 6350 per month) for about 75,000 messages. Any exchange of messages beyond 75,000 messages each month is free, you will be billed only USD 75 (Dated 1st May,2024)  Know more
• Google cloud storage & Big query – Pay per use. Your messages (incoming and outgoing) are stored here, with any media files sent or received. Your dashboards are powered by querying the data stored here.

Here is a pricing calculator which you can (make a copy and) use to calculate your monthly budget by adding your number of users, number of messages you want to send and frequency.

3. Growing/ Hiring team- As you scale up, you will definitely need to expand your team as well. Here are a few responsibilities typically included in the job description for the owner of a chatbot:

• Conduct comprehensive user research to gain insights into user needs, motivations, and pain points.
• Develop the chatbot’s design based on insights from user research and oversee its implementation (Content creation is part of it).
• Continuously monitor and analyze chatbot data while gathering user feedback periodically.
• Drive enhancements and innovations within the chatbot using feedback-driven insights and evolving user requirements.

Organizations that operate at scale (with more than 5,000 users) usually have more than one team member. Teams are typically segmented for content creation, flow creation, monitoring and evaluation, etc.

I hope this blog has given you a rough idea of how to run a chatbot program with Glific. Please let us know if you think we need to add more details that would help you better.

Other external blog which might interest you: https://medium.com/voice-tech-podcast/conversation-design-workflow-how-to-design-your-chatbot-in-10-basic-steps-721652b056d