How to scope your next chatbot project

How to scope your next chatbot project

How to scope your next chatbot project

When sitting down to decide what information your chatbot should know (i.e. scoping your chatbot), you’re actually trying to predict what questions your future users will ask. And by giving your users a keyboard, you’re opening up the list of possible questions to include literally everything and anything

 

When a user asks a question, your bot will use natural language processing (NLP) to try and determine the intent (or purpose) of your user’s question. These intents are the building blocks of your bot’s knowledge. 

 

For your bot to recognize an intent, you need to identify what questions your users will ask, build an NLP model based on that information and train your bot with +40 expressions per intent. 

There’s a big difference between what your bot could know and what it should know. But to get to the should, you need to start with the could. Below we share our favorite places to look for possible intents and our tips on selecting the most impactful ones. Because at the end of the day, your chatbot needs to save you time and money.

Gather all the possible intents

The scoping phase of a chatbot project is when you gather all the possible intents. This is the time to dig out a magnifying glass and play Sherlock Holmes. Just like a detective, you need to poke around in the dusty corners, interview witnesses and listen for the truth. 

 

This is the phase where you’ll look into all the possible questions your bot could answer. If you’re feeling impatient and want to know what your bot should answer, skip ahead to the next section.

Start with the FAQs

There’s a reason companies have FAQs– they’re the select set of questions that users ask the most. Not only do you already know that these are high-volume questions, but the content is also structured and the answers are there! It’s almost a perfect outline for your chatbot’s scope. All you need to do is rework the questions to match a single intent and rewrite the copy to fit a conversation.

Chat with your customer support team

No one knows your users better than your customer support agents. They interact with and solve your users’ questions and issues daily. Most customer support teams have a knowledge library of questions and answers beyond the basic FAQs. Similar to an FAQ, you can assume that users ask these questions repeatedly, and there’s already a clear outline of how the bot should answer.

 

Customer support teams are also your gateway to the “information universe.” They manage the access to the digital archive of all past user conversations, and in most cases, they classify these conversations with topics, tags, and have template responses. This is a friendly warning that reading through real user conversations can lead you down the black hole of information overload, but we swear that this is the best way to understand who your bot will talk to and how they’ll respond to it. 

 

In case you’re still not convinced, the final reason to involve your customer support agents in scoping your chatbot is that they’ll be your bot’s direct colleagues. When your bot can’t close a user request on its own, your customer support agents will take over the conversation via an offload flow. You need to know how they work to design a bot that will help them, and they need to understand how the bot will work to help users better.

Review company documentation

Is there an onboarding manual? Or a handbook of company processes? Since this is information that your company wants to share with employees, partners, customers and more, it’s a great thing to read for inspiration. You’ll be pleasantly surprised how many intents you can find in a basic About-CompanyX.pdf.

Identify the manual processes

Chatbots exist to automate repetitive tasks. Basic manual processes inspire some of the best intents and future conversational flows.

 

A common manual process is data capture. You can easily put the bot in charge of collecting and updating user data, so that your human colleagues can focus on more complex tasks. Managing appointments, routing to the right queue and sending reminders are other processes that a chatbot can easily take over.

Turn user questions into chatbot intents

Questions and intents are similar, but not the same. A question is literally what your user asks, while an intent is the purpose of what your user asks. For example, a user asks Hey Google! What’s the number for Amy’s Pizza? The Hey Google… is the question, but 01-restaurant-contact-info is the intent.

When scoping a chatbot make sure you understand the difference between question, intent and response

Transforming a question into an intent isn’t always easy. The key is to capture the essence in one to three words. A good rule of thumb is that if your intent doesn’t fit on a post-it, it’s too long.

Choose the best intents

Now that you have a collection of possible intents, it’s time to decide which intents will have the largest potential impact and therefore, the highest priority. For a pilot project, we recommend starting with ~20 intents. It’s always better to get a small set of intents working well before making a dumpster of half-understood intents.

Look for the trifecta

The best intents are the ones that are repetitive, easy to recognize and easy to answer. When you find this trifecta, put this intent at the top of your implementation backlog.

Repetitiveness, recognizability and easiness are the trifecta for choosing intents to include in your chatbot's scope

But how do you know if you have a trifecta? We use a systematic framework to visualize which intents will have the largest potential impact.

 

Recognizability: Can a bot easily recognize this intent? Is it clear and unique, or very vague, or similar to other intents?

Repetitiveness: How frequent is this intent? Will it be triggered 200 times a day or twice a day?  

Easiness: How simple or complex is it to implement this intent? Do we need to use APIs, create a microservice, etc.? 

 

Review each intent and give it a score from 1-10 for recognizable and repetitive (1 being the least and 10 being the most). Then give the intent a score from 1-5 based on how easy it’ll be to implement (1 being the easiest and 5 being the most difficult). 

 

Now it’s time to map each intent on a graph; recognizability is on the y-axis and repetitiveness is on the x-axis. Once you have the coordinates for an intent, mark it with its easiness score.

A cluster in the top right quadrant will appear. These are the intents that you should prioritize for your chatbot. Start with the easiest intents to implement, and then work your way up through the more difficult ones.

How to decide which intents to include in your pilot chatbot's scope

Any intents that don’t make the list for the pilot should go on your backlog so that you can pick them up to include in future releases. 

A chatbot’s knowledge isn’t static

Launching your bot is only the beginning. This is one of our favorite diagrams of all time because it’s so simple, yet so accurate 👇

How people think a bot works vs. how it actually works

The best source of new intents for your chatbot comes from its users. Once your bot is live, take time to review the incoming conversations. This allows you to make sure that the bot’s recognizing existing intents correctly and to see what other questions users ask that are currently not in scope. 

 

Look for patterns and figure out how the bot could answer that question instead of putting it in a not understood flow. Your priority should be to make every experience a neutral or positive experience for the user. Looking at the missed questions gives you all the necessary information to increase your bot’s knowledge. The best bots grow continuously and are updated regularly to answer actual user questions. 

5 ways to use conversational artificial intelligence in banking

5 ways to use conversational artificial intelligence in banking

5 ways to use conversational artificial intelligence in banking
It’s no surprise that the financial industry is perfect for applying artificial intelligence (AI)– banks, crypto companies and financial institutions have thousands of customers who use their services multiple times a day.

We often receive the question from prospective banking clients: How can we use conversational AI? Here are 5 of the best ways to implement conversational AI in banking, insurance and crypto companies.

Customer support

Setting up a chatbot to answer repetitive customer questions such as: What is the BIC code?  and Where can I find my routing number? is a no-brainer for any bank.

These questions come in at high volumes, they’re easy to recognize and the answers are relatively simple– usually, a short text, link or image can answer the user’s question. Let’s assume that What is the BIC code? represents 5% of the incoming questions. If you use a bot to automate that answer, you reduce your incoming question volume by 5%. FAQs are a great area to start with for your MVP (minimum viable product) because even with a small scope of ~20 intents, you can immediately experience the benefits of automation by reducing the volume of incoming questions and the time spent by your customer support agents answering them.

Plus, most customer support tools (Sparkcentral, Zendesk, Freshdesk, etc.) have out-of-the-box connections to chatbot building software, so you can keep your existing technology stack. BTC Direct’s Toshi and KBC’s Kate are two of our favorite customer support chatbots automating conversations in the financial industry.

Data capture

Another area where banks tend to lose a lot of time is during data capture flows. Many financial procedures require additional information before completing a request. Sometimes asking for a user’s name, birthday and email can take up to 10 min. By using a virtual assistant to automate the data capture, you avoid having your agent wait for the customer to respond and the customer wait around to have the agent ask the next question. It’s a win-win for both sides!   Some typical baking processes that require additional information and could benefit from automated data capture are:
  • Raising daily limits
  • Requesting a new card reader
  • Ordering foreign currency
  • Canceling or blocking a card
 
We’re probably biased, but Argenta’s Charlie is our favorite example of automated data capture in a banking app. Another great example is Belfius’s myBo.
Argenta's chatbot Charlie answers user questions

Financial transactions

Forget logging into an app or, even worse, opening your computer. Encrypted messaging apps, like WhatsApp, Signal and Telegram make checking your balance, paying bills and transferring money between accounts easier than ever. And the best part? Customers can use an app they already have instead of installing a new one!

 

Check out Capital One’s Eno and Rawbank’s Rawbot to see how they use conversational AI to execute financial transactions.

Filing insurance claims

Filing an insurance claim is never a fun process. But it’s even worse when you don’t know what to do and have to wait for someone to help you. Developing an insurance claims chatbot is one of the more complex implementations, but the time saved and improved customer experience make this a huge cost savings opportunity in the long run. Instead of spending hours processing a single claim, AI automates the complete data capture, so a claims agent only needs to spend a few minutes reviewing the case.

We can also use AI to detect fraudulent language and build flows within the chatbot that stop users from submitting fake claims or adapting existing claims to increase their insurance payout.

For an excellent implementation of an insurance claims chatbot, take a look at Belfius’s myBo.
Belfius insurance claims chatbot

Coaching and budget management

Chatbots make great coaches for many reasons: they’re always on time, they never tire of nagging and most importantly, they don’t judge. We’re human, and sometimes we need a (not so) gentle reminder to keep ourselves in line, especially when it’s about budget management. Sticking to a budget and reaching financial goals, such as saving for a house or retirement, require constant discipline.

Building a coach-like virtual assistant is one of the most challenging implementations because it requires integrating with your bank’s infrastructure and security requirements, engineering If This, Then That(IFTT) triggers and preparing for how the bot will handle conversations outside its scope.

Our personal favorite is Cleo. She’s sassy, honest and just the kind of girl to help keep your spending habits in line. If you want something a bit more traditional, check out Hope.
Cleo chatbot

Meet Charlie, Belgium’s most helpful customer support chatbot

Meet Charlie, Belgium’s most helpful customer support chatbot

Argenta logo
Charlie is Argenta’s virtual assistant. She lives within the Argenta mobile banking app and helps answer customers’ questions and complete transactions. But what makes her so helpful? She only chats when she’s confident she can answer the question.
The problem

How can we make a chatbot that improves customer experience?

Argenta is the fifth-largest bank in Belgium and its customer support team receives 20,000+ messages per month. It was impossible for their team of 23 agents to keep up. They were overloaded and overworked.

During the discovery phase of the project, we identified three areas where the agents were losing the most time:
    1.Basic FAQs
    2.Basic intent detection and data capture for specific transactional flows
    3.Conditional data detection and capture for increasing daily limits

Argenta is known for its excellent customer service and wanted to ensure that customer satisfaction remained stable or improved when introducing the bot to its customers. With this specific challenge in mind, we introduced a few custom solutions that take Charlie from being a basic bot to a great bot.
The solution

Charlie is a chatbot with a clear scope and near-perfect NLP model

Charlie only chats when she’s confident she can help

Latching onto the concept that Charlie should maintain the status quo or improve customer satisfaction, we pivoted and approached her introduction differently: She’s a very respectable lady and only introduces herself when she’s at least 85% sure she has the right answer.

 Conversation where Charlie is 97% sure of the right answer

Conversation where Charlie is 62% sure of the right answer and immediately hands it off to a live agent to support
How is this possible? We developed a unique introduction and fallback system so that if she knows the answer, she shares it. If she doesn’t, the conversation gets routed to a live agent, and the customer has the same experience they would normally have pre-Charlie.

The only change for the customer is that if their question can be automated, they receive an instant answer instead of queueing for a live agent. It’s a win-win for the customer and the support team because both save time! The customer doesn’t waste time queueing for a live agent and the live agent doesn’t waste time answering a question that Charlie could easily automate.

Charlie has pre-defined areas of expertise

By not introducing herself unless she’s confident in the answer, Charlie doesn’t need to know everything. Together with the customer service team, we defined a clear scope of easy-to-answer, frequently asked questions. This translated into 30-40 intents covering approximately 15% of all incoming questions.

Beginning with a small set of intents enabled us to test, train and monitor Charlie’s natural language processing (NLP) model to near perfection. When we first introduced her to real customers, we reviewed every interaction to ensure that she said the right thing and created a net positive customer experience. After a few weeks of intense monitoring and some minor tweaks, it became clear that Charlie knew what she was talking about and was ready to work independently.

Charlie clarifies doubt instead of defaulting to “I don’t know”

The vocabulary associated with banking is complex and nuanced. Since many banking intents overlap with one another, we usually recommend working with entities, such as debit card, credit card, limit, etc. Using contextual entities, we built a unique entity fallback system. If Charlie’s between 70-85% sure of the intent, then she rephrases the intent as a question to confirm her suspicions. If she recognizes an entity, she responds with a custom menu that proposes the most asked intents related to that entity.

Conversation where Charlie is 74% a keyword (entity) in the user’s question

Conversation where Charlie is 75% sure of the intent and confirms her understanding with a question

Charlie automates data capture to save everyone time

Basic data capture is a time-sink for a live agent. It can take up to 10 min. for a user to respond with their name, address, email and phone number. That’s valuable time that an agent could use on more complex conversations.

One of customers’ most frequently requested items is a new card reader. The batteries expire over time and customers ask for a replacement. This question is easy for a bot to recognize, so instead of bothering a live agent with this simple request, Charlie detects when a user is asking for a new card reader and responds with a data capture flow that captures the user’s name, address, email, etc. Charlie then forwards that information and the new card reader request to a specific team within Argenta, where they check the request and prep the card reader for dispatch.

Converting a frequently requested manual process into a conversational flow saves Argenta time and provides users with an immediate solution

Another frequently asked question is how customers can increase their daily transactional limit. For certain amounts, it’s an easy fix that Charlie can automate. But for amounts that exceed Є25,000, the customer needs to answer some questions. Charlie can detect when a customer wants to increase their daily limit and can also detect if the amount exceeds Є25,000 so she can already get answers to the questions before handing the conversation off to a live agent to make the final limit adjustment.

Charlie helps a user adjust their limit to Є13,000

Charlie helps a user adjust their limit to Є30,000

The Results

Charlie is a chatbot loved by colleagues and customers

Charlie saved the customer support team 24 days (192 hours) of work in the first month and a half, and they’re grateful for her help. Independently, Charlie manages 20%+ of the incoming customer conversations end-to-end. 

Our new colleague Charlie is doing really well. Because of her quick reaction to simple questions, we can focus on the real work 😉

Ilke Schiltz, Customer Support Agent @ Argenta
The most triggered flow was increasing daily transactional limits, followed by requesting a new card and then a new card reader. As Charlie continues to engage with and help customers, we’ll broaden her banking knowledge so that she continues to save the customer support team time.

💾

24 DAYS

OF WORK SAVED IN THE FIRST 1,5 MONTHS

💪

+20%

CONVERSATIONS HANDLED FULLY BY CHARLIE

😊

+95%

CONTACT CENTER
CSAT

Other work that will spark your interest

Conversational AI Consultant

Conversational AI Consultant

Brussels | Full time

As a Conversational AI Consultant, you’ll work with the team to deliver conversational AI projects for our global clients. This job is hands-on and begins with an immersive bootcamp where you'll learn to master our methodology, processes and technical tools.
We’re looking for people who contribute to our good vibes and are eager to solve problems. So if you like to experiment with new technology and work directly with clients to answer their business needs, then you should definitely apply.

👨‍💻 What you'll do

  • Work directly with clients during scoping, briefing and implementation phases
  • Use conversational A.I. tools to build flows and NLP/intent models
  • Identify and analyze business needs to provide high-level budget estimates
  • Apply business logic and integrations to transform simple chatbots into personalized conversational agents

🕵️‍♂️ What we're looking for

  • 0-3 years of work experience
  • Prior consulting or agency experience is a plus
  • You know how to interact with corporate clients
  • You can manage deadlines
  • You have an analytical mind and the urge to create order when confronted with large data sets
  • You have a keen business sense
  • You have basic Javascript / HTML / CSS coding skills (We’re open to someone with limited prior experience but with a strong willingness and ability to teach themselves)
  • Knowledge of API integrations is a plus
  • You have some experience in user acceptance and business acceptance testing
  • You understand or at least have a strong interest in IT, data, operations, automation and the A.I. landscape
  • Fluent in English, spoken and written (other European languages are a plus)
  • You’re curious and a quick-learner

💸 What you'll receive

  • A competitive salary in line with your experience
  • Flexible working (50% remote, 50% office)
  • Modern mobility package of your choice
  • Learning and development package
  • Mentoring and support from 3 senior profiles with a combined 10+ years experience in conversational A.I.
  • The opportunity to grow with and help build a young, innovative company

Conversational AI Intern

Conversational AI Intern

Brussels | 3 to 6 months

As an intern, you’re at the beginning of your career and we recognize that you’re here to learn. We don’t expect you to have 3 years of industry experience, a PhD in data science or to have visited Mars. We're simply looking for someone who's eager to discover the world of conversational AI.
We’re looking for people who contribute to our good vibes and are eager to solve problems. So if you like to experiment with new technology and want to build solutions that answer business needs, then you should definitely apply.

👨‍💻 What you'll do

  • Use conversational AI tools to build flows and NLP/intent models
  • Identify and analyze business needs to provide high-level budget estimates
  • Apply business logic and integrations to transform simple chatbots into personalized conversational agents
  • Create integrations between conversational agents and back-end systems (databases, help desk software, CRMs, …) using (basic) APIs and scripts
  • Document learnings and share them with the team
  • 🕵️‍♂️ What we're looking for

  • You have an analytical mind and the urge to create order when confronted with large data sets
  • You have a keen business sense
  • You have basic Javascript / HTML / CSS coding skills (We’re open to someone with limited prior experience but with a strong willingness and ability to teach themselves)
  • Knowledge of API integrations is a plus
  • You understand or at least have a strong interest in IT, data, operations, automation and the AI landscape
  • Fluent in English, spoken and written (other European languages are a plus)
  • You’re curious and a quick-learner
  • 💸 What you'll receive

  • Experience working with AI technologies, especially NLP (natural language processing)
  • Mentoring and support from 4 senior profiles with a combined 15+ years’ experience in conversational AI
  • Access to conversational AI courses
  • The opportunity to grow with and help build a young, innovative company
  • Transport to the office and lunches are on us