5 Often-Overlooked Steps to Building a Useful Chatbot for Your Law Practice

If you’ve decided you’d like a law firm chatbot, but aren’t sure where to start, this step-by-step article will help your planning.

It is easy to get lost when you comparison shop chatbot platforms or dive into chatbot mechanics. So make sure you step back to plan what your law firm is trying to achieve and how best to get there.

There’s no theory in this post. It is just a nuts-and-bolts action plan (with a handy companion worksheet!) for creating a useful chatbot for your law practice.

Five step guide to plan a chatbot for your law practice

5 Steps to Building a Useful Chatbot for Your Law Office

First, take a simple guided tour through the steps you’ll need to plan and build a useful chatbot for your firm with a free “5 Steps” checklist.

Get it Now!

Why Build a Law Firm Chatbot?

Before we jump in, you might ask yourself, “Why build a chatbot for your law practice in the first place?” or “Do people really want to be chatting with disembodied chatbots?”

Well, here’s the scoop.

People say chatbots are the new apps. But in many ways, they are even better. When you consider the oversaturation of apps—and that you don’t need to download a chatbot, learn a new user interface, or worry about what operating system you’re using it on—it becomes obvious that chatbots eliminate troublesome user-adoption friction.

Consumer surveys show that chatbots also address modern consumer expectations.

picture of text box with text reading reaching desired outcome

When asked “What’s most important when communicating with a company?,” 68% of consumers responded that “Reaching desired outcome” was paramount, followed by “ease of experience” (48%), “speed” (44%), and “convenient time” (39%). (Ubisend 2017 Chatbot Report.)

Chatbots deliver these outcomes, in large part, because they operate 24/7/365. They also remove friction and—at least if they are thoughtfully designed to solve users’ real problems—do so immediately. In fact, 69% of consumers would consider talking to a chatbot over a human being because a chatbot can provide an instantaneous answer.

picture of law firm chatbot quote with statistic

Certainly, we must adapt to our profession the scope of what a law firm chatbot can provide your firm’s clients. A chatbot can’t engage in the practice of law, and it can’t guarantee results. But there are still many tasks a chatbot can automate—and information it can relay to clients or potential clients—in useful ways.

Step 1: What Is Your Chatbot’s Purpose?

Let’s start with your goal. What are you trying to achieve with this chatbot? Are you building a bot to capture more leads? Capture basic information from clients for intake or other reasons? Gather information to draft legal documents? Provide standardized information to clients when they reach case milestones?

Because a chatbot can automate conversations and simple tasks, it can be a very powerful tool. But like any tool, if misused, it can create problems.

For example, a chatbot shouldn’t be used to substitute for an initial client meeting. But it can lay the groundwork for a successful initial client meeting. Your bot can work on gathering basic facts while you carefully define your client’s real problem and propose a workable solution.

As with any goal, it’s a good idea to define some key metric to measure performance. Here are a couple basic examples to get you started:

  • Lead Conversion Rate = Total Number of Leads / Number of Website Visitors x 100.
  • Client Conversion Rate = Total Number of Retained Clients / Number of Leads x 100.

Step 2: Where Does Your Bot Live?

Your chatbot’s home is where it lives online. It can live on Facebook Messenger, Slack, WhatsApp, Skype, Alexa, or any number of other platforms. This is not to say that your chatbot cannot have more than one home. With most chatbot builders, you can launch your chatbot so it is available on more than one platform at the same time.

Choosing your chatbot’s home is important because where your it lives should depend on where your audience lives. You should select a home where your bot is more likely to connect with your intended audience.

picture of choices where to place a law firm chatbot

So, where is your audience? Are you trying to reach new business clients? If so, Slack might be a good home for your chatbot because 77% of the Fortune 100 use it. Do you want to connect with new immigration clients? WhatsApp boasts 98% non-US users, some of whom may be interested in immigrating to the United States. New family law clients? Facebook allows you to advertise directly to potential clients who have recently updated their relationship status. Brainstorm places that will maximize your bot’s potential engagement with your intended audience.

Step 3: What Is Your Bot’s Personality?

Although it may sound oxymoronic to talk about a legal chatbot and personality in the same sentence, it is absolutely an important consideration in building your chatbot.

For example, if your chatbot is for internal use only, its personality may not be important. You may prefer that it be more utilitarian in its demeanor. But let’s assume for our purposes that your chatbot will perform in front of consumers.

pictures of various emojis to illustrate law firm chatbot personalities

At first, you might wonder whether “giving your bot a personality” means you should make it silly or snarky. Maybe. But personality is more than that. Personality is a gestalt. It is the sum of components that interact uniquely to give a specific impression. Your law firm’s chatbot’s personality should help accomplish its purpose.

Your Chatbot’s personality is an equation:

Name + Visual Style + Backstory + Conversational Tone = Chatbot Personality.


I prefer a human name over a task-oriented name. For example, my first voicebot offered general divorce information. I called her ‘Larissa,’ after a friend’s legal assistant. Her name, in my mind anyway, created the perfect blend of a friendly person and super-competent legal assistant. I could have named it ‘DivorceBot,’ but anything ‘DivorceBot’ would have gained in clarity of task identification it would have lost to “Larissa” in personality.

Also, check whether your bot’s name is already used or identified with a similar service. No matter its personality, your chatty little friend won’t know how to respond to a cease and desist letter.

Visual Style

Now for the fun stuff. What does your chatbot look like?

Is your bot avatar a little robot? Does it have your managing partner’s face? Or maybe you’ve created a character that embodies a helpful (and hyper-responsive!) assistant. Whatever you decide, just be sure your chatbot’s face is consistent with your overall goal and reflects your firm’s overall branding.

Beyond the bot’s avatar, other design elements require your attention, too: the color palette (match your firm’s colors), typefaces, and design style (for example, don’t mix flat icon designs with drop-shadow designs).


What is your chatbot’s backstory? Taking this personality thing a little too far? Not really.

Is your chatbot an assistant, a paralegal, or a partner’s alter ego? Is this your chatbot’s first job? Do you force it to work 24/7? Does it get benefits? What are its aspirations? Is your chatbot happy working for the firm? These sound like silly questions, but they help to build a fuller character from which dialog will flow.

picture of linked in profile of billy bot law firm chatbot

Perhaps the most delightful example of chatbot personality with backstory is Billy Bot, who even has his own LinkedIn profile. Billy Bot is a “junior clerk” in London who live tweets from events that he “attends” and gives updates about his training and progress. And he has more than 1,500 Twitter followers!

Conversational Tone

It may be obvious but it’s worth emphasizing: a chatbot’s primary mode of communication is conversation.

Just as websites use visual design to cue a user about what action to take, a chatbot uses conversation. Conversational design is an art. And it takes practice, improvisation, and creativity. If someone in your office has the gift of gab, you may want to take notes to capture turns of phrase and emphasis, then incorporate what works into your bot’s dialogue.

Words matter. Choose words wisely so your bot conveys the right emotions. Change up responses where you can so the dialog will still feel fresh if the user has to do it again.

What is your chatbot’s conversational tone? Is he serious, academic, deadpan, and matter-of-fact? Or light, happy-go-lucky, and enthusiastic? Maybe she is deeply sympathetic, a touch sarcastic, or a bit aggressive. Whatever tone you choose, remember the tone should reflect the nature of the discussion. It could even change during the conversation if appropriate.

Word of Warning

Did I forget to say, “Don’t try to give people the impression that your chatbot is a real person?” You might inadvertently leave this impression by forgetting to tell users at the conversation’s start. Or maybe you might think people will be more likely to use the chatbot if they think a real person is behind it? All I can say is: don’t. Don’t try to trick people into thinking that your bot is a real person.

Why not? First, with the state of technology right now, the user will almost certainly be seriously disappointed. Even with machine learning and neural networks—which are not typically incorporated into most chatbot platforms—a natural conversation about any topic (even the most basic) is difficult to pull off without it feeling artificial and leading to errors. Second, most attorney-client communications are not free-ranging discussions about politics, religion, or fashion. They are typically goal-oriented. So, it’s entirely consistent for a legal chatbot to be directed toward achieving a specific result. A bot should be useful above all.

Step 4: What Is Your Chatbot’s Conversation Structure?

We have conversations everyday. But, like walking, people converse intuitively and don’t typically “plan” their conversations and repartee. I say “most people” because, unlike most people, lawyers are trained in the art of listening, questioning, and communicating. Didn’t think your Lawyering Skills class would come in handy today? Well, it does when designing the conversation your chatbot will have with a user (read: an actual or potential client).

When we do a client intake interview or take a deposition, we know there is a difference between open-ended questions, leading questions, and closed questions. Similarly, there is a way to craft chatbot conversation to guide users down a defined path and to reach a specific goal.


Before you start designing your chatbot’s conversation, you need to do a little preparation.

First, you need to answer some fundamental questions. Who is your audience (Step 2)? If your intended user is an existing client, you will have different conversations than if they are a potential client. What is your user trying to do(Step 1)? What information is necessary for them to do it?

Do you have any written materials from which you can draw? For example, if you have a potential-new-client script you provide to new staff members so they can screen calls effectively, efficiently, and politely, you already have some wonderful content to draw from. Don’t be afraid to repurpose existing procedures and scripts.

Diagram Your Dialog Tree

Now that you understand your audience and their goals, you should map how these conversations will unfold. Lucid Chart and Gliffy are a couple of great programs for diagramming dialog trees. Both are paid services that offer free trials.

Conversations don’t always meander in the direction you expect. A diagram helps you identify dead ends and problem areas. Then you can create graceful ways to redirect your user back to the right path.

Like any good story, your conversation must have a beginning, middle, and end.

A Greeting, and Managing Expectations

Use your greeting as an opportunity to welcome your user and make clear what your chatbot can and can’t do. Give them an overview of the process and its goals and explain some of your chatbot’s functions. For example, explain how to restart the bot if the user runs into trouble and how to ask for help from a real-live person.

This is also the time to make sure you explain that your user is talking to a bot, not a human being.

Glide Path to Goal

The conversation should guide the user toward a goal. You should avoid open-ended questions unless absolutely necessary and, even then, they should only capture information from the user, not change the conversation’s direction. I prefer closed questions with a limited range of answers.

So-called “quick replies” are great for keeping the conversation on track. You bot uses a quick reply when it asks a question and presents pre-defined answers, like ‘yes’ or ‘no,’ or multiple choice answers, like ‘corporation’, ‘limited partnership.’ or ‘sole proprietorship.’

And quick replies have an unexpected benefit, too: because the user understands she is conversing with a computer—and that the goal of the conversation with a computer is to advance her case—she is more likely to comply with the conversation’s limited scope. As we all well know, real-time conversations with real-live clients and their human(-ish) lawyers sometimes get sidetracked. When they do, the data gathering needed to analyze an acute legal problem can be lost behind a client preferring to tell a story or share less-relevant facts.

In a weird way, chatbots—particularly law firm chatbots—can accomplish what we always wish we could: creating a glide path from which clients cannot deviate.

Goal Achieved

Once the user has completed the conversation with your chatbot, it’s time to thank him for his patience with the conversation and to provide him with a deliverable or a call to action. A deliverable can take many forms: scheduling an appointment, providing a document or checklist, or creating a report, letter, or completed form. It gives a sense of closure and provides a tangible benefit for participating. Also, be sure your chatbot explains next steps (if there are any) so the user will know what more needs to be done.

Error Handling

“I didn’t get that” gets old. Fast. Instead, use your chatbot’s error-handling responses as an opportunity to keep things light and show some personality. Instead of “I didn’t get that,” why not “Oops, I didn’t see that coming! Can we try that again?” Add a few variations on the theme to keep it fresh.

Step 5: What Tools Will You Use to Build Your Bot?

In my last post, 2 Chatbot Platforms for Lawyers, No Coding Required, I explored how to build a chatbot with user-friendly chatbot building platforms.

Not all chatbot builders are created equal. Some are visually-oriented while others are clearly geared for coders. In general, the more “accessible” chatbot building platforms are less powerful than those developers favor. The platform you choose can limit what your chatbot can do. But that isn’t necessarily a bad thing.

Most of the chatbot builders below are no-coding affairs. But I have included some more advanced platforms, too. I’ve listed some features for each, as outlined in more detail below:

  1. Free or Paid: Is the service is free or paid? If paid, what terms apply?
  2. Coding Required: Does the chatbot require coding skills?
  3. Publishing Platforms: Where will your chatbot live online? Your website? Facebook Messenger? Somewhere else?
  4. Artificial Intelligence: Does the platform offer artificial intelligence—like natural-language processing, natural-language understanding, or machine learning—that allows your chatbot to interpret language and respond intelligently?
  5. 3rd-Party Integrations: Will your chatbot integrate with 3rd-party services like Gmail, MailChimp, Office 365, etc.?
  6. Supported Languages: Does the chatbot builder support foreign languages?
  7. Recommend Use: I give a recommendation about how to best to use the platform.

Here is an overview of 7 chatbot building platforms, starting with FlowXO.


FlowXO is a visual builder suitable for both beginners and proficient users. It has some useful conversational templates that can help you get started easily. Although FlowXO is light on AI, it makes up for it by integrating with over one hundred external services, including some that let you integrate AI into the mix.

  1. Free or Paid: Both. Free plan for 500 interactions, for up to 5 bots or active flows; $19/month for up to 15 bots and 5,000 interactions.
  2. Coding Required: No. FlowXO is a visual chatbot builder that requires no coding skills.
  3. Publishing Platforms: Website, Facebook Messenger, Slack, Telegram, and Twilio SMS.
  4. Artificial Intelligence: None, but FlowXO can integrate with natural language processing platforms like Dialogflow or Wit.ai (summarized later) to interpret language and respond intelligently.
  5. 3rd-Party Integrations: Over 100 plug-and-play integrations with 3rd-party services like Gmail, Mailchimp, Infusionsoft, Office 365, Stripe, and Trello.
  6. Supported Languages: 17 languages, including English, Spanish, French, Italian, Chinese, Japanese, Korean, Greek, Turkish, and Russian.
  7. Recommended Use: Novice to Proficient. FlowXO is best suited for a beginner looking to get started with chatbots and has plenty of room to grow. Just remember that it will take many hours to explore and master all of the features and hundreds of integrations.

Dialogflow (formerly API.ai)

Google bought API.ai and re-named it Dialogflow. Although not technically a chatbot builder (it’s intended to be a natural language processing platform for text and voice bots, like Google Assistant and Alexa), it does a pretty good job of creating standalone bots.

  1. Free or Paid: Free.
  2. Coding Required: It depends. You can build a basic bot with no coding required. Coding skills are necessary for advanced functionality.
  3. Publishing Platforms: Website, Facebook Messenger, Slack, Viber, Twitter, Skype, Line, Kik, Google Assistant, Alexa, and Twilio SMS.
  4. Artificial Intelligence: Very Good. Powerful natural language processing and machine learning. Dialogflow employs an intents-and-entities system to understand words you train it to understand, extract information, and provide relevant responses.
  5. 3rd-Party Integrations: Yes. Dialogflow does not support plug-and-play integrations, but it does allow you to integrate 3rd-party services using webhooks and an inline code editor to host and execute code through Cloud Functions for Firebase.
  6. Supported Languages: 15 languages, including English, Spanish, French, Italian, Chinese, Japanese, Korean and, Russian.
  7. Recommended Use: Proficient to Advanced. Dialogflow is best used with a chatbot you create on another platform to process users’ natural language responses. With Google’s massive data set, Dialogflow’s ability to understand just gets better and better.


Manychat’s primary strength is the powerful marketing features, like sequences. Sequences are series of messages that ManyChat will send automatically over time. You can set the message delays from several minutes to several weeks, and use them to nurture leads or provide materials in multiple steps.

  1. Free or Paid: Both. Free for limited use of growth tools and message sequencing (similar to an email autoresponder) with Manychat branding. The paid plan provides unlimited access to growth tools and message sequencing and removes Manychat’s branding.
  2. Coding Required: No. Manychat is a visual chatbot builder that requires no coding skills. It’s actually one of the more user-friendly platforms I’ve tested.
  3. Publishing Platforms: Facebook Messenger.
  4. Artificial Intelligence: Limited. Manychat can essentially link keywords to other phrases, but that’s about it. No natural language processing.
  5. 3rd-Party Integrations: None.
  6. Supported Languages: English.
  7. Recommended Use: Novice to Proficient. Easy for a beginner to get started with chatbots. Manychat is best used to improve your visibility on Facebook and Messenger, to grow a following, and to keep top of mind with relevant content.

QnA Maker

QnA Maker isn’t really a chatbot-building platform. It makes it super simple to get a simple FAQ service up and running quickly. You input your FAQ page’s URL and QnA Maker automatically formats the questions and answers into pairs that can be published in the form of a chatbot.

  1. Free or Paid: Free.
  2. Coding Required: A little bit. QnA Maker creates an service that is accessible through its API. You still have to have to create a bot on Microsoft’s Azure cloud service to be able to publicly access the FAQ service you just created.
  3. Publishing Platforms: Website.
  4. Artificial Intelligence: Limited. QnA Maker uses basic natural language processing to connect user queries with the appropriate answer.
  5. 3rd-Party Integrations: None.
  6. Supported Languages: English.
  7. Recommended Use: Novice. QnA Maker is intended for a novice to get a taste of the potential for chatbots. I think it’s a good idea for everyone to try out QnA Maker because it also eases you into the more technical aspects of APIs, webhooks, and more advanced chatbot frameworks.

IBM Watson

You probably remember hearing about Watson when it trounced accomplished Jeopardy players. Watson, developed by IBM, is a system capable of answering questions posed in natural language. Watson was named after IBM’s first CEO, Thomas J. Watson, and it is one of the more powerful chatbot building platforms.

  1. Free or Paid: Free for up to 10,000 API calls per month, then $0.0025/call.
  2. Coding Required: Coding required. Although it is technically possible to launch a Watson chatbot on Facebook without coding, the process requires configuring dependencies in json, generating tokens, grabbing information from the developer section of Facebook, and having a general understanding of how APIs work. If this sounds like Greek to you, you’ll have to hire a developer.
  3. Publishing Platforms: You can publish your chatbot on any platform in which you can code an integration.
  4. Artificial Intelligence: Strong. Like Dialogflow, Watson (available on IBM’s Cloud, formerly IBM Bluemix) employs an intents-and-entities system to understand words you train it to understand, extract information, and provide relevant responses. With Watson you also enjoy an extensive list of services to augment your bot, like a Tone Analyzer, Language Translator, Personality Insights, and Speech to Text.
  5. 3rd-Party Integrations: As many as you can code.
  6. Supported Languages: English, Spanish, French, German, Brazilian Portuguese, Japanese, and Simplified Chinese.
  7. Recommended Use: Advanced. Watson is definitely for more-advanced users. That said, it is definitely worth the extra effort if you have time to train yourself or hire a developer. I would use Watson if I needed to navigate complex, multi-domain conversations, to analyze documents, or to create an e-discovery service.


Chatfuel is fun to use and is one of the more popular visually-intuitive chatbot builders with no coding required. Its claim to fame is that it allows you caton create a Facebook Messenger chatbot in 10 minutes or less. If you’ve read my other articles or have heard me speak, you know I highly recommend it as a gateway chatbot builder.

  1. Free or Paid: Both. Like Manychat, Chatfuel is free. If you want to remove Chatfuel’s branding, it will cost you $30/month/chatbot.
  2. Coding Required: Chatfuel is a visual chatbot builder that requires no coding skills.
  3. Publishing Platforms: Facebook Messenger.
  4. Artificial Intelligence: Limited. Chatfuel can essentially link keywords to other phrases, but that’s about it. No natural language processing.
  5. 3rd-Party Integrations: About 20 plug-and-play integrations, including Slack, Google Calendar, Twitter, Instagram, Youtube, and many more with Zapier and a JSON API.
  6. Supported Languages: 50 languages, including Chinese, English, French, German, Greek, Hebrew, Italian, Japanese, Korean, Persian, Russian, Spanish, and Tagalog.
  7. Recommended Use: Novice. I highly recommend Chatfuel for the novice looking to dip a toe into the chatbot space. Although its actual artificial intelligence is rudimentary, it can automate routine conversations and you can extend its functionality with some great 3rd-party services. Chatfuel is a great place to start.


Wit.ai, which Facebook acquired and which provides the AI engine for Facebook Messenger-based chatbots, is a free service. It uses a simple natural language processing platform and APIs for developers to build chatbot applications.

  1. Free or Paid: Free.
  2. Coding Required: Yes.
  3. Publishing Platforms: You can publish your chatbot on any platform in which you can code an integration.
  4. Artificial Intelligence: Very Good. Like Dialogflow and Watson, Wit.ai uses an intents-and-entities system to understand words you train it to understand, extract information, and provide relevant responses. What makes Wit.ai a little different is that it adds ‘stories,’ which allow you to provide examples of conversations so it can learn from those examples and provide a smoother conversational experience.
  5. 3rd-Party Integrations: As many as you can code.
  6. Supported Languages: 50 languages.
  7. Recommended Use: Advanced. For advanced users only.

For more information on other chatbot building platforms, here is comparative table of 25 chatbot builder platforms. (Olga Davydova, 25 Chatbot Platforms: A Comparative Table)


I know, I know. That was a load of information, and you probably feel a bit overwhelmed. To be sure, building a chatbot is not a project to be undertaken lightly. It requires not-insignificant supplies of diligence, effort, and research. But, if you follow the steps I’ve outlined, you’ll be far ahead in building a useful chatbot for your law practice.

Remember to ask yourself these questions:

  • What goal are you trying to achieve?
  • What audience are you trying to reach?
  • What is your chatbot’s personality?
  • What is your chatbot’s conversational structure?
  • What tools will you use to build your chatbot?

Like a brilliantly-architected chatbot conversational structure, I hope this 5-step plan helps you find your direction. And, if you ever get lost, I’ll be here to redirect you.

Tom Martin
Tom Martin is the founder of LawDroid, a chatbot development and consulting company for the legal industry. Tom’s company focuses on making useful bots to help lawyers capture more leads, intake more clients, and handle more cases. Tom is an alum of Lawyerist’s inaugural TBD Law conference and a 2017 Fastcase 50 honoree. He is also a co-founder of Vancouver Legal Hackers and advisor to A2J Tech Fellows. Tom has written and spoken about the evolving AI and bot space and the implications for the legal industry. LawDroid was recently awarded a contract, funded by Legal Services Corporation, to build the US’s first voice bot to increase access to justice. Tom will be speaking in March 2018 at the British Legal Technology Forum in London. You can follow Tom on Twitter @lawdroid1 or read his blog at LawDroid.