- Decide whether a chatbot is right for you
- Gather requirements
- Involve stakeholders
- Identify what type of chatbot you need
- Choose a platform
- Carry out fundamental conversation design steps
- Decide on how you plan to collect feedback, track behaviour and analyse data.
Decide whether a chatbot is right for you
There’s been a lot of hype in recent years. It doesn’t help with all the major market analysts (Looking at you Gartner and Forrester) all shouting how 80% of businesses will adopt chatbots by 2020 (at the time of writing: May 2019, we’re at roughly 60%! – will we make it?!)
Consequently, we’ve personally had numerous clients who bitten in fear of missing out only for us to help them find out – a chatbot isn’t the right solution for the pain they’re looking to solve.
Truth be told, chatbots were fabled to be the death of apps. Instead, we’ve found that a hybrid approach is likely to be the future.
So make sure to double-check before progressing any further: ask yourself and your stakeholders:
“Do we really need a chatbot?”
To help you answer this question, we’ve put together a cheat sheet:
Go through the criteria below to see if using a chatbot is the right medium for solving the user challenge.
If you tick most of the boxes, it’s likely a chatbot can help with the pain you’re solving.
| Check each box if true: | Pro of conversation | Con of conversation |
| Users are happy to talk or type about this pain:Users already have person to person conversations about pain or feature | Conversations let users tackle the pain in a way that suits them. | Text & spoken conversations don’t always do well in public spaces (the train) or shared spaces (such as Slack) |
| Users will want multiple ways of accessing to solving their pain through different channels & mediums. | Conversations can be transferred between different channels and picked up after interruptions without necessarily interfering with the user experience or flow. | Large investment usually required for Omnichannel. A lot of back end logistics needed and careful data privacy to navigate to achieve such user experiences. |
| The pain or feature is difficult or clunky to findUsers might need to navigate multiple apps or widgetsUsers might need to navigate multiple devices | Conversation can save the user more time than normal user interfaces. Conversations are intuitive and thus can be the best shortcut. | A picture speaks a thousand words and conversations alone may not be enough. |
| Users want to do this when they could be interrupted or while multitasking Users can do this while their eyes are busy | Conversations can be done when users are multitasking. Ever typed a text while looking at someone else? | Conversations don’t always suite tasks that require focus or take some time to achieve as they can turn laborious without some shift in pace, cadence or experience. |
Develop a chatbot strategy
Once you’ve qualified a chatbot as the right solution, you need to start creating a high-level overview and roadmap for your chatbot strategy. We cover the 10 steps in building a successful chatbot strategy here.
Some key takeaways from the article:
- What is the competition doing? What unique selling point does your brand have to assist customers via chatbot technology?
- Ensure to build a strategy for the entire customer journey, from the discovery of the pain, taking action to fix it and maintenance. Where does the chatbot fit along this journey?
Involve stakeholders
From our experience, many clients come in with too high expectations of chatbot capability vs the cost involved to deliver such a chatbot. Expectations of cost, technical capability and timeliness will certainly vary and it’s important to keep stakeholders involved in terms of what’s changed.
Many projects fail or even fail to get off the ground due to poor communication with stakeholders.
Carry out fundamental conversation design steps
What is conversation design?
Conversation design is a method of designing something based on how humans share information. The advantage of using conversation design when solving problems is that it’s very intuitive for people to use it since language is second nature to humans!
We have a full article going into the depth steps of conversation design. In quick summary we’ll briefly explain each of these and the importance to the overall chatbot design process:
Gather requirements and identify the use case
The first step of the chatbot design process is gathering requirements. This is all about asking the right questions and gathering all the information and answers you have in one place ready to process and utilise.
To name a few key questions:
- Who are your users?
- What are their needs?
- How are they completing these tasks today?
- What does the before and after state look like if you use a chatbot to solve the problem?
- How do they talk and text?
As you identify your key use cases, ensure you keep in mind that the use case is achievable within the remit of your resources such as:
- Technical limitations
- Level of effort
- Timeline & budget
- Continuous support
Decide what type of chatbot you need
There are 3 types of chatbot
Logic only? Hybrid or pure ML?
Create a chatbot persona
This is where I personally have the most fun during the conversation design process. Creating the chatbot persona.
A chatbot persona is what gives the conversational assistant some character. Many make the mistake of thinking you should design your chatbot to feel more human – but that doesn’t always work effectively.
Fortunately, this step is fun because it’s also mandatory. If you don’t properly design your chatbot so that talks your user’s language and empathises with their character, it will sink faster than a lead balloon.
We run through the steps and process in detail in our conversation design article.
Create conversation flows
Conversation flow is where you draw out all the conversations you expect your chatbot to have with your user.
This is normally created in a fast and agile manner where your first draft would be quickly tested via low fidelity or “wizard of oz” testing.
We do this to ensure that the flow designed makes sense, feels natural. Meanwhile, we also aim to achieve the goal in the most efficient manner.
Depending on the type of chatbot you need (whether it’s logic only or hybrid), you will have to adapt your conversation methods to either linear and non-linear.
Compare and choose a platform
What platform you use can affect what design techniques you use and the overall capability of your chatbot.
For example if you want to have your bot chat on Facebook Messenger – it has a quicky reply character limit on it’s response button templates. It also has word limits and restricts other design aspects such as colour, button sizes and font.
This can vary to more flexible channels such as your own custom web widget or on an App – where you have much more creative freedom and generally speaking, a better-designed bot.
Advanced conversation designers will be aware of the limitations of each messaging channel and NLU platform. Or at the least can reference to the documents quickly to check on what’s possible.
Architecture
Your business architecture will also affect the platform you choose.
There are a ton of considerations to make when picking the chatbot platform and designing the architecture:
- How will users be identified? Across different channels and sessions?
- How and where will their progress be saved?
- Do you have outdated back end systems that you need to integrate into?
- What CRMs and databases do you need to connect to?
- Does the chatbot need to operate on multiple channels or device?
Data
Typing in closely with the chatbot platform considerations is how you plan to manage and store data.
- What information is available? (e.g., titles, descriptions, dates & times, topics)
- What’s the format of the session information? Is it plain text, audio, or other?
- If the content is plain text, was it written to be seen or to be heard?
- How long is it? Or how long does it take to read?
Does the data need to be need to be anomynised?
Your business will likely need to anoymise data in many use cases due to something called Personally Identifable Information (PII).
Personally identifiable information (PII) is any data that could potentially identify a specific individual.
PII can be sensitive or non-sensitive. Non-sensitive PII is information that can be transmitted in an unencrypted form without resulting in harm to the individual. Non-sensitive PII can be easily gathered from public records, phone books, corporate directories and websites.
Sensitive PII is information which, when disclosed, could result in harm to the individual whose privacy has been breached. Sensitive PII should therefore be encrypted in transit and when data is at rest. Such information includes biometric information, medical information, personally identifiable financial information (PIFI) and unique identifiers such as passport or Social Security numbers.
Removing PII is usually done with something called a message broker before it gets anywhere close to your databases, cloud storage or software platforms.
Related Questions
What is involved once you’ve done the design and you need to implement it?
Great question! We write all about the implementation process here. In a summary the key steps involved are:
- Integrating your chosen NLP tool or platform to your backend systems and other software
- Deciding on your hosting requirements
- Data privacy and security
- Designing your deployment management and processes
- Any legal processes or requirements.
