The Rise of Digital Humans
In recent years, digital humans have become increasingly popular in various industries, from entertainment to customer service.
There have been rapid advances in digital human platforms which offer increasingly lifelike capabilities and rich functionality. So it's easier to create a digital human with current technologies but like any conversational AI project creating a successful digital human for your business takes time and effort.
Creating a digital human requires a thorough understanding of the brand, users, and the digital human's role and tasks. All humans are different and so are the digital ones!
This blog post will guide you through the process of crafting the persona of your digital human, helping you understand the key elements that make a digital human effective, and highlight the pitfalls to look out for.
1. The Digital Human's Job Function
Digital humans have the potential to revolutionise the way businesses interact with their customers.
New roles are emerging rapidly, but here are some of the most common current applications of digital humans in business:
- Customer Support: Digital humans can be used to provide 24/7 customer support, answering questions and resolving issues for customers. This can help improve response times, increase customer satisfaction, and reduce the workload for human customer support staff.
- Sales and Marketing: Digital humans can be used to promote products and services, answer questions about the company and its offerings, and provide information about promotions and discounts. This can help increase brand awareness and drive sales.
- Employee Training: Digital humans can be used to provide training and onboarding for new employees, helping to ensure that they are up-to-date with company policies and procedures.
- Event Management: Digital humans can be used to manage events, such as trade shows and conferences, providing information about the event, answering questions, and directing attendees to the right location.
- Virtual Receptionist: Digital humans can be used as virtual receptionists, greeting visitors and directing them to the appropriate person or location. This can help improve the efficiency of the reception area and provide a more professional image for the company.
2. The brand your digital human will be representing
The first step in creating a digital human is to understand the brand.
Digital humans are not just computer-generated characters, but also an extension of the brand they represent. Therefore, it's vital to have a clear understanding of the brand's values, personality, and tone of voice.
This will help ensure that the digital human accurately represents the brand and aligns with its overall message and use case (or role that you want the digital human to fulfil).
It is also important to consider the brand's target audience.
Understanding the demographics, interests, and behaviours of the target audience will help make sure that the digital human appeals to them and is relevant to their needs. For example, if the target audience is young and tech-savvy, it might be appropriate to create a digital human with a more modern, hip, and casual personality... if the target audience is older and more traditional, a more formal and professional persona could be more suitable.
3. The users your digital human will be interacting with
The next step in creating a digital human is to understand the users.
This means understanding the needs and expectations of the users, as well as their motivations for interacting with your digital human.
For example, users could interact with a digital human for entertainment, information, customer service or even an onboarding or training role.
Understanding these motivations will help to ensure that the digital human is designed to meet the users' needs and provide a positive and engaging experience.
Also, keep in mind the context in which the users will interact with the digital human. This includes the platform, device, and location, as well as the users' overall state of mind and purpose. For example, users might be interacting with your digital human in a relaxed, leisurely setting... but they might also need to talk in a more serious and focused setting, such as a customer service interaction.
Understanding this context will help to ensure that the digital human's persona is appropriate and relevant to the situation.
4. Where the digital human fits into the customer journey
The role and tasks of the digital human are critical in determining its persona.
The user journey(s) that you will need to support will determine your digital human's skills, knowledge, and overall personality.
For example, if the digital human is designed to provide customer service, it should have excellent communication skills, an empathetic personality, and a deep understanding of the brand and its products or services.
If the digital human is designed for entertainment, it may have a more playful and creative personality, with an emphasis on humour and engagement.
You should also consider the specific tasks that your digital human needs to perform.
For example, if it's designed to provide information, it should have a comprehensive understanding of the topic and be able to answer questions and provide recommendations in a clear and concise manner.
If the digital human is designed for entertainment, it may be able to perform a variety of activities, such as telling jokes, singing, or dancing.
Understanding the specific tasks will help to ensure that when it comes to the implementation phase you create your digital human is equipped with the necessary skills and knowledge to perform them effectively.
Tips. Insight. Offers. Are You In?
5. Choosing a Digital Human Platform
When it comes to creating a digital human for your company, there are several platform choices available, e.g. Uneeq, Trulience, and SitePal. These platforms offer different levels of customization, complexity, and cost.
Most platforms allow the flexibility to choose different voices, genders, age groups, and different levels of customisation, e.g. appearance, voice, and personality. This might also include flexibility around things like skin colour, hairstyle, clothing, and accessories.
The voice of your avatar
The text-to-speech(TTS) capabilities of the platforms are often provided by big-name providers, such as Google or Microsoft. This tends to mean that there is a good choice of voices to match your persona.
The face or body of your avatar
Platforms tend to provide a number of faces or body types to choose from. There may also be options for whether you are opting for a head and shoulders or full body style.
Depending on the platform and your budget, you can also have your own digital avatar created which will match your brand and persona perfectly.
Other platform considerations
Other factors to look for are emotional state management capabilities, channel integrations and analytics, as well as the quality of the documentation and support services. There is a surprising amount to consider.
6. Conversation design for your digital human
Without a voice, your human is not going to be great to talk to! Many digital human platform providers offer out-of-the-box chat but to achieve a truly memorable experience you are going to need to create a conversational service to use with your digital human.
Designing this experience is a vital part of your digital human creation. Unless you are going to hand the whole conversation over to a Large Language Model (LLM) powered service, which is possible... but for the purposes of this post we are sticking to an intent-based approach.
Conversation design for your digital human involves defining the types of interactions that the digital human should be able to engage in, as well as the responses it should provide. If you want to learn more about conversation design, here are some useful books.
Effective conversation design should take into account the brand identity, target audience, and tone of the digital human, as well as the types of user inputs it is likely to receive (more on that below).
Your goal is to create a conversational flow that is clear, intuitive, and provides a positive user experience.
Unless you're going to use an LLM-powered chat service that handles your chat capability you will need to design at least some of your conversations with an intent-based approach.
This requires careful planning and scripting of the digital human's responses, as well as testing and refining the conversation design to ensure that it is effective.
We recommend using software to help you design your conversations for your digital human: anything which supports flow chart/diagramming can work. But, the best results come from specific conversation design tools, e.g. fabble or Voiceflow (a Bot Forge favourite), which offer a range of tools to streamline the design process. These include SSML tag creation and voice response testing... you can even export your designs to your conversational AI platform of choice.
The key takeaway here is:
design for every exchange to fail by being smart with graceful fallbacks and always including a means to repair a conversation that as gone off track
Creating the Natural Language Understanding (NLU)
Once you have your design, the next step is to create your chat service.
Most of the platforms support integrations with the common conversational AI platforms, with many streamlining the process of integration by providing extensive SDKs and developer examples.
This means integrating with the likes of Lex, Watson, Dialogflow, Cognigy or any other service which exposes its chat service as an API is feasible.
The Natural Language Understanding (NLU) of a digital human is the key to its success. It's the component that helps the digital human understand and respond to user input in a human-like manner. It determines the digital human's ability to understand user intent, extract relevant information from user inputs, and generate appropriate responses.
Let's explore the different aspects of creating a successful NLU for a digital human, including training your model, and creating the right responses and tone for your digital human using techniques such as SSML and platform behaviours.
A) Training your model
We've covered training your NLU model in other posts so it's beyond the scope of this guide to go into too much detail, but it's worth a quick run-through as it's a critical step.
The NLU model on whatever platform you're using is responsible for understanding user inputs and generating appropriate responses, so it is important to ensure that it's trained effectively.
To train the NLU model, you'll need to provide a large dataset of user inputs and corresponding responses. This dataset should be diverse, representing a wide range of user intents and scenarios.
B) What if I don't have any training data?
If you don't have training data then you have 2 options:
- Real users (the best option): Plan your intents and release your experience to a subset of "friendly users" as soon as possible and use the transcripts to train your model
- Use AI to create your training data: In the world of ChatGPT and GPT-3 (other large language models are available) it's possible to create some pretty amazing training data. Some conversation design platforms and conversational AI platforms provide this sort of feature out of the box, e.g. Voiceflow.
The training process can be time-consuming and definitely requires careful attention to detail, but it is essential for creating a digital human that provides a positive user experience.
Invest in effective NLU model training, and you will ensure that your digital human is capable of understanding its audience, and it can provide accurate and relevant responses.
C) Creating Responses
By this point, you should have your conversation design, and the other key details for your digital human's persona. Hopefully, you know who your digital human is and what they need to understand and do.
Crafting your responses is where you can really enrich your human, helping to create an appropriate and potentially mind-blowing persona.
Ensuring specific language and intonation helps to reinforce the personality traits in the conversation.
This process requires writing scripts for each interaction and response, taking into account the brand identity, target audience, and tone of the digital human. The scripts should be written in a way that is clear, concise, and easy for the user to understand.
In addition to telling your digital human how to communicate, your conversational service can also send commands that can control speech and control other aspects of the digital human's behaviour (depending on the capabilities of the platform itself).
D) Using SSML
Speech Synthesis Markup Language (SSML) is a markup language that provides a standard way to describe the pronunciation, volume, pitch, and other aspects of speech.
SSML can be used to enhance the voice and speech of a digital human, making it sound more natural and human-like. For example, SSML can be used to add pauses, emphasis, and prosody to the digital human's speech, making it sound more like a real human.
SSML can also be used to improve accessibility for users with disabilities. SSML can be used to add alternative text to speech, making the digital human accessible to users who are blind or have low vision. This can also be useful for users who are deaf or hard of hearing, as SSML can be used to add captions or subtitles to the digital human's speech.
SSML will become a large part of any voice assistant's responses. We find practically any response will benefit from some SSML to help them provide a believable, lifelike, and engaging experience.
E) Platform Behaviors
By using behaviour tags in your response dialogue, you have the ability to influence the emotional state of your digital human and how they interact while speaking as well as change the environment they live in.
Different platforms have different user interfaces, interaction styles, and input methods. Some examples of behaviour could be happy, sad, disapproving, agreeing, enquiring, and excited. You could even use them to make your avatar look in different directions.
Once you know the syntax of the tags needed, you can create your responses to leverage these platform-specific behaviours.
F) It's all about personality
By considering the right responses, tone, and platform behaviours, it's possible to create a digital human that effectively understands and interacts with its target audience as planned.
Techniques such as SSML and platform behaviours can help enhance the voice and speech of the digital human, making it sound more natural and human-like by helping its personality shine.
7. Challenges & Common Pitfalls
Digital humans have the potential to revolutionize the way businesses interact with their customers, but they also come with a unique set of challenges.
In order to ensure a positive user experience, it is important to understand the potential problems that users may encounter once the digital human has been implemented. In this section, we'll visit some of the most common problems that users may face when interacting with digital humans.
Lack of understanding
We've found users naturally try to talk to a digital human.
Depending on where your digital human lives, the most common problem we've found is inconsistent performance over the microphone when users are interacting with a digital human over web interface. This can be caused by a number of factors:
- Noise interference: Voice input can be disrupted by background noise, making it difficult for the device to accurately recognize the user's commands.
- Accented speech: Voice input systems may struggle to recognize speech from users with strong accents or those who speak a different language.
- Ambiguity: Voice input can be ambiguous, as the same words can have different meanings depending on the context. This can lead to incorrect responses or misunderstandings.
- Technical limitations: NLU input systems may struggle to recognise certain words or phrases because they haven't been part of the training.
One of the biggest challenges in creating digital humans is ensuring consistent responses.
This requires careful planning and scripting of the digital human's responses, taking into account the brand identity, target audience, and tone of the digital human.
If the digital human's responses are inconsistent or do not align with the brand's personality and tone, it can lead to confusion and frustration for the user.
Another common problem that users may encounter is limited functionality.
If the digital human is not capable of handling a wide range of tasks and interactions, it makes it difficult to provide the information and support that users need. This can result in users feeling frustrated and unsatisfied with the digital human's performance.
Inevitably, technical issues can also be a major problem for users when interacting with digital humans.
This includes issues with the platform or device that the digital human is deployed on, as well as issues with the digital human's programming or implementation. Technical issues can cause the digital human to malfunction or behave in unexpected ways, resulting in a user experience you'll remember for all the wrong reasons.
Conversational AI service speed can also be a factor here. If there's a lag on the round trip to the conversational AI service this can really affect the digital human's ability to reply naturally.
Lack of personalisation
If your digital human is not able to respond to user input in a personalised manner, it may not be able to build a strong connection with the user.
This can result in a lack of engagement and a poor user experience.
A lot can go wrong
Digital humans have the potential to provide an innovative and engaging way for businesses to interact with their customers. Understanding the potential problems is your most important weapon in the battle against poor user experiences.
Tackle these challenges head-on in the planning and testing phases, and keep an eye on them by monitoring your human regularly. If you're after a seamless user experience, constant improvement and regular iterations need to be your friends.
In conclusion, crafting a successful digital human persona requires a deep understanding of the brand, target audience, and the role of the digital human.
By considering these important factors, businesses can create a digital human that effectively represents their brand, connects with their target audience, and provides valuable support and information to users.
Creating a Natural Language Understanding (NLU) for the digital human is a critical aspect of crafting its persona. By considering the right responses, tone, and platform behaviours, businesses can create a digital human that effectively understands and interacts with its target audience. Techniques such as SSML and platform behaviours can help enhance the voice and speech of the digital human, making it sound more natural and human-like.
However, it is important to be aware of the potential problems that users may encounter when interacting with digital humans. These include inconsistent responses, limited functionality, technical issues, and a lack of personalisation. Take these challenges on and you can create a digital human that provides a positive and seamless experience for users.
Digital humans have a wide range of potential applications in business, from customer support and sales and marketing to employee training and event management. By leveraging the power of digital human technology, businesses can improve their interactions with customers, increase brand awareness, and drive sales. Whether used as a virtual receptionist, customer support agent, or salesperson, digital humans have the potential to revolutionize the way businesses interact with their customers.
With the right approach, digital human technology has huge potential to provide a valuable and innovative solution for any business looking to connect with its customers in a new and engaging way.
About The Bot Forge
Consistently named as one of the top-ranked AI companies in the UK, The Bot Forge is a UK-based agency that specialises in chatbot & voice assistant design, development and optimisation.
If you'd like a no-obligation chat to discuss your project with one of our team, please book a free consultation.