Calling for alignment in Conversation Design
NOTE: This paper is in development. It is currently being edited by industry stakeholders to create full alignment on it. Although there is support for the paper as is, some of the wording may change over time. In 2021, an advisory board will be installed to ensure the paper stays up to date as new interfaces and technologies emerge. At the end of 2020, the authors and supporters will be shown on this website. It is already published here for full transparency.
AI Assistants, like chatbots and voice assistants, are developing into a core element of customer conversations. Around the world, chatbots and voice assistants are powering billions of conversations a day.
We need to make sure that these are good conversations. Not only do we want to understand what people say, we want people to feel understood. Not only do we want our applications to be secure, but we also want people to feel safe.
Only when we focus on human-centricity, we can successfully move towards level 3 AI Assistants, where we have more personalized and transactional experiences. Only when people feel comfortable talking to chatbots and voice assistants, we can focus on critical journeys and create personalized and contextual experiences.
Advancing the field of conversation design is key to this. Conversation Design is the practice of making AI Assistants more helpful and natural when they talk with humans. It combines an understanding of technology, psychology, and language to create human-centric experiences for chatbots and voice assistants.
The world of Conversation Design is fragmented
Around the world, thousands of people work in conversation design. They are all trying to discover how to design conversations that make AI Assistants valuable for both the organization and the people interacting with them.
But this community is fragmented. People use different platforms, tools, job titles, workflows, and terminologies. This fragmentation stands in the way of a faster advancement and adoption of conversation design as a serious domain. It stands in the way of creating valuable human-centric AI Assistants.
This paper, therefore, calls for alignment on workflow, job titles, notation standards, and responsibilities. Alignment on these elements will increase liquidity in the conversation design market which benefits the conversational AI industry as a whole.
Alignment on workflow, job titles, and responsibilities, achieves the following goals
- Designers know which skill set to acquire, can promote themselves better, and are in a position to advance their careers faster.
- Helps organizations find, hire, and onboard qualified conversation designers faster and scale their teams more easily to leverage conversational AI technology.
- Helps universities offer their students better courses on conversation design and students will have a better career perspective.
- Technology companies know that there is a workforce that is ready to unlock the potential of their products and also gain better understanding of their ultimate users.
Industry alignment on human-centric workflow
A human-centric workflow is crucial when designing AI Assistants. We don't just want to focus on developing and deploying technology, we want to understand who the people are that are going to interact with our conversational experiences. We want a balance between technology, psychology, and language. We want to understand what people say and we want people to feel understood.
As an industry, we need to align on workflow. This workflow makes sure that we design conversations that are both human-centric and at the same time make best use of conversational AI technology. This workflow describes how the roles interact and team members work together and make best use of each other’s expertise.
It is important to state that this workflow is voice first. A voice first approach ensures that teams can design with the bare minimum in terms of interface. It can therefore be used to design for all kinds of conversational interfaces. It doesn't matter if you are designing for voice, chat, social robotics, or any other conversational interface. Conversation design is both technology and interface agnostic, it is simply the way to design, deploy, and manage conversations at scale.
Key elements within this workflow:
- Stay consistent with bot personality
- Understand the mindset of the audience
- Create sample dialogue (role-play) to discover the most natural flow of the conversation
- Quick validation of the conversation with Wizard of Oz tests
- Clear focus on long tail design with clear handovers
- Standardised error handling processes to make sure the conversation can always continue
- Use real utterances for training language models
- Continuous optimisation through data analysis to improve on matching and experience
The Conversation Design Institute has developed a workflow that is built on these elements. It is currently being used by enterprises around the world that are deploying AI Assistants at scale for both chat and voice user interfaces.
Job titles for designing AI Assistants
Designing for AI Assistants is not a job for one person. It requires a team with a well balanced skill set that works together in a structured way. These are three of the key roles that are fundamental to making AI assistants helpful, natural, and successful. These are the roles we recognize, develop, and promote in the industry.
That doesn’t necessarily mean that these are always three different people. Oftentimes, one person will have multiple responsibilities within a team. Yet as teams grow, these specialties require different people.
The AI Trainer turns data into understanding. The AI Trainer identifies and priorities intents for automation. It is the most technical of all roles. This person works with conversational AI technologies, analytics platforms, and tools that will help improve the language model.
The AI Trainer:
- implements and deploys the conversations
- ensures the engine understands the user’s input
- identifies intents and improvements to conversation
- monitors the quality of handover flows
- collects feedback and advises conversarsation designers on improvements
- knows the technology and tools well and is able to extra data for analysis
How technical this role is depends on the technology a team is using. Some interfaces are more user friendly than others, and there is a collection of tools that makes it easier for non-technical AI Trainers to train and improve AI Assistants.
The Conversation Designer maps the flow of the conversation. They have a human-centric mindset and turn knowledge into conversational flow. This person often has a background in UX design.
The Conversation Designer:
- identifies needs for both the user and bot
- creates sample dialogue to clarify the most natural flow
- maps the flow with the help of flowcharts and conversation mapping techniques Flowchart design to structure the flow
- tests to validate the flow with users
- designs handover experience and processes
- outlines error handling steps and processes to make sure the conversation always continues
- is responsible for continuous optimisation based on feedback from users and the AI Trainer
The Conversation Designer turns knowledge and business process into human-centric flows. This is the structure of the conversation that can then be built upon by the conversational copywriter.
The Conversational Copywriter is responsible for turning words into resonating dialogue.
The Conversational Copywriter:
- defines requirements and develops a conversational strategy
- develops bot personality to ensure the proper tone-of-voice
- writes dialogue that is clear, consistent, and engaging
- understands how to apply copywriting techniques to guide users towards desired behavior
- designs long-tail strategies to protect the Pareto principle, where we ensure that 80 percent of the user go through 20 percent of the conversational paths to keep dialogues manageable
- writes engaging copy for difficult conversations like fall back or handover messages
The Conversational Copywriter turns words in clear and resonating dialogue. The AI Assistant needs to speak in the right tone of voice and make users feel comfortable engaging in conversation. The Conversational Copywriter often has a background in copywriting, linguistics, communication, or marketing.
A clear workflow that lets the AI Trainer, Conversation Designer, and Conversational Copywriter work together
It is important to state that this workflow is voice first and can be used to design for all kinds of conversational interfaces. The benefit of designing voice-first is that you force yourself to work with the bare minimum. The design can then be leveraged by various technologies and conversational interfaces.
It doesn't matter if you are designing for voice, chat, social robotics, or any other conversational interface. It is both technology and interface agnostic, it is simply the way to design, deploy, and manage conversations at scale.
Conversation design is the key ingredient for getting value from AI Assistants. It is therefore Conversation Design Institute’s goal to standardise conversation design across industry and provide the structure and processes to support automated conversations. It's therefore important that we organize ourselves better. Aligning workflow, job titles, and underlying responsibilities increases liquidity in the market and advances the field faster. This benefits ALL stakeholders in the industry.
- 1. Designers know which skill set to acquire, can promote themselves better, and are in a position to advance their careers faster.
- 2. Helps organizations find, hire, and onboard qualified conversation designers faster and scale their teams more easily to leverage conversational AI technology.
- 3. Helps universities offer their students better courses on conversation design and students will have a better career perspective.
- 4. Technology companies know that there is a workforce that is ready to unlock the potential of their products and also gain better understanding of their ultimate users.
This paper is co-authored and supported by stakeholders in our industry. It is the flag to rally behind for the common goal of recognising, developing, and promoting the role of the conversation designer.
Conversation designers in the world, unite!
The world of AI Assistants is developing rapidly. Technology is being developed, use cases are being discovered, and new interfaces will be added in the next few years. The Conversation Design Institute will, therefore, have an advisory board that comes together every 3 months to discuss the manifest and make changes where needed.
Key stakeholders from across academia, enterprise, consultancy and technology have a seat on the advisory board.
Open-sourcing job descriptions
Based on this manifest, we will initiate a project to open-source job descriptions. These job descriptions can be used by enterprises to help them with finding and hiring the people that are going to help them get value from AI Assistants.
- AI Trainer
- Conversation Designer
- Conversational Copywriter