The Future of Conversational AI
Predictions from the experts
The Conversational AI space has made great strides in the underlying technologies, use cases, and adoption.
We are moving past the days of basic question and answer experiences and simple decision trees. Chatbots, voice assistants, and Interactive Voice Response (IVR) solutions have advanced to incorporate sophisticated Natural Language Understanding (NLU) that not only understands a user’s Intent, but enables the chatbot to respond appropriately in a way that satisfies the user.
Conversational AI solutions are used across a wide variety of use cases and industries — including customer service, appointment booking, triaging medical symptoms, signing up for an account, and more.
The demand for conversational AI continues to grow as users prefer to communicate over digital channels. Businesses are developing chatbots, voice assistant, and IVR experiences to increase customer satisfaction, reduce operational costs, and achieve business goals.
As the year comes to a close, leaders in the Conversational AI space share their thoughts on the biggest trends of 2021, as well as their predictions for next year.
The experts include:
- Alan Kiernan, CTO/Co-founder Cation Consulting / Parly
- Alex Weidauer, CEO/Co-founder, Rasa
- Braden Ream, CEO/Co-founder, Voiceflow
- Cathal McGloin, CEO/Co-founder, ServisBot
- Cathy Pearl, Design Manager Google Assistant, Google
- Jake Tyler, CEO/Co-founder, Finn.ai
- Joan Palmiter Bajorek, CEO, Women in Voice
- Laetitia Cailleteau, Global Lead for Conversational AI, Accenture
- Lisa Falkson, Senior VUI Designer, Amazon
- Maaike Coppens, Chief Design Officer, OpenDialog
- Maria Crosas Batista, Conversational AI Product Owner, Nestle
- Sherry Comes, Managing Director Conversational AI, Deloitte
Trends of 2021
Increased enterprise investment in conversational AI
Our experts saw enterprises significantly increase their investments in conversational AI through building out their own internal teams. As Alex Weidauer puts it, large enterprises are “doubling down” on conversational AI. Joan Palmiter Bajorek also points out how enterprises shifted from having one or two people experimenting with chatbots, to hiring whole teams rapidly.
Enterprises are seeing the importance of having conversational AI domain expertise. On a related note, Braden Ream saw interest triple in the past year, as conversational design accelerated as a branch of UX. Conversational AI is not just about the NLU, but the actual conversation flows to enable an intuitive experience that satisfies users and helps guide them back to the “happy path” when things go awry.
Covid-19 accelerated adoption
A common theme mentioned by our experts was the impact of Covid-19 on conversational AI. The demand for chatbots and virtual assistants increased as businesses reacted to Covid-19 social distancing and shelter-in-place orders. As Laetitia Cailleteau describes it, there was a Covid “surge.”
However, this increase in demand also resulted in an increase in simple, tactical solutions to solve problems related to the pandemic, which may not have been the best examples of conversational AI. As Maria Crosas Batista points out, there was an increase in reduced-scope experiences to deal with specific pandemic situations. Alan Kiernan also saw a notable increase in lower grade, decision tree chatbots to handle pandemic related customer support. As Cathal McGoin highlights though, while these experiences may not have been the best, it is not an issue with the technology, but that one must build and train chatbots correctly, and integrate them into business systems, to provide real customer value.
Multimodal design — the rise of mobile
Our experts saw an increase in multimodal use cases and design approaches. In particular, there was an increase in building for mobile experiences. As Cathy Pearl indicates, people are getting more comfortable switching modes within a customer journey — starting with voice on a device, then swiping through options on a screen, and tapping and typing responses. Alex adds most of the successful virtual assistants in the enterprise are tightly integrated into mobile apps.
In addition to multimodal experiences, there have been other trends in improving accessibility and usability. As Lisa Falkson points out, there has been an increase in designing for non-standard speech patterns as well as designing for the elderly.
Voice taking off for utility use cases
After years of optimism, the experts are finally starting to see voice assistants take off for real, utility use cases, like customer service. As Maaike Coppens states, people in the voice industry have been eagerly awaiting voice to take off, and we finally saw that happen as voice experiences moved beyond “gimmicky” use cases to utility ones, and multimodal experiences.
While smart speakers are useful and exciting, there has been a shift to designing for other voice experiences. As Joan mentions, there has been a movement from smart home devices to voice enabled mobile experiences. There is also a significant demand for high quality, automated, IVR systems.
Leveraging data for better user experiences
Our experts saw advancements in the underlying technologies and use of data to improve user experiences.
Natural Language Understanding continues to improve. As Alex points out, “the pace of research is remarkable.” Jake Tyler saw the use of Transformer based models for NLU as one of the biggest trends that drove real world improvements in conversational AI.
In addition to improvements in NLU, conversational AI made heavier use of data. One trend Sherry Comes points out is “extreme personalization” — leveraging data before, during, and after to offer a highly personalized conversational experience. Keeping track of who the user is, and what they have done before, can enable a better user experience. For example, if you call an airline after booking a flight, the IVR will ask if you are calling about the flight you booked, and thus streamline the call journey.
Predictions for 2022
More strategic use cases
While the pandemic may have brought about an increase in tactical chatbots to solve immediate needs, our experts see enterprises moving to more strategic investments in 2022. Alex predicts more and more enterprises will move to strategic implementations. Alan also sees enterprises seeing the benefits of conversational AI and refocusing from pandemic related use cases to future roadmaps.
Empathy, inclusion, and accessibility
A common prediction, or hope for the future, from our experts was to see more empathy, inclusion, and accessibility within conversational AI experiences. For example, Cathy looks forward to more diverse voices and personas, as well as normalizing and exposing people to a wider variety of speakers, genders, and communication styles. Lisa adds to this with the hope to see improvements in voice recognition of non-native speakers. In addition, Laetitia hopes to see a more formalized code of conduct, or ethics charter, for conversational AI interactions.
On a related note is the use of chatbots for AI powered care. For example, Joan hopes to see more conversational AI experiences for mental health use cases. Laetitia also predicts the rise of conversational AI for companionship and care.
Increase in multimodal experiences
While our experts saw an increase in multimodal experiences in 2021, they are predicting even more investment in 2022. As Jake predicts, voice and text interfaces will continue to blend as users expect they can talk to digital experiences and/or click to interact.
It is not only multimodal, but omnichannel experiences that will become more important. As Alex indicates, delivering consistent experiences across a wide variety of channels will be increasingly important and represent a significant challenge.
Leveraging data for optimization and personalization
Data and analytics are especially important in conversational AI as communication is free-form and unstructured — users can say whatever they want, and however they want.
Jake envisions a continued move towards a data centric approach to conversational AI, wherein performance optimization will focus on the quality of data, over new, or better algorithms. Examples of this occur in vertical and domain specific solutions with pre-trained models, like finance, healthcare, or travel, for instance.
Context and personalization are also key in conversational AI experiences. The data known about a user, their behavior, mode of interaction, and history, can enable better user experiences. As Maaike points out, context is a key requirement at the center of what makes or breaks conversational experiences. Similarly, Maria would like to see more personalized, predictive experiences, where chatbots anticipate what a user will say, and follow up afterwards to ensure a great experience.
A relatively new concept in conversational AI is multi-bot orchestration, wherein a “master chatbot” directs users to other use-case-specific chatbots. For example, a financial institution may have separate chatbots to handle commercial, consumer, and mortgage use cases and a master chatbot that orchestrates the interactions across them.
Our experts predict an increase in adoption of these multi-bot orchestration architectures. As Sherry describes, too many companies have everything in one stack, and those days are in the past. Chatbots need to be able to communicate across stacks, and with each other.
Expansion into other technologies
Some of our experts are excited to see conversational AI expand into additional technology areas like AR/VR and the Metaverse. Alan predicts conversational AI will expand into these virtual experiences. Maaike envisions similar expansion into AR/VR, and quite possibly into NFT with “art you talk with.”
Conversational AI is an exciting space that continues to evolve and advance. I look forward to seeing how the underlying technologies improve and new use cases emerge, as well as how our experts’ predictions play out in the future.
In their own words
Alan Kiernan, CTO/Co-founder Cation Consulting / Parly
2021, with covid, was the year of simple tactical responses to customer requirements for most organisations, with long term objectives pushed out, until immediate business concerns were met.
The notable increase in lower grade decision tree based chatbots, as a means to offload customer support queries related to Covid was an obvious quick win for the industry in the chatbot space, unfortunately, the majority of these were tactical, rapid deployments; and thus did not avail of conversational AI.
An increase in social platform based bots, particularly whats app and facebook apps, many of which were simplified conversations in nature, hot word based detection leading to broad drill down menus.
Voice assistants remained limited in the evolution of expansion into high value business services/customer interactions, remaining within the realm of high volume, low complexity responses. (Customer services Vs Purchases/Complex business processing tasks.)
I expect to (and would like to) to see the evolution of consumer engagement into the “Metaverse” serve as means to expand the use of conversation AI within social and virtual society exchanges; and as a byproduct rapidly grow the underlying technology and services to be more at the forefront of the UX industry. The tools and services already exist but their usage could be accelerated with an increased demand.
Example : Summarian Hosts (serviced by lex and polly) on a Lumberyard built ecosystem(metaverse), with a common social function/objective is a compelling prospect, with conversational AI servicing as a means of ecosystem interaction (aka a concierge service with a very broad reach) rather than the user fumbling with A/R controls. If the NLU can be broad enough, such a concierge would greatly reduce the friction/barriers for entry.
Part of the challenge at present is the limitations of initial metaverses, where user demographic and mission may be limited to social gaming and experiences but will evolve into a virtual environment which adopts social, business, education and entertainment experiences; and awesome potential awaits.
Outside of Metaverse..
I (happily) expect that utilisation of NLP in existing proven domains, such as customer interactions with chat, telephony and smart speakers will continue to grow, as providers continue to make the realisation of benefits simpler to unlock and business refocus on post-covid BAU and future roadmaps. In addition to the inclusion of conversational AI in group/social activities such as group watching on streaming services and devices, with concierge services.
Alex Weidauer, CEO/Co-founder, Rasa
1) Large enterprises doubling down on conversational AI with serious investments in their own teams
2) Most of the really successful assistants in enterprise are tightly integrated into mobile apps
3) NLP is getting better and better — the pace of research is remarkable
1) More and more enterprises will move from tactical implementations of conversational AI to strategic ones
2) Delivering a consistent experiences across a vast amount of channels (true omnichannel) will become increasingly important and represents a huge challenge
3) Increased consolidation in the conversational AI startup space
Braden Ream, CEO/Co-founder, Voiceflow
Conversation design is accelerating as a branch of UX Design. By our rough account it nearly tripled the past year based on the number of companies hiring, and designers in the space.
Continued growth of the conversation design field as the UX side of conversational AI.
Cathal McGloin, CEO/Co-founder, ServisBot
Use cases were mainly focused around the problems brought on by the pandemic — chatbots to help overcoming service delivery problems — which led to a lot of disillusionment (Gartner puts chatbots in the trough) — but the reality is that tech alone can’t solve the service delivery issues if you don’t build and train the bots correctly and more importantly integrate them to business systems to provide real value to the customers. So a lot of chatbot failures this year.
Emergence of everything that helps overcome chatbot disappointment — multi-modal, more integrations, better data to train, multi-bot architectures, better conversation design, better expectation setting — leading to emergence from Gartner’s Trough of Disillusionment and more useful solutions as the technology matures. Also consolidation amongst vendors in the space.
Cathy Pearl, Design Manager Google Assistant, Google
There has been an increase in multimodal design thinking. A conversational experience with tech is not just about voice interactions, but includes everything from text-based chatbots to interacting with a smart watch via voice and tap, to a range of interaction modes on a mobile phone. People are starting to get more comfortable with switching modes within one “journey”: perhaps talking to a device to kick off the experience, then swiping through options on a screen, and tapping and typing responses. We need to give users flexibility to switch between modes as they prefer, and not assume we know which one is best for them in their current context.
I would like to see the industry embrace experimenting with a more diverse set of voices and personas, normalizing and exposing people to a wider variety of genders, speaker locales, and communication styles. I loved the recent Wired article “The Future of Digital Assistants Is Queer”.
Jake Tyler, CEO/Co-founder, Finn.ai
Move to wider spread use of transformer based models drove real world improvements in many Chatbot/ virtual assistant applications
The focus will continue to move towards a data centric approach to performance optimization, focused on the quality of data over new/ better algorithms. This approach will mean more time curating, cleaning and labelling data to drive performance vs. Iterating on the model to work around the noise in the data set.
As a result of #1, the market will trend towards vertical or domain specific solutions that bundle in conversational AI into applications. Virtual assistants for banking, travel, healthcare are good examples of this, as are IT helpdesk, HR and knowledge centre applications. The key benefits bring that these pre-trained solutions help businesses realize the value of conversational AI without the complex AI training requirement
Voice, messaging and traditional GUI interfaces will continue to blend with users expecting they can talk to digital experience and/or click. As this happens interactions will move out of channels that are limited to a single model, such as traditional 1800 phone support and into channels that are multi-modal (chat, apps, Alexa etc)
Joan Palmiter Bajorek, CEO, Women in Voice
2021 saw great investment from enterprises in many sectors. Instead of reallocating a developer and content writer to play with making a bot or two, we saw whole teams hiring rapidly. NVIDIA, CVS Health, Walmart, and Facebook/Meta went on a hiring spree this year for director roles and filling out whole product and ML teams. We also saw a movement away from smart home devices and back to voice-enabled mobile experiences. This is especially seen through recent Voice Talks demos with Google Assistant.
Mental health, I’d really love to see the conversational AI field make strides in this vertical. Woebot and Kintsugi are pushing this field forward. Building AI for good and the mental health is both a high impact and lucrative space for much-needed innovation.
Laetitia Cailleteau, Global Lead for Conversational AI, Accenture
Voice is taking off,
Vendors are consolidating,
Focus on contact center transformation and disaster management (i.e. Covid surge),
Companies realising this is not about Tech only
AI powered robots/VA — the next major movement after labor arbitrage.
EU laws (Schrems II) accelerating the insourcing of contact mgt powered by robots, supervised by super humans.
Conversational AI — enabling contact mgt to evolve from a back office/cost center to a strategic element powering the experience economy
The evolution of human and machine relationship / formalisation of code of conduct/ transparency/ Ethics charter
More open source , creation of standards and industry collaboration to avoid “reinventing the wheel” and put more focus on “changing the game”
Conversational AI is the “new app” — or “Engagement layer” (way for brands to engage customers — have conversations in natural ways in the channel of their choice )
Conv AI powered robots — the rise of companionship and AI powered care.
Lisa Falkson, Senior VUI Designer, Amazon
I saw a continued trend towards designing for accessibility, the elderly, and those with non-standard speech (stutters, etc.). One thing that was brand new and exciting in 2021 was the addition of movement to Echo family of devices, i.e. the Echo Show 10 that pivots on a base, and the Astro robot. I think this introduces a new level of complexity (but also usefulness!) to voice-first products.
I think there will be continued work on non-standard speech. Would like to see specific work on detecting and improving recognition of non-native speakers.
Maaike Coppens, Chief Design Officer, OpenDialog
For several years now, each year has been proclaimed as being the year of conversational AI and more specifically voice. Industry professionals have been eagerly waiting, claiming “years of” as if to protect ourselves from the very frightening “trough of disillusionment” … Then came 2021, a year where conversational AI did truly grow substantially.
Voice technologies moved beyond gimmicky use cases to concentrate more deeply on utility (IoT, in-car, CCAI, … ) and cross-platform collaboration.
Increased funding and acquisitions of conversational technology startups, sent a strong signal of validation that conversational AI is here to stay.
When it comes to conversation design, 2021 brought a substantial amount of new resources, and networking opportunities — empowering those new to the field to experiment, practice and learn. The increase in numbers of both conversation design job listings and conversation designers has been truly encouraging.
Most of all, 2021 was the year of innovation for Conversational AI! The industry’s maturity made way for new ideas and models, like the OpenDialog framework, to emerge — beyond intents, flowcharts and happy paths — moving slowly but steadily to truly natural multi-turn interaction models.
2022, will see us into a world where conversational technology is no longer the new kid on the block but an expected mode of interaction for any new application. Just like you wouldn’t fathom building an app today where users couldn’t tap or swipe, the same will become true for conversational interaction.
One key requirement will remain at the centre of what makes or breaks conversational interactions : context-awareness at multiple levels. 2022 will, therefore, continue to see research into how conversational agents can accomplish context-awareness in the long-term and at scale, beyond substantial amounts of man-powered training.
In 2022, the interplay between other modals and conversational technology will also grow significantly. Natural conversations involve a lot more than just speech or text. Conversational technology is only part of the complex puzzle that is natural interaction. The combination of conversational technology, spatial computing, and AR/VR — make for exciting conversational problems to solve in which context-awareness will continue to be a key challenge!
Last year, we have also seen the NFT market taking off substantially — concentrating mainly on the creation and collection of digital art (exception made of the odd tweet). 2022 will probably see conversational technology taking a more substantial dip in the NFT waters — reaching from “art you can talk with” to domain-specific full-fledged conversational agents.
My wish is that amidst our ever increasing ability and facility to converse with machines, both in the real and virtual worlds, we keep point on conversations that truly matter to move the industry forward responsibly : user-centricity, kindness, inclusion and sustainability.
Maria Crosas Batista, Conversational AI Product Owner, Nestle
E-commerce capabilities that allow businesses to sell their products during pandemic/post-pandemic situation. Chatbots with reduced scope to deal with particular situations during pandemic.
More natural conversations where chatbots anticipate what users will say or will go and follow-up to ensure a great experience. Businesses need to understand, though, what’s feasible or not in the AI/ML space. I’d like to see Businesses using less generic chatbots and more specific, listening to what consumers really want to know and what Businesses want to say to them.
Sherry Comes, Managing Director Conversational AI, Deloitte
Extreme personalization! Basically using data before, during and after conversations to highly personalize each conversational experience. The days of calling into a conversational IVR, or engaging with a conversational assistant, and getting the exact same highly frustrating conversational experience for every experience are finally coming to an end!
Companies have too many AI platforms and technology stacks to have all conversational bots remain within one single stack. I would like to see more innovation around the orchestration of hybrid conversational bots. The days of having a single bot, or a few bots all within the same stack are in the past. We now have hybrid bots that need to interact with each other across multiple stacks and with different technologies and platforms. I would like to see conversational middleware that allows all conversational bots, through open and/or standard interfaces, be able to interact, speak and understand one another.