IDC’s European Software Survey 2020 suggests that a big uptick in demand for conversational AI platforms is on its way. 23% of European organisations said they planned to implement conversational AI systems (the tools to build text- and voice-based AI agents for mobile apps, websites, contact centres and other internal processes) in the next 18 months, compared with just 8% that say they already use the technology. Competition for new users is heating up, and there is probably no vendor that has its eyes more on the prize than Google Cloud.
This week sees Google Cloud’s latest conversational platform, Dialogflow CX, become generally available to global users. Dialogflow CX is a step change from Dialogflow Essentials, with Google putting a greater focus on ease of use, collaboration and improved integration into new channels such as voice to further enhance Google Cloud’s Contact Center AI.
Ease of Use
First off, Dialogflow CX uses machine learning to learn the intents of potential questions an agent might receive. Instead of entering many variants of the same question when creating a new intent in Dialogflow CX, developers need to input just a few example phrases or questions because built-in machine learning expands on your list with other similar phrases. This makes the process of building intents considerably easier to get right and saves on time thinking up every possible variation of the same question.
Also, the UI of the Dialogflow CX’s console, the main area where agents are developed, has been significantly upgraded with a new intuitive visual builder. The end-to-end agent design can be done through the visual builder, which makes it easy to build and manage complex agents. This is a big step up from the form-based UI experience of Dialogflow ES.
New Channels
Dialogflow ES did have an API for third parties to build integration into their or other systems, making it possible to implement a virtual agent across different channels. Dialogflow CX goes one step further, adding direct “one-click” integration into the CX console for contact centre software providers such as Avaya and AudioCode, a communication and workplace systems software provider (more contact centre integrations are planned in the near future). Most importantly, these one-click integrations are now powered by native support for telephony features such as DTMF and live agent handoff.
The clear intent with these integrations is to provide a means by which developers can easily leverage their conversational agents over the voice channel without developing a customised solution. By building out better and deeper integrations with trusted software vendors in the contact centre and workplace systems domains (which already own most of the relationships for voice channels), Dialogflow CX is well placed to serve not just the demand for chatbots, but the emerging voicebot demand too.
Multilingual Agents
An impressive feature of Dialogflow CX is that it currently supports multilingual conversational agents, which enables users to design the conversational experience once and then use it across multiple languages. Dialogflow CX also plans to provide more support in the future, including “auto-translation” of training phrases, prompts and so on across languages. This will significantly improve the customer experience of building and maintaining a multilingual conversational agent.
If it Ain’t Broke … Improve
Google has been wise not to throw out any of the best features of the original Dialogflow technology. Developers still get tools for managing access and collaboration of different team members with levels of privileges across CX. The analytics dashboard is still there, but with an improved UI. To top it off, Google has expanded the depth of the prebuilt agents. For developers that want to deliver a sophisticated bot in a short timeframe, often the fastest approach is to rework an existing prebuilt agent. New prebuilt agent templates demonstrate the full range of features that Dialogflow CX offers, acting as key learning examples for anyone trying to get to grips with how a feature may work.
Our Experience
After spending a few hours experimenting with Dialogflow CX, we would expect even beginners to find the platform straightforward enough to use and be able to build a simple agent. Using the more advanced CX features such as fulfilments, integration and webhooks, which are targeted towards advanced users, would require more training. Google would be well advised to create a range of learning resources to help developers take their CX skills to the next level. Getting started is very easy, as Google Cloud offers free cloud credits to begin using Dialogflow CX.
Where to Go Next
The conversational AI platforms software segment and its end product quality continue to be weighed down by the technology’s very hierarchical approach to designing agents. Consultants, management staff and specialised developers still predominantly drive implementations of conversational platforms. Contact centre agents and team members at the coalface of interacting with customers directly daily, who may hold some of the best knowledge about how to structure interactions with them, have mostly been excluded from the process of conversational agent design improvement and management.
Dialogflow CX further lowers the barrier to entry for using a conversational AI platform alongside the existing collaborative tools, enhancing the technology’s collaborative capabilities. However, it remains to be seen whether these steps will be enough to meaningfully enable a wider range of employees, from a selection of business units, to play a part in designing, testing and ensuring the continual improvement of virtual conversational agents. Until such a point, virtual agents could remain towards the bottom of the list of preferred means of interacting with a company for many consumers.
Upcoming Research
IDC’s European AI strategies service will publish research studies on the key trends shaping the European conversational AI platforms market over the first quarter of 2021. To find out more, please contact IDC Senior Research Analyst Jack Vernon or Neil Ward-Dutton, VP of IDC’s AI and Intelligent Process Automation European Practices.