IBM and Oracle both forecast that by 2020, 80% of interaction between businesses and users will be automated. Although that prediction may be a bit too optimistic, Robotic Process Automation (RPA) is currently widely used across industries to speed-up mundane processes and improve customer relations in the form of virtual assistants, such as chatbots. The next logical step is the implementation of RPA-assisted “voice bots”, however adoption has been slow. So, why don’t more organizations and businesses deploy voice assistants to further automate customer interactions?
The perceived problem with using voice bots
Although there will be eight billion AI-powered voice bots by 2023 (meaning nearly every human on earth will interact with AI), many businesses currently fear that voice bots will frustrate and repel customers for multiple reasons:
- Users might hang up because they get frustrated communicating with a robot;
- The robot could be “dumb” and unable to comprehend complex verbal questions and answers;
- The robotic “iron voice” might annoys users;
- Why call when everyone prefers to communicate via text/chat these days?
These fears mostly seem to stem from consumers’ interactions with Interactive Voice Response (IVR) technology that is based on Dual Tone Multi-Frequency (DTMF) touchtone user entry. IVR via DTMF has been widely employed by the communications industry for decades, but superior technologies that automate customer services and interactions are rapidly becoming more capable and accessible to businesses of all types.
The actual problem of not using voice bots
Voice assistants, or “voice bots,” have the potential to provide numerous advantages to businesses in retail, hospitality, financial services, and really any customer-facing industry – so practically all businesses serve to benefit from the implementation of voice assistants.
Most businesses who currently employ bots of any nature (chatbots, text-reading bots, voice bots) cite several major advantages provided by their automated assistants. First – The bot can work 24/7 and never needs to take a vacation (bots are also available on holidays when the rest of the workforce is largely unavailable). Second – bots never tire or become stressed out – meaning they can handle large quantities of tasks and requests in a timely fashion. Many businesses are subject to peak loads, and bots can quick and efficiently notify customers of any changes such as cancelled flights, new products, deliveries, etc. Third – bots don’t nearly require the same amount of human resources employee’s time as hiring a new employee. Training a bot may take several days but once fully trained, a bot can be used for as many simultaneous requests as possible. Bots also don’t need to be managed or paid – meaning their deployment is up to three-times faster than onboarding a new employee.
Most businesses’ major fears regarding voice bots are related to the bots’ ability to comprehend information and communicate responses in an effective manner. As mentioned above, some businesses fear that voice bots will be too “dumb” to understand complex questions and formulate answers.
Ironically, “dumb” bots are often more effective than “smart” ones. A smart bot can ask and answer open ended questions, often requiring the bot to operate beyond the parameters of its given algorithm. A dumb bot only understands “yes,” “no,” “I don’t know,” or similar short, pre-programmed answers. While the smart bot has a wider range of capabilities and customer interactions, it takes much longer to train and costs a great deal more to implement than a dumb bot. In addition, when a smart bot is presented with a phrase or request that its algorithm does not recognize, the bot can get confused and lead the conversation in erroneous directions – meaning live agents must be available to handle any out-of-bounds requests. Sometimes, less is more. In this case, it’s much more efficient for a “dumb” bot to lead the customer to the right solution through the positioning of select closed-ended questions that can be answered with short and direct responses.
As voice recognition and voice synthesis technology will only improve, the fears of a robotic “iron voice” that drives away users will also fade. Recent breakthroughs in AI and deep learning have resulted in highly effective bots that can mimic human voices, intonations, and cadence quite well. We’ve all received that robo-marketing call from a spam number that sounds like a legitimate human being at first, only for us to realize it’s a bot when it keeps repeating the same phrase.
How to create & deploy an effective voice bot
To create an effective voice bot, you will need a number of components: accurate voice synthesis and speech recognition technology, as well as connecting a Natural Language Understanding (NLU) platform (such as Dialogflow) to the bot to aid in resolving complex scenarios.
The NLU connected with the bot should have the ability to integrate with a broad range of messaging and communication apps and platforms, and to understand and respond in a variety of different languages. NLU’s should also feature an easy-to-use interface for programming and develop the bot. Interfaces should include a knowledge integrator that auto-populates the bot with data; interaction capabilities that allow the bot to naturally handle interactions without requiring specific programming; history and performance improvement tools.
When deploying the voice bot, do not expect it to work out of the box in CRM. But rest assured that integration is quite simple. From scratch, along with testing, setting up specific automated voice bot capabilities should only take a few days depending on the complexity of the bot’s capability requirements. The overall cost of developing a bot ultimately depends on the level and scale of the technologies involved, which should be chosen based on the specific needs of the bot’s capabilities.
There are more and more situations that now require automated calling and customer interaction, drastically increasing the demand for effective voice bots as customer service calls are often now being transferred to virtual assistants for resolution. Thanks to recent breakthroughs in AI, deep learning, and voice synthesis technology voice bots are more accessible and deployable than ever. It’s up to businesses that want to stay ahead of the CRM curve to begin developing and deploying their automated customer service team of the future, i.e. voice bots.
About the Author
Alexey Aylarov is the co-founder and CEO of Voximplant. In 2008, he graduated from Bauman Moscow State Technical University with a degree in Computer Systems and Networks. In 2010, he co-founded Zingaya, and in 2013 – Voximplant. Alexey is an expert in telephony, audio, and video communications, messaging, chatbots.
Sign up for the free insideAI News newsletter.
Great read and I especially liked the ‘dialed down’ tone of the article. While it’s true that many processes are being automated, not as much of the interaction will be automation so fast. I enjoyed the smart vs dumb bot debate, and have to agree with your conclusion, that sometimes less is more. Still looking forward to seeing how the market develops these solutions and what voice bots will look like in 5 years time 🙂
Thank you for this insightful and well-researched article. You have shown us the current case analysis on the usage of voice bots across several industries, as well as how to build effective chatbots in a well-put-together how-to.