For companies planning to deploy a chatbot to streamline workflows or increase productivity, it can be tempting to create a wizard that can do anything (even coffee).
Experts advise instead to focus on specific use cases - and let the bots look like bots. This would be the best way for them to create value.
The bots more and more present
In the enterprise, bots and virtual agents gain ground in customer services, human resources tools, and those used to manage procurement and procurement. Various companies, like Slack, have partnered with vendors like Oracle or SAP to create chatbots for almost any use case. Other systems will surely follow as the ROIs show the value of these automated self-service interactions.
"Chatbots are also becoming essential for stimulating information discovery," says Mark Kurtz, director of growth and Gage, an American marketing agency.
Using a bot differs from a classic search for information. "For example, an employee may search for something specific in an ERP system, but the chatbot may discover related information that the user did not even know existed. This can lead to new ways of solving business problems, "he says.
Make a chatbot for specific use cases.
Before developing and deploying a bot, some caveats are needed. As popular wisdom says, "He who is average everywhere will be good nowhere." The lesson is especially for chatbots.
"Instead of building a generalist bot, like Siri, who wants to respond to everything, it's better to focus on a specific use case," says Kriti Sharma, vice president of bots and artificial intelligence at Sage. "Even if the bot does one good thing, it's enough for the users".
To get started, she advises companies to focus on the tasks that users perform frequently, rather than the ones they do once a year or once a quarter.
"If the bot helps to do something like regularizing end-of-year taxes, it will be hard to create a craze [and an adoption around the bot]," she diagnoses.
On the other hand, a chatbot does not have to be purely passive. It can be proactive and, for example, send reminders about late bills.
The user experience first
Be careful not to get lost on the road either. For Boaz Hecht, founder of SkyGiraffe, San Francisco-based publisher bought by ServiceNow , the natural language dimension in chatbots would be overrated today. Even counterproductive.
"When I think of robots connected to SAP, I do not care about the conversational part: I care about the functionality part," he says.
Worse, inflating the interface in natural language could generate perfectly useless skills for users, waste resources and lengthen development. Boaz Hecht also advises companies to choose frequently asked questions so that the chatbot can focus on them and respond to them as efficiently as possible.
On the other hand, giving your chatbot a sense of humor through AI can improve the user experience, tempers Kriti Sharma. For example, a chatbot in an accounting tool can use accounting humor to make the experience more enjoyable.
Involve users right from the start
In any case, the overall idea of many effective bots is to map out a user journey: create an expense report, answer questions about the expenses of a particular category and open the possibilities, explains Kriti Sharma.
To achieve this, organizations need to collect input from early-stage users in order to establish a roadmap for that journey, for example by asking for feedback after an invoice has been paid.
"The design and the scope of the bot will change as users give feedback on what they want to build as an assistant," she warns.
AI but also human
Without falling into the excess of the AI at all, to create a successful user experience, you will still - according to Mark Kurtz - integrate an engine of artificial intelligence that allows the bot to learn vocabulary, discussions after discussions.
This learning should also be accompanied by a human whose mission is to train chatbots to match the intention of the user with the answers.
It is also very important to pay attention to the area of practice of the assistant. The bot will need to have enough automation and, if necessary, connection with existing systems to extract the data and ask for clarification from users only when needed, adds Kriti Sharma.
In the same vein, "a good bot will be able to find information about the user and modulate its actions accordingly," she concludes.
The goal is the simplest for the end user. Another option, this simplicity can also, at first in any case, go through use cases where the answer will not need to be personalized and where the bot will not have to connect to systems.
Do not forget: it's a bot!
Experts also agree that trying to get a chatbot to a human being will only hurt the user's experience.
The bot should always present itself as a bot and explain its capabilities, (like helping with the creation of reports).
"A bot is a bot," says Kriti Sharma. "Some people say that they are able to handle queries in real natural language, but in the end, in real life, users are generally disappointed."
Conclusion
For companies that want to have chatbots for self-service interactions, be they internal or external, simplicity, efficiently take advantage of artificial intelligence and automation, and not pretend it This is a human are three essential points for bots to provide a positive experience.
Keeping the user in mind and targeting frequently performed tasks will also help the chatbot gain popularity. This will allow you to build other business cases on a solid foundation.
0 Comments
Post a Comment