Laurent Burdin is the founder of Berlin/Hamburg based “Space and Lemon” Innovation Agency. Space and Lemon focuses on digital trend scouting and Chatbot & AI development for larger corporations. As we’re more and more asked by our clients to develop innovative social activities, we’ve asked Laurent some questions on chatbots…and the state of artificial intelligence.
Chatbots seem to be the flavour of the month in the industry: when would you say they are relevant for organizations and when would you say they are too gimmicky
Any professional and any organisation need to confront itself with chatbots NOW in 2017 because they will have a business impact in 2018/2019. Bots platforms are gaining grounds on various formats: project management tool like slack, messaging like Facebook messenger, website chat like Zendesk Chat or speakers like Amazon Echo. Their connection with AI modules make them strategically key for organisation. They have to adopt, test and start using the immense capabilities of deep learning.
Many developers around the world are testing bots and –granted- some are very gimmicky like this Alexa skill “Alexa, can I get a beer before 4 pm? Etiquette says no beer before 4”.
Any organisation should start a bot/AI project in 2017: build a bot conversation, link it to relevant datas and to AI modules. This can start with a small, gimmicky bot. Then grow expertise, usecase relevance and bot training, and finally to discover huge opportunities and business impact.
Chatbots can be developed through diverse technologies and at different scales: what are the 3 questions a decision-maker should ask before scoping any project?
One plan and 3 questions.
Decision makers need to put a plan together. I call it “Bot and AI wake-up plan”:
- setup a BOT/AI team, train them (and her/himself) on Bot/ AI architecture and develop a sequence of usecases and conversations.
- Then ask 3 questions
- can my present architecture provides with interfaces and datas required for the conversations?
- where is the self-learning and AI part in the bot?
- which input (text, voice), tool or framework shall I choose (see graphics with the Bot/AI ecosystem players).
Definitely separate the UX work, the architecture set-up from the choice of text or voice platform. Hence avoid management decision we saw 10 years ago at the launch of mobile apps: “I too need an app in the store for my company, no matter what!”.
Chatbots require a great UX, a lot of content, and a precise value added: who should lead this kind of projects within organizations?
Ideally a dedicated team with UX, developers and data scientist professionals. My hypothesis is that any organisation will have an AI team in the next 3 years.
Today, the natural leadership goes to the digital team in the organisation. They have the practice of user experience, digital content and IT architecture on websites, mobile apps or social medias.
The crucial decision is to start, understand the bot and AI ecosystem and accumulate experience within the organisation. Systematically planned, and not with one shots to please the management or with the utopic projection that a bot can do anything.
Regarding Artificial Intelligence, chatbots seem to be mostly used in the “Artificial Narrow Intelligence“: when do you think a tipping point will occur and that we’ll enter “Artificial General Intelligence” stage?
The progress of Artificial Intelligence in this digital context is based on the immense computing capabilities, a mature data management practice and a “clash” of ambition of tech giants with billions of cash. Look at Alexa from Amazon, a bot and AI ecosystem that, in its first mainstream usage, the Echo speakers (in the US, UK and Germany), is a threat to Google search dominance. Google needs to react, develop and launch its own system. And Microsoft needs to position itself, etc.
Back to narrow or general? The present development are mainly “narrow” in a sense that the conversation takes place on a defined subject and the answers are ruled-based (“if you say keyword x, I answer phrase y). However new bot developments show more intelligence behind: opening the domains of conversation and generating answers from a machine learning module.
There is no tipping point to reach the general AI (matching a human brain) phase, it will happen step by step, with more and more developments training the machine. If “general” AI is asking a virtual assistant “anything” and it understand me, my emotions, my context, my intentions and gives me an accurate answer… where are nearly there! Regularly try and train the new Google Assistant. The perfect answer? In 3 years and in your car.