August 15, 2017

I, Robot Negotiator

Chatbots. They’re the text-based cousin to Alexa, Siri, and the other electronic smarty pants personal assistants. But realistically, many chatbots are just FAQs redesigned as text bubbles. Inoffensive, but not really all that useful.

To really give chatbots a purpose, they need to be able to understand context and reason. To essentially think on their own. They need a heapin helpin’ of AI.

That’s where Facebook comes in. They’ve been experimenting with chatbots for a while, and it seems like they’re having success in getting chatbots to do something we all hate to do: negotiating.

Using neural network tech, based on a set of recorded negotiations, they were able to train their chatbot to properly negotiate with someone.

They didn’t stop there though, adding the ability to “think ahead” about potential outcomes enables them to plan out their best method of negotiation. One of those methods, which it learned on its own by trying to achieve its goal, was the ability to “bluff.”  Hmm, how human.

And now the more negotiating it does or sees, the better it gets.

So imagine an insurance company negotiating rates using a chatbot. After likely seeing hundreds if not thousands of negotiations a day, it will know far more how to get you to the right tipping point — advantage, insurance company.

If you’re a marketer, the uses for this could be endless. Automating media buys, pricing projects, etc. And with the data crunching, automated learning, and lack of emotion, one would assume anyone using that chatbot can save a lot, in money and time.

That said, one wonders about the qualitative side of things. Decision making can be a fickle thing. And the results of repetitive actions getting to predictions might wear out, or morph, over time.  

Finally, what might it do in the face of trying to predict reactions when confronted with something truly original?  

Well, we’ll have to invent something for that someday, won’t we?