What if Twitter understood what you were saying and could then take action on your messages? What if Twitter wasn’t just a place to post your random thoughts, but an A.I. bot that actually helped you get your work done? That’s the concept behind Akibot, a new enterprise microblogging service. At first glance, Akibot may look very much like your typical Twitter clone, but it does something very different: it combines the collective intelligence provided by microblogging with an artificial intelligence engine that lets the service take action on the messages posted.
When Akibot’s developer Marcelo Pham first heard about Twitter, (surprisingly, only a few months ago!) he thought that it sounded like a silly idea. Yet the more he thought about it, the more he thought it began to make sense…just not the way that Twitter had envisioned it. Instead, Marcelo saw microblogging as a “very tiny step towards the machine reading our minds.” He then began to work on a concept for an enterprise microblogging platform that would take Twitter to the next logical step: analyzing what users are posting. He then added another function: intelligence. The end result is Akibot, the first “semantic actionable microblogging platform for the enterprise.”
Examples of Akibot in Use
To understand how Akibot works, imagine the following scenario: you post a message that reads “There will be a meeting next Wednesday morning at 1 PM regarding the new development project.” In other enterprise Twitter clones like Yammer and Present.ly, only those others reading the stream of posts would see the message and would then be able to act on it, if need be. In Akibot, however, the system itself would understand the message and would create an appointment on the team calendar for you. It could even send you and your colleagues a reminder in the form of a text message or email when the meeting time drew near.
Another example goes like this: say a colleague posts a message stating “here is the latest Penske file http://xxxxxxxxx,” – pointing to the resource hosted on the company’s intranet. A week later, another user could ask “Does anyone know where the latest Penske file is?”, and Akibot could then respond with a message pointing to the location previously posted.
Akibot can also function as a time-tracking tool. All you would have to do is post a message letting everyone know when you’re beginning to work on a particular project and then post another when you’re done.
The microblogging service could even update your CRM system with information about customers and your interactions with them. Again, all you’d have to do is post the information to Akibot.
How Does Akibot Work?
In order for Akibot to do what it does, it seeks out various keywords in a post, but not using simple search or in a “brute force” sort of way. Instead, it looks at the sentence structure as a whole to determine meaning. Akibot’s main module is called the “preprocessor” which uses common elements of natural language processing (NLP) combined with two proprietary modules: a “contextual analyzer” and a “context>action” dictionary. The contextual analyzer take the results from the NLP module and finds the context using noun/pronoun/verb structures and then the “context>action” dictionary stores the relevant data and takes action on the item (e.g. it sends a reminder, updates your business software, etc.)
Because Akibot understands natural language – that is, the way people naturally speak – end users posting their messages don’t have to use any special syntax in order for Akibot to understand them. However, if you do end up posting something Akibot doesn’t understand, it will just ask you to explain and then learns from that explanation so it never has to ask again.
Even Simpler Than Twitter?
There are a few things that Akibot does differently than Twitter. For example, there is no “follow” functionality. By default, everyone follows everyone else, but can “opt-out” from following certain other users if they wish. Since Akibot is meant to be used within a single company, this makes sense. The system is also designed to be uncomplicated so there are no groups, no tabs, no browse functionality, and no search.
Like Twitter, though, Akibot supports private messages, but no special syntax is required here, either. To create one of these types of updates, a user simply clicks on “private” when posting.
Still in Private Beta
The company is still brand-new and there are no exact launch dates yet regarding when it will become publicly available. A lot will depend on the feedback provided by the initial crop of beta testers. Also, if Akibot was to receive funding (they have none now), development could proceed at a faster pace, notes Marcel.
At the moment, the company is considering offering Akibot for free for up to three users and then any additional users would cost $1/per user per month.
If your company wants to join the private beta, you can sign up to be considered on Akibot’s homepage under the “Signup” option.