Talla's own Chief Data Scientist, Byron Galbraith, made an appearance on the A.I. in Industry podcast to discuss why your knowledge base needs artificial intelligence. For those of you who read faster than you listen, here's a quick rundown of the topic.
The biggest problem facing corporate knowledge bases is information discovery, which is a polite way of saying that users can't find the information they're looking for in their company knowledge bases. This problem has three root causes:
- The information is present, but hard to find
- The information is present, but incorrect
- The information is missing
The common solution to this problem, ironically, is to not look for answers in a company knowledge base. Instead, employees ask other employees for answers, which is always inefficient, occasionally annoying, and often a drag on productivity. (This is why the phrase "RTFM" became popular in tech support circles.)
Users go to other employees because those employees either know the needed information outright (and can compensate for what's missing from, or incorrect in, the knowledge base), or can direct the user to a documented answer (and can compensate for what's hidden in the knowledge base).
Unless and until a knowledge base can provide answers more easily and more quickly than simply asking an employee, the knowledge base won't be terribly useful. Fortunately, artificial intelligence can finally make knowledge bases smart enough to replace your "in the know" coworker.
First, A.I. can make your knowledge base search queries better. From automatically correcting spelling errors to parsing related terms, A.I. can understand that a search for "praking payback" is really a search for your "parking reimbursement policy".
A.I. can go further and add context to this same search. If an employee in the Boston office asks another Boston employee about parking reimbursement, they both innately understand that the topic as hand is the Boston parking policy, not the policy for New York, Chicago, or San Francisco. A conventional knowledge base wouldn't know that, but an A.I. knowledge base could recognize that the user in question is a Boston employee, and be sure to return the policy for the correct location.
This is the low-hanging fruit or A.I. assistance. This functionality addresses the issue of information that's present in the knowledge base, but which employees just can't seem to find.
A true A.I. knowledge base goes further, noticing search terms that don't ever return good results, and suggesting to editors that a knowledge base entry needs to be created. Even more advanced A.I. software can perform a topic modelling exercise, looking for common phrases and terms in your corporate documentation, noting how those term interrelate, and suggesting a basic taxonomy or tagging system for your knowledge base. A.I. can go further still, and, after suggesting a taxonomy or organizational structure for your knowledge base, note which topics have out-of-date content, or content that was authored by departed employees and thus needs a new "owner" to review them.
All of this makes A.I. proactive in not just searching your knowledge base, but in populating and maintaining the information within the knowledge base.
If you'd like to spend 20 minutes or so deep-diving on these techniques and the value of artificial intelligence in your knowledge base -- with all the commensurate academic and technical terms -- check out Byron's podcast appearance.
If you'd like to discuss what an A.I. knowledge base can do for your organization, contact Talla today.