When regulating artificial intelligence systems, their potential impact on society and individuals is taken into account
Well-organized information management makes finding and using information easy, allowing for quick and correct decisions and ensuring organizational goals are met. (PriceWaterhouseCooper, 2014). I argue that in reality, no one has managed to achieve this. The main problems are:
- Non-use – information is collected and processed but not used.
- Deficient info – information is incorrect, outdated, or incomplete.
- Overview – there is no overview of information assets and their usage.
- Accessibility – information cannot be found or accessed.
- Usability – information is not in a format convenient for the user.
Organizations constantly spend resources to solve these problems, but as data volumes grow at an accelerating pace, it seems the problem is always one step ahead. It is like a vicious cycle – while solving one problem, two more have emerged in the meantime.
AI assistant makes information accessible
By AI assistant, I mean a system that has access to company documents and other information sources and can answer user questions about this info using a familiar chat program (Microsoft Teams, Slack, or web chat). Such a system helps improve information accessibility in several ways:
- finding info becomes “possible” – over time, tens of thousands of documents may have been created in an organization, and finding the right info is like looking for a needle in a haystack. For employees, this is often an impossible task, but modern search systems (semantic search) that search for texts according to their contextual meaning can find the right piece of info.
- searching for info becomes convenient – people are used to using chat programs daily in both professional and personal communication. Asking a question and reading the answer through a chat window is significantly more convenient than navigating large and old document repositories.
- info becomes understandable – internal company documents are usually drafted by experts in their field, meaning the info is difficult for a third party to understand. Artificial intelligence can simplify the wording of the source info and make the answer more understandable for the reader.
Thus, three positive developments occur simultaneously, which together help make employees aware of the policies, rules, process descriptions, and other documents created in the company.
Artificial intelligence helps find information, phrases it in an understandable language, and does so through a simple and familiar user interface.
AI assistant helps improve the quality of source information
Suppose the AI assistant is ready and employees are asking it questions daily. You can be sure there will be questions for which no answer exists in your datasets. You can also be sure that some of the source info is outdated, incomplete, or incorrect, and this will carry over into the answers provided by the AI.
But what exactly are employees asking, and which answers are of low quality? To find out, a couple of essential functionalities need to be added to the AI assistant:
- logging of questions and answers – a system that saves employee questions and answers in an anonymized form;
- option to rate answers – the possibility to leave thumbs-up/thumbs-down feedback for each answer.
Now a new and interesting world opens up:
- First, we can start generating statistics by question topics – what the real information need of employees is and how it changes over time.
- It is also possible to analyze employee ratings of answers by topic – we get info on which source documents need to be supplemented or updated.
- Additionally, we can find questions for which no answer was found – this shows what new information should be created.
While company work instructions are usually created and updated because regulations require it, now for the first time, data owners can prioritize their work according to users’ real information needs.
“AI assistant usage statistics show what information users actually need and which parts need to be updated.”
For data owners, the new arrangement means that work becomes more relevant and pleasant: we no longer create info because rules dictate it, but because last week 25 employees actually needed that info.
Summary
With current technologies, it has been very difficult to achieve information management that ensures user awareness and the creation of relevant info.
AI assistant completely changes information management: on one hand, AI helps find info among thousands of documents and makes it usable through a familiar chat interface. On the other hand, usage statistics provide data owners with insights into real user needs and document quality.
