{"id":9270,"date":"2023-12-20T14:03:58","date_gmt":"2023-12-20T11:03:58","guid":{"rendered":"https:\/\/tehisintellekt.ee\/gpt-tehnoloogial-pohineva-ai-assistendi-vaartuse-komponendid\/"},"modified":"2026-02-01T14:51:32","modified_gmt":"2026-02-01T11:51:32","slug":"value-components-of-a-gpt-based-ai-assistant","status":"publish","type":"post","link":"https:\/\/tehisintellekt.ee\/en\/value-components-of-a-gpt-based-ai-assistant\/","title":{"rendered":"Value components of a GPT-based AI assistant"},"content":{"rendered":"\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

A virtual assistant based on GPT technology is an AI solution that can answer employee questions about company policies, rules, processes, and guidelines.<\/p>

Organizations must keep up with technological development to survive. However, technology investments cannot be purely faith-based. To start a project, there must be a clear logic of value creation that explains how it helps reduce costs or increase revenues. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

A leap in employee efficiency<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

When starting the virtual assistant project, we conducted a study where we asked employees to measure how long it takes them to find answers to questions regarding company policies, rules, or processes. The test involved 65 employees from different fields and language backgrounds. <\/p>

The results showed that the median time for employees to find answers to company-related questions is 18 minutes. It is worth noting that the time distribution has a “long tail”, meaning that for 9% of the questions, finding an answer took longer than 2 hours (searching independently, asking a manager, and eventually getting an answer), and for 6% of the questions, no answer was found at all (they tried to search, failed, and moved on to other tasks). <\/p>

If finding an answer manually takes 18 minutes and using a virtual assistant takes 30 seconds, this means an average saving of 17.5 minutes per question. From this, we can derive the following value logic: if an employee needs to find answers to an average of two questions per day, the virtual assistant saves them more than 140 working hours a year (i.e., 252 working days, 2 questions, 17.5 minutes saved = 8820 minutes or 147 working hours). <\/p>

The actual gain depends on how often an employee needs to find answers from company information sources, but the principle is clear: a virtual assistant creates the potential for a leap in efficiency.<\/p>

This has a particularly strong impact on new employees, who have the most questions and know the least about where to find information. While it usually takes months for new employees to settle in, it can be argued that with the help of a virtual assistant, they can solve at least some tasks at a good level from day one. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Managers can focus on management<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

If an employee cannot find an answer to their question, they eventually turn to their manager. This means that managers are in the role of answer brokers on a daily basis. If a virtual assistant helps the employee find the desired information, managers can focus on management, and they also have the assurance that their employees can work on their tasks instead of actively waiting for a piece of information to arrive. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Taking the data management process to a new level<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Data management is mostly a tedious and heavily regulated field. Much documentation is created simply because rules require it. It can be argued that a virtual assistant creates the premise for a new type of data management. <\/p>

Employees, by interacting with the virtual assistant daily, unknowingly provide an assessment of the relevance and accuracy of various datasets. It is quite simple to compile statistics on what people ask about and where the information gaps are the largest. <\/p>

Usage statistics and feedback from the virtual assistant add a new dimension to data management: for the first time, the creation and updating of datasets can be prioritized based on the real needs of users (the need is greatest for what is asked most often or where users rate the quality of answers the lowest).<\/p>

Employee feedback on the virtual assistant adds a new dimension to data management: for the first time, the creation and updating of datasets can be prioritized based on the real needs of users!<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

New functionalities<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

In addition to making work more efficient, generative artificial intelligence creates the opportunity to perform activities that have not been possible until now. Below, I mention two significant advancements we have seen in our projects: <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Ask in \"any\" language<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

If a virtual assistant is built on new large language models, it becomes possible for the first time to ask about an Estonian document in Polish, for example. Or vice versa: about a Polish document in Estonian. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t

\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Making source information more \"readable\"<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Company guidelines and rules are typically written by specialists in their field. This means they are often difficult for an employee from another unit to read. Using the logic of large language models, a virtual assistant allows complex source information to be conveyed in much simpler language. In other words, the technology not only makes data faster to find but also easier to read. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

The virtual assistant creates the prerequisite for the emergence of a learning organization<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Let’s take a step back and look at the value of the virtual assistant in the bigger picture. I argue that generative artificial intelligence makes it possible for the first time to realize certain management science concepts that were developed decades ago but have not been practically feasible until now. <\/p>

The learning organization is a concept popularized by Peter Senge in his 1990 book “The Fifth Discipline”. The idea of such an organization is that the knowledge created by people remains within the organization, becomes accessible to all employees, and through this, the organization constantly becomes smarter. <\/p>

In reality, this concept has never fully materialized because people leaving the organization largely take their acquired knowledge with them, and new people are unable (and unwilling) to work through massive old datasets.<\/p>

The virtual assistant changes this game because artificial intelligence has no objection to processing massive datasets. This means that the knowledge the organization has collected over time becomes easily accessible to all employees. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Summary<\/h4>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Technology investments should not be driven by the “fear of missing out” but by a clear logic of value creation that explains how it helps reduce costs or increase revenues.<\/p>

The most important quantitative and qualitative value components of a GPT-based AI assistant are the following:<\/p>

  • The AI assistant helps significantly speed up finding information and compiling answers, dramatically increasing the efficiency of both new and existing employees.<\/li>
  • When a machine helps answer employee questions, it frees managers from the role of answer brokers and allows them to focus on management.<\/li>
  • A potential is created to take corporate document management to a new level. If employees use the AI assistant and provide feedback, the creation and updating of company guidelines can be prioritized based on real needs. <\/li>
  • In addition to making work more efficient, generative artificial intelligence creates the opportunity to perform activities that have not been possible until now. For example, it allows asking about a Polish document in Estonian or vice versa. It is also able to convey an answer written in complex language by specialists to the user in easier-to-understand language. <\/li><\/ul>

    In the strategic picture, generative artificial intelligence has for the first time opened the opportunity for the emergence of a “learning organization”, where knowledge created by people is stored in the organization and is easily accessible to both existing and new employees, making the organization more efficient day by day.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

    \n\t\t\t\t\t\t
    \n\t\t\t\t\t
    \n\t\t\t
    \n\t\t\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t

    Latest articles<\/b><\/h1>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t\t\t
    \n\t\t\t\t
    \n\t\t