Artificial intelligence, ITSM and, in general, where does LEAN?

Instead of a preface or where the tentacles of LEAN come from

A couple of years ago, my colleague talked about how LEAN works in our Service Desk division. But somehow he kept silent that LEAN works with us in all service projects, and not just in the Service Desk. In general, LEAN is a very useful tool for finding areas of improvement in work processes, and what is important is a good team-building tool.

Introduction

Once upon a time, in a far, far galaxy ... And yes, Lean Thinking will be with you!



In general, understanding the loss of work of ITSM processes, the team came to the conclusion that for some reason they were wasting people’s time on a task with such a strong monkey-job. More precisely, to coordinate incoming requests to the team stack from all sources. And everything seems to be clear and can be done. But what's the catch? Create classifiers and configure routing based on them, and you will be happy ... And here we were confronted with a problem: "accuracy" suffers and we can’t completely remove the coordinating person, it’s straightforward.

On the way to the right decision or preparation

Well, accuracy and precision .... We follow the Kano principle and decide what we can do with the most reasonable effect: the classification matrix — the decision to set the class by searching for support words in the description, etc. And hallelujah!



70% of the osprey applications blocked by the robot! - Everyone is happy: "We are cool, we are gods ...". We really implemented it and have lived like that for several years. But time passes, and the loss is that, here it is. Now we want both classification and give us the accuracy of a person.

We begin to solve the problem of overlapping the remaining array of applications. Remember that this is approximately 30%.



So, their main problems:

1. Direct user requests without description structure.
2. New types of queries require time to describe.
3. Requests, similar to others, in the classifier go to the wrong team ...

Already it becomes clear where our story is heading? And so, time goes on, and LEAN does not bear the costs ...

So, the essence of the problem

What is a query is a text that needs to be processed, and the results need to be decided on its class. For example, the use of the already described classifier for certain phrases and words requires a rather long preparation of the classification matrix and its constant updating.



Started to think how to be. The team realized that it lacked the ability to solve this problem. Then they turned to colleagues from the optimization department. We have such a team as at Toyota factories, they help the whole company in streamlining processes: they search, dig, etc.

«    ,   ,  ,   …» -   .
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
     

We begin to storm new heights with the use of a storm. Brainstorm is a very useful tool, the 5W method intensifies the storm before the storm! And what did we decide:



Our initial problems:

1. The problem of accuracy, or rather the technological weakness of the existing solution, and there is no way to improve it.
2. The problem of the cost of support - it is necessary to constantly update the classification matrix, to monitor deviations.

What are the suggestions for a solution:

1. It is necessary that the machine can make a decision on the quality of the proposed option.
2. The solution must be self-taught at minimal cost.
3. Support for the solution is not otherwise different in cost from the previous solution.

We begin to sort out the options.



From technology, you can think about statistical analytics with BI elements. Expensive, and why is there a monster with ERP elements? The problems are then similar to the tasks solved by “artificial intelligence” and the mechanisms of “machine learning”. Well, our optimization department, without a doubt of success, called the guys from the digital solutions department for the next meeting.

Solve the problem

For a couple of weeks, data architects and data engineers went through a considerable number of frameworks and rolled out a solution - the first assessment and model:















A month later, we docked our ITSM and Artificial Intelligence and completed testing.



As a result: we do not need query coordinators at all, since the robot now processes 99% of all incidents and does not create a negative feeling of routine for the remaining 10-15 incidents per day. The team is satisfied, not being distracted from the main tasks, the employees received the elimination of the routine, simply stating that this “archaic tool” is already outdated and interferes with work.

Output

Together with the team, constant monitoring of their processes is invaluable. Not only allowing to find costs and eliminate them, but also to form an understanding and need to use new technologies. By solving the tasks of eliminating even the smallest, but completely routine problems, we really create value. And the value is not only for the customer, but also for employees and the company.