Blog | Publish date: 17 Aug 2021

What do Tegnell and AI have in common?

No, this is not a joke. Both of them represent areas where there is currently a lot of dialogue about what ‘responsibility’ means. Having said that, dialogue about responsibility is probably taking place on a daily basis in most organisations, as many of them have made major changes to their operations and need to look at “who does what”. In this blog post, Stefan Palm reflects on the concept of responsibility, and what it means for the individual and the company.

I personally believe that one fundamental mistake is made in these dialogues about responsibility. People are not sure what responsibility means. This is understandable in the media, as you can write a lot more around it if there is anything that is unclear. But most organisations would benefit from reaching a consensus on what responsibility means.

Now you might think that this is not that difficult. Of course you know what responsibility means.
My point is that the person you are working with may have a different opinion of what responsibility means, and this is what can cause problems.

Who is actually responsible for what?

Let’s have a look at a few examples to get us thinking.
We will start with a simple question. Who is responsible for the finances in a private company? You could answer this in different ways. Here are a few possible angles:

  • It is the senior management of the company, who run the company on behalf of the owners.
  • It is the finance department, which deals with the budgetary process and delivers the financial results.
  • It is the various operations managers who have a clear responsibility for their financial results.

I hope you are starting to realise that this question of responsibility is not that easy. Next example: Who is responsible for IT security in an organisation?
Once again, it is not difficult to find a few different angles:

  • The senior management at the company is responsible for everything, including IT security.
  • The IT department is responsible for all IT issues, including security.
  • Our CSO has set clear guidelines for this, and every operations manager is responsible for ensuring compliance with them in their own area.
  • In our organisation, we trust our employees, and have distributed a responsibility model, where each individual is responsible for following our IT security guidelines.

You might recognise one of these situations, as I have not made them up. They come from real-life experience.

That’s enough examples for now. Instead let us move on to introducing a few elements that are good to include in your dialogue about responsibility. The first thing you need is a good definition that you can all agree on:
“Responsibility exists, where there is the knowledge and resources (ability) to make changes”.
This is quite a ‘strict’ definition and you might feel that it is far removed from your own experiences.

Instead, you are responsible for something, but you do not have the resources at your disposal. Or you are responsible for a new area, but you are not able to gain the new knowledge you need in this area when you ask if you can go on a course.

RACI matrix

The problem is that we use the word ‘responsible’ in a careless way. What responsibility do I actually have? This is when it can be a good idea to produce an RACI matrix so that we can work in a smarter way. Responsible, Accountable, Consulted, Informed is the definition given in Wikipedia (but I have seen other versions over the years).

Responsible, where you write the answer to the question “Who is responsible?”. As I hope you now realise, this is a thorny issue. But it is good to use the definition I gave earlier. However, for some issues, you also need to consider legal aspects. When it comes to the law, what you personally think is not as important; it is a court of law that will determine who is responsible.

Accountable, the person who makes sure that something has been done. This is normally the person who owns the measurement figure linked to the issue. Because if we do not measure it, it will not have been done, isn’t that right?

Consulted, a person who may have an opinion on this issue, who we should talk to before making a decision. The challenge here is to narrow the list; many people may have an opinion, but who should we listen to before making a decision?

Informed, who should be informed, i.e. who is affected by the decisions we make in this issue. Also remember to check with them to make sure they have understood what is being communicated.

No, it is not always easy to reach a consensus on this, but this is why this is such a good exercise in my opinion. And as with many processes/models/tools, it is not about getting it ‘right’ from the start; what is normally most important is to get started. This will allow you to learn from your experiences and do better next time.

Explainable AI

The RACI matrix is not a new tool, but as I said, I recently became aware of it when I was studying ‘Explainable AI’. This is an exciting field in AI that deserves more attention in my opinion. Looking back, AI algorithms have acted like black boxes. They have consumed data and produced statistical results in the form of a classification or estimation. We were not able to explain how this happened. But there is growing concern and a growing insight that we need to understand how AI models work, following a number of bad implementations. This has led to the field of ‘Explainable AI’.
Many different frameworks are now available that you can use to analyse and simulate behaviours in your AI models before you deploy them. This field also includes the ability to monitor an AI model in production to determine when the behaviour starts to deviate from what it is supposed to be delivering.

You could say that this gives us an ability to control AI technology. That does not sound like such a bad thing, does it? But do not forget that it is us as people who are responsible for what AI does today and also what AI will be able to do in the future.
And if you would like help to understand what AI can do for you, get in touch with us here at Softronic.

Blog post written by Stefan Palm – copyright Softronic.