Cookie Preferences

We use cookies to provide a better user experience and personalised service. By consenting to the use of cookies, we can develop an even better service and will be able to provide content that is interesting to you. You are in control of your cookie preferences, and you may change them at any time. Read more about our cookies.

Skip to content

Director’s Cut: 4. Good AI stands on top of good management – how do you get there?

Director’s Cut is a no-fluff blog series. Mika Tarhala, Arter’s Managing Director, shares the reality of leadership. What he’s learned the hard way, where he’s succeeded, and why structures are everything.

AI does its work on the basis of the information you give it. In this blog I explain what we did in Arter’s management team before AI started to deliver results, and how we used AI in that work, too.

Up-to-date information is the starting point for managing AI

In the autumn I noticed that an old eNPS figure kept recurring in the drafts AI had produced for me. It had made its way into several drafts before I noticed it.

AI read the figure from an instruction I had written myself, and used it faithfully every time.

A small error, but it showed clearly how AI works. AI needs the same as a new employee: good onboarding and up-to-date information.

Compared with a person, AI does not ask if it does not know. It guesses, or repeats what you once told it.

The management system now has two readers: people and AI

Management, operational and quality systems are familiar to many organisations. In many companies, strategy, operating principles, instructions, processes and metrics have been described and are used systematically.

We used to manage people. Now we also manage AI tools, and soon agents.

A person can cope with an outdated instruction, because they know to ask a colleague if they are unsure whether it is current. AI does not know that the instruction is old. When the management system is read in every task, it cannot be something that is updated once a year.

Organisations that have already modelled their operations have a head start

Many organisations have already done a large part of the work that using AI requires. They have modelled their operations, for example, in line with the ISO 9001, ISO 14001, ISO 45001 or ISO 27001 standard. The processes have been described and the information that guides the work exists.

The information was gathered and maintained for managing people. Now the same information helps to manage AI, too.

Case Arter: annual planning with AI

Before our annual planning day in October, AI interviewed each member of the management team separately.

The problems had been analysed in depth and written down concretely before we started working on them together.

After the interview, each of us spent half a day planning 2027 with AI. Claude was not allowed to write the goals, but it knew our strategy and the 2026 operational plan.

After the homework, I used AI to pull the five plans together. And on the annual planning day itself, instead of presentations, we could use the time for discussion and decisions.

None of this would have worked without the management system, from which AI read the strategy and the plans. I would guess that the same tool starting from a blank slate would have produced a tidy general plan that would suit any software company.

With AI, preparation speeds up and the bottleneck moves to decisions

The day also showed what the speed of AI does to leadership. When preparation sped up, the bottleneck moved to decisions and to communicating them. Our old way of working had served us well, but it suited a slower pace of work.

On the annual planning day we agreed at what level each decision is made, and that every decision always comes with an agreement on who communicates it and to whom.

These decisions have now been put into our management system, and from there they guide AI-driven work across the whole company.

What information should a management system contain to serve AI as well?

Our management system has eight layers, and all of them need to stay in order at the same time, because if one layer contains outdated information, it affects all the layers below it.

The lowest layer tells you who owns the information. At our company every process step has an owner who decides how AI is used in that step, and the check is done within the step. In practice this means that a possible error stops at the point where it arises.

We recognised the need for shared concepts when the word “customer” meant something different depending on which process step we were looking at. People sense the differences in context even when the terms are the same between process steps, but AI cannot recognise this.

We have several fact layers, and they form a hierarchy. The fact layer shared by the whole company is at the top, with the management’s and the functions’ own layers beneath it. Each layer has its own owner. In the event of a conflict, the facts of the higher layer win, and an individual project’s instruction cannot override the higher layer.

Information classification tells AI what it may share externally. Our internal information is labelled, and when AI prepares something that goes outside, it leaves out the labelled information or turns it into a general lesson.

None of this is new to a quality professional or to anyone who knows the ISO standards like the back of their hand. Roles, terms, control of documented information and classification are all found in the standards. AI can be seen as a new stakeholder with its own requirements for how operations are described.

Example content of a management system for the age of AI

  1. Standardised ways of working. The same things are done the same agreed way every time. Then both people and AI can rely on how the work proceeds.
  2. Rights for people and agents. It has been agreed what each person and each AI agent may see, do and decide.
  3. Processes that people and AI understand. The flow of work is described so clearly that both a new employee and AI can follow it.
  4. Traceability from strategy to everyday work. For every task you can see which strategic goal it relates to and why it is being done.
  5. An information management model and information classification. It has been agreed where information is stored, who maintains it and how confidential it is.
  6. One truth. For every matter there is one up-to-date and reliable source of information, not several contradictory versions.
  7. Shared concepts. Words mean the same thing across the whole organisation. For example, “customer” or “project” are understood the same way everywhere.
  8. Ownership. Every matter has a named responsible person, and it is clear who decides what.

How we built the content of the management system at Arter with AI

Eight layers with owners may sound like a complex and heavy system. It is precisely the burden of maintenance that usually brings a management system down. The content goes out of date because updating feels like separate work. As the information ages, use of the management system fades.

Our own experience is that AI lightens the building and maintenance of a management system by about 80%:

  • Describing a process now takes a couple of minutes, where by hand it took about an hour. You can draw the process on a flipchart, take a photo of it and let AI turn it into a draft.
  • You no longer need to start from a blank sheet. When an organisation’s existing instructions and policies exist, half of building the whole management system can be done with the help of AI.
  • I built my own fact layer and project structure with Claude in a couple of weeks. It is a rough-and-ready version, but every AI project of mine now reads the same rules and the same fact layers.

Maintenance can also be done by hand, and some 600 of our customers are doing exactly that right now. The benefit of AI lies in quality and in the lightness of the work. When updating information stops feeling like separate work, the management system stays up to date and in use by everyone.

Even if the world changes, a human still presses the Approve button

AI agents are on their way, but one thing stays the same:

People lead the work of AI. In the management system, AI writes only drafts, and every piece of content has an owner who is responsible for keeping the information up to date. In the end, it is the human who presses the Approve button.

When you need to change how AI operates, you change the management system

With a management system, information changes in one place, and the change shows up in all AI-driven work without any single tool or agent having to be updated separately.

Get to know ARTER, the management system for the age of AI.

All tracks pointing the same way. 🎿

Author of the Director’s Cut blog series: Mika Tarhala

Mika Tarhala is Managing Director at Arter Oy. With over 15 years of experience in the software and consulting industries, he has built a SaaS business from scratch to market leadership, leading it from Finland, with a distributed team, and for two years from Mexico. Today, he is building Arter as part of the European TSS group.

Mika’s leadership philosophy is captured in a single sentence:

Structure enables clarity, clarity builds trust, and trust drives growth.