The Real Knowledge Base · from our own projects
Should you manage AI centrally, or let each department choose?
Let a department choose for itself as long as the consequences affect only that department. Decide together as soon as other departments need the same data or systems. Set those shared agreements down centrally, including who decides and who remains responsible. Not everyone needs to use the same tools for that. That way it is clear who chooses and which customer data and prices apply.
Why the question matters now
AI often enters a company through individual departments. Marketing adopts a writing assistant, sales a tool for leads, administration something for invoices. Each tool works on its own. Together they can produce different versions of the same customer, price or text. If you let everything be decided centrally, you get the opposite problem. Then a department waits for a decision about its own work.
Where the line lies
The question is not central or decentralised, but who bears the consequences. A writing assistant for your own texts mainly affects that department. That choice can stay there, within the ground rules. A tool that changes customer data or sends quotations also affects sales, administration and your customer. That calls for a joint decision. That way every department knows in advance where it can choose for itself.
What you agree on together
- Ground rules. Which tools and customer data may be used, and whether AI may train on that data. Agree who approves the work. Apply stricter checks where the brand is at risk. Allow more flexibility where mistakes can be corrected without harm.
- Data that takes precedence. For each data item, one version to use, for customers, products, prices and brand rules. With someone who keeps that version up to date.
- Shared systems. Agree which data departments share and who is responsible for it.
Also agree who decides when two departments want different things. Set down the criteria independently of the tools. Then they still stand when a different AI model comes along.
What you leave to the department
The department knows the work. It knows what takes the most time and where a mistake does damage. So let it choose where to start. Weigh not only the time, but also what a mistake costs. How to weigh that per task is explained in Which tasks do you leave to AI, and which stay with your team?
How we do it ourselves
At THE REAL MAG. all our knowledge, clients and systems come together in one AI brain on our own server. It grows with the company. Our bookkeeping, lead follow-up and the monitoring of our ads also run on our own systems.
Where it goes wrong
- Each department chooses a tool, with its own customer data. Then the same customer exists in several versions.
- Letting everything be decided centrally. Then a department waits for a decision about its own work.
- Treating IT as an opponent. A restriction often has a reason. Look together at what is possible within it.
- No agreement on who decides. Then whoever starts first also decides for everyone else.
What this means for you
Where departments affect each other, they need to consider their choices together. In the future-proof scan we look at your processes, tools and data in your workplace. You decide in advance what we look at.
Frequently asked questions
Does IT have to manage the AI in our company?
IT looks after the technology, security and data. Which tasks AI is given and what is good enough is decided by management and the departments. So work together, with someone who understands both sides.
What if departments already use their own tools?
First map out what is running and with which data. If a tool only affects its own department, it can stay, within the ground rules. If it uses shared data, assess together whether it can stay within the agreements.
Where do you start if nothing has been arranged yet?
With the ground rules, and with one process that takes a lot of time and where a mistake can be fixed. That is where you see what works. Then you take that to the rest of the company.
Sources and accountability
- This article describes our own way of working, as set out on our TID and CID pages.
- The example of our own company also comes from there.
- Our advice on these trade-offs is based on this way of working.
- These sources do not quantify the benefits of shared agreements. That is why there is no figure here.
Where do your departments make conflicting choices?
The future-proof scan gives you an honest verdict and a plan for the next steps.
Request the future-proof scan