Ask the AI assistant in sales and the one in service about your company's delivery process. The chances of getting two different answers are high. Not because anyone was careless, but because the system was designed exactly that way.
The department as a knowledge container is not an organisational mistake. It is one of the success stories of industrialisation. Taylorist division of labour, specialised expertise, clear responsibilities: each unit knows its area and owns its knowledge. That was the right design for an economy built on human collaboration.
AI works differently.
🐛 The structural mismatch
A language model does not think in departments. It processes what it finds. If sales gives it its picture of the company, service gives it theirs, and finance gives it another, it works with fragmented context. Every department is right from its own perspective. Together they do not add up to a coherent picture.
What was a smart separation for human collaboration becomes systematic noise for AI. Not because departments are doing anything wrong, but because the tool needs a different foundation than humans do.
People bridge silos through conversations and personal networks. That connective layer emerged informally over years. AI cannot read it. It needs what is written down.
This is not a management problem that can be organised away. It is a structural difference between what was optimised for human collaboration and what working with AI requires. What got us here will not get us further.
🦋 The new layer
Departments do not disappear. Specialised knowledge stays valuable. What is missing is a layer that did not need to exist before: a shared company context that holds across departmental boundaries.
🦋 Not a replacement for departmental knowledge, but a connective layer above it. What applies to everyone: company profile, communication style, core processes, products and pricing.
🦋 Maintained, not documented once. Prices change, products come and go, processes get simpler. A living foundation, not a filing system.
🦋 With a responsible person who approves drafts. Uncoordinated context is worse than none.
🦋 Start in one area, not across the whole company at once. Where the benefit becomes visible fastest: often pricing, offers, communication style.
What this looks like in practice
A manufacturing company rolls out AI assistants in sales and service. After three months the leadership team notices: the answers are good, but inconsistent. Customers receive different information about lead times and pricing.
The cause is not technical. Both departments briefed the assistant as well as they could. They simply used different sources.
The organisation does not respond with a new tool. It builds a new layer: a shared document covering what applies across departments. Offers, pricing, standard processes. One person is responsible. Departments contribute drafts. Four months later the inconsistencies are gone. (This example is fictional.)
Benefits at a glance
🦋 Consistent AI responses across departmental boundaries.
🦋 Knowledge that stays in the organisation even when people leave.
🦋 New colleagues work from the current foundation from day one.
🦋 Changes are updated once, for everyone.
Conclusion
Department-owned knowledge as an organisational principle was right for its time. Working with AI, the same structure produces systematically worse results. The tool has changed; the organisational principle has not. What got us here will not get us further. Departments stay. What gets added is a new layer: knowledge that holds and is maintained across departmental boundaries.
👉 Where in your organisation does most of the knowledge still sit within individual departments? We are happy to discuss where a shared layer would make the most difference. Get in touch via the contact form.
(AI generated)