AI only sees what it can access
Many have rolled out Copilot or ChatGPT and see little effect in daily work. Usually the model is not the problem. The information it needs sits in five different systems.
Central X
Most businesses already have the systems they need. The problem is that they do not talk to each other. People move information by hand, nobody sees the whole picture, and AI delivers less than you hoped. We develop Central X, the core that connects your systems and makes AI useful, for businesses, organisations and municipalities.
AI does little as long as your data is scattered across email, spreadsheets and line-of-business systems. An AI assistant can only answer from what it can access. Central X brings together the information from the systems you already use, so AI can work on your own data, with control over what it sees and does.
Many have rolled out Copilot or ChatGPT and see little effect in daily work. Usually the model is not the problem. The information it needs sits in five different systems.
Once the information is connected, the same core can interpret documents, answer questions about your own data, draft replies and flag what slips. You start with one task, automate it step by step and add more.
You decide what AI can see and do, and everything it does is logged. We agree which AI services are used and where data is processed before anything goes live. Data is stored in the EU/EEA.
Businesses
Jobs, customers, hours and invoice data in one picture, with AI doing the groundwork.
Organisations
Members, cases, documents and knowledge in one place, so nothing depends on the right person being at work.
Municipalities
Cases and information across line-of-business systems, with AI that interprets enquiries and drafts replies. We know the public sector requirements from apps we have developed for Oslo and Øygarden municipalities, among others.
Then there is usually more to gain than you think. It does not mean anything is wrong. The routines work, people know their jobs, and the work gets delivered. But much of the capacity goes into moving information between spreadsheets, email and line-of-business systems.
Customers call to ask about status, and people spend hours answering.
The same information is entered several times, by several people.
Documentation is scattered across folders, email and phones.
Completed work and materials do not always make it into the invoice data.
Your systems do not talk to each other, so someone moves data by hand.
You have rolled out AI tools, but see little effect in daily work.
The core pulls data from the systems you already use, where they offer an API. You choose how much it should do, and how much should still be approved by a person.
Most start at level one and move up as they see what it delivers. You do not need to decide now how far it should go.
Your systems stay as they are today. We develop a layer on top that pulls in data and keeps the overview up to date. The lowest threshold, the fastest effect, and you do not replace anything that works.
A fit when the systems themselves are good enough, but nobody sees the whole picture.
The same core, but we also develop what is missing entirely. Often a mobile view for people out on jobs, digital forms, or invoice data compiled automatically from the entries.
A fit when parts of the workflow lack a system.
The core grows module by module, and gradually takes over tasks that today sit in separate tools for project management, tasks and internal communication. You decide the pace, and how far it should go.
A fit when you pay for many tools that each solve one piece.
| Today | With a core |
|---|---|
| Customers call to hear where their case stands | Customers log in and see status, documents and progress themselves |
| Forms on paper and in PDF | Digital forms filled in where the work is done |
| Invoice data is put together afterwards | The invoice data is compiled from entries along the way |
| Documents in several versions | One source, with history per customer, project or unit |
| Manual keying of incoming email and PDFs | AI interprets the content, a person approves |
| No combined overview | Real-time status for management |
| Everything sits in the heads of a few people | The knowledge sits in the system, and follows the job |
AI is most useful when it has access to combined, up-to-date information. In the core, it does. We start with AI suggesting and a person approving, and you decide when more can run automatically.
Emails, PDFs, orders and invoices are read and turned into suggested entries. A person checks and approves, instead of typing everything in themselves.
“What was agreed with the customer in March?” The core answers and shows where the answer comes from. Knowledge no longer depends on the right person being at work.
Status reports for management, and replies to customers wondering where their case stands, are drafted from the data in the core. Someone reads them through and sends them.
The person out on a job dictates on their phone. AI fills in the form, and the person confirms before it is saved.
The core alerts you when a case has stalled, or when hours and materials are missing from the invoice data. Then it can be corrected before the invoice goes out.
What AI does is traceable, and you can see where a suggestion came from. Before anything is put into use, we agree together which AI services are used, where the data is processed, and that it is not used to train the models.
A short chat about how you work today, and where there is most to gain in overview and automation. If your current solution is good enough, we will say so. No cost.
For larger needs, we map one process area from start to finish, together with the people who know it. You get a solution outline, a costed plan and a fixed-price offer. The document is yours regardless, even if you choose another supplier.
At a fixed price, on the scope that has been described. You follow the progress along the way. The first version is production-ready, with access control and logging in place. Who owns what is set out in the agreement before we start.
New modules are added where the bottlenecks show up, not where we guessed in advance. We run the solution on a fixed agreement: monitoring, updates and bug fixes.
If it turns out that an off-the-shelf system covers your needs better, we will tell you. See also system development and PWAs for people in the field.
See also all frequently asked questions about price and process.
Tell us briefly how you work today. Together we will see whether there is anything to gain in overview and automation, and we will tell you if your current solution is good enough. That conversation costs nothing.

