Local AI for everyday office work
What do you currently do by hand: copy details from enquiries, look up prices or write quotes? We pick one task and test it on your company computer or server. You check the results.
Review the proposed information before exporting or sending it.
A fictional example
“Please quote 120 cardboard boxes, each 40 × 30 × 20 cm. We would like delivery by 22 Sep 2026.”
Check the row before you download it.
Colleagues search from their desks. The application looks on the office server, in files they have permission to read.
Choose a task for us to test on your equipment. These three fictional examples use prepared answers.
The app puts the customer’s request into a table. You check what’s missing before confirming the order.
“Please send 20 pairs of gloves in sizes M and L.”
Of the 20 pairs, how many should be medium and how many large?
Ask, for example, what payment term appears in a contract. The application shows the exact passage it used for the answer and searches only the files you have permission to read.
How long do we have to pay this invoice?
Within 15 days of receiving it.
Supply agreement, section 3Payment is due within 15 days of receiving the invoice.
A customer calls to reschedule. The app turns your phone note into a short handover for the colleague who will call them back.
“Customer asked for Thursday at 10. Not sure yet if anyone else is coming.”
Requested time: Thursday, 10 am. We don’t know how many people are coming. Check the calendar, then call the customer to confirm.
The application picks out the item, quantity and delivery date from the customer’s message.
If the customer leaves out the quantity, the field stays blank.
Try it yourself: review a request.
We can start with a front desk booking form, a workshop estimate or the spreadsheet you use for orders. We keep the fields and terms your colleagues know. In the assessment, we time the task with and without the app, including checks and corrections.
We define the fields, templates, output format and review steps. The application enforces permissions and approvals; model instructions are not a substitute for those controls.
You choose which files the application can search. It shows the passage it found and links to the source. When a price list changes, you can update the file without retraining the model.
We can adapt an existing model using checked examples of a repeated task, such as sorting requests or drafting in an agreed format. We compare the result with the initially configured application, using separate examples not used for training. We keep the simpler setup if training does not help. We do not build a model from scratch.
Before any adaptation, we agree which data is used, where it is processed, who can access it and how long it is kept. A local installation does not automatically mean training or support also happens locally. We do not transfer data outside the agreed boundary.
We check the document volume, how quickly you need results and how many people will use the application at once. The test helps us choose suitable equipment.
We start with the equipment you have. If it cannot handle the chosen task, we explain what needs changing and why.
In an installation verified to run fully on-site, documents, searches and answers are processed on the approved equipment, without sending them to an external AI service.
Downloads, updates, backups and remote support may use the internet. We agree separately on how they work and which data can be accessed.
We start from the computer you already have. What matters is the documents you process and how many colleagues use the application at the same time.
| Setup | Approximate number of users | What we assess |
|---|---|---|
| Your existing computer | 1-5 users | We check the processor, memory and whether the task needs a dedicated graphics card. |
| A dedicated office computer | 5-25 users | We test use by several colleagues and check what equipment can handle their requests. |
| Several computers | 25+ users | We check the number of users, data access rules and how work resumes if one computer fails. |
Tests establish the supported user count and speed. Any equipment and licence costs are listed separately in the proposal.
We start with one activity and agree how to check the result before installation.
We look at what repeats, how many documents there are and which software you use now.
We configure the application and test it on the intended equipment. Further model training is evaluated separately if testing shows it would help.
We install it on the agreed equipment and prepare an export or connection supported by your existing software.
Your team learns how to review results and what to do if the application or computer is unavailable.
Updates and ongoing support can be agreed separately based on what needs maintaining.
The answer depends on the task, documents and computer. These are the things we check.
We measure the time needed to process, review and correct test documents. We check the processor, memory, graphics card and use by several colleagues. The proposal records the requirements and test results.
Some tasks may run on your existing computer without a dedicated graphics card. We check its processor, available memory and usual workload, then test representative documents with the number of people who would use the application at the same time.
They can, if the task and equipment support it. One office machine may serve several colleagues, but we check user count, document access and what happens when requests arrive at the same time.
If the program runs on that computer, you cannot use it while the machine is off or broken. Your team needs a manual way to work and good backups. If many people use it, we can look into setting up a dedicated office computer.
We can assess PDFs, scanned documents, Word files, spreadsheets, CSV, email and document photos. We test the language, format and image quality before agreeing which files the application will support.
Not necessarily. We configure the application first and test representative documents. Further training is optional. We compare both versions on examples not used for training and keep the one that handles the task better.
It means showing the people who use the application how to check results, what to approve and what to do when information is missing. It does not mean training the model.
Not necessarily. Everyday use of the application and model training can require different equipment. If training needs other resources, we agree in advance where it takes place and where the data goes. We never transfer data without your agreement.
No. Adding approved documents may update what the application searches, but it does not automatically train the model. Search and additional training are different processes.
No. The application searches approved files and shows the passages used in its answer. You can update documents without changing the model. Further training is optional and evaluated separately.
A missing field should stay blank or be flagged. Search should show the source document or say it could not find one. Someone reviews and approves information before it is exported, sent or entered into another system.
Before work begins, we agree which examples are needed, where they are processed, who can access them and how long they are kept. A local setup for everyday use does not automatically make training, maintenance or support local too. We do not move data outside the agreed boundary.
An installation verified to run fully on-site can process documents without an internet connection once its components are installed and tested. Updates and software downloads may need internet access. We agree separately on online backups or remote support; there is no hidden switch to an online AI service.
We check how your existing software accepts and exports data. A file import may be enough, or the applications may need a direct connection. When data is already structured, we use ordinary processing rules where AI is not needed.
At handover, you get instructions for using the application and returning to manual work. We agree separately who handles updates, backup checks and ongoing support.
No. A public chatbot is a separate project, with information intended for visitors, access rules and availability requirements. It does not get access to private company files.
The application can prepare data for review and export. The accountant or responsible person checks and approves entries. It does not make tax decisions or approve payments. Any connection to accounting software is agreed separately.
Local AI is one part of our work. We also build internal tools, websites and mobile apps, automate agreed workflows, modernize existing systems and help with security and maintenance.
Enter the number of people who handle documents, their daily hours and their hourly cost. Adjust the percentage to see what that time could be worth.
The €900 assessment pays for a test on your equipment. A custom app for one task starts at €3,500. You can request a quote without an assessment. Websites, repairs and other work are priced separately.
€900 credited against implementation
€900
Estimated duration: 2 weeks
We pick one time-consuming task and test it on your company equipment. You find out what the application can do, whether you need another computer and what the full project would cost.
Read a sample assessment and test resultsRequest an assessmentOne complete workflow
From €3,500
Estimated duration: 4 to 6 weeks
We automate one document task or build a first version of an internal application. This includes review steps, user access, installation, testing and showing your team how to use it.
Request an application quoteConnect several tasks and applications
From €8,500
Estimated duration: 6 to 12 weeks
We connect the application to the systems in your quote once we’ve checked they can exchange data. We test backup recovery, document how it works and hand it over to your team. Ongoing maintenance is agreed separately.
Request a quoteThe €900 assessment fee is deducted in full if you commission the assessed project. Equipment, licences and optional maintenance cost extra. The proposal records the work, schedule and credit terms.
Tell us what comes in, what you need at the end and roughly how often you do it. A description is enough. Please leave out documents and customer details.
