Blog · 17 November 2026 · 6 min read
Maintaining a local AI: support or new development?
Anyone considering a local AI often fears that, along with the computer, they are buying a lifelong tie to whoever installed it. That is not the case, and it is worth clarifying before signing: what is really needed after go-live, what falls within a support contract and what is a separate job.
In short: no, you don't have to pay someone for ever. Once installation is finished, the system is yours and runs by itself: support is optional and is for those who want someone to keep it up to date and step in if it stops working. Keeping what has been installed working and up to date is support. Any addition is new development: one more connection, user, archive or feature.
Do I have to have a support contract after installation?
No. A local AI built with open source programs keeps working even if nobody touches it again: there are no licences that expire and no supplier services it depends on. Chat, models and document archive stay on your machine, with the administrator credentials in your hands.
Support is therefore a choice, not a condition. It makes sense if there is nobody in the company who wants to deal with updates and checks, or if you prefer to have a point of contact who steps in when something goes wrong. If, on the other hand, you have an in-house technician, or another supplier you trust, they can take over the system using the handover documentation.
What happens without maintenance? In the short term, nothing. Over time, however, software and models age: you miss out on security fixes, new features and better models.
What does support usually include?
Keeping what has been installed in good health. In practice, four activities:
- Updating the software. Ollama, the engine that runs the models, and Open WebUI, the chat interface, publish new versions all the time: the former every few days, the latter every few weeks. It doesn't pay to install each one as soon as it comes out. You update at scheduled intervals, after reading the release notes and with a fresh backup in hand.
- Assessing new models. Every few months, more capable open models come out for the same amount of memory. At each intervention we check whether one of them improves the company's work and, if so, replace the one in use.
- Checking overall health. Disk space, backups completing successfully, document syncing, connections that respond.
- Stepping in when something stops working. If the chat or the engine stops responding, in most cases they restart by themselves. When that isn't enough, the technician intervenes remotely, after you have opened up access.
A small allowance of time for questions about use is usually included too: how to set up an assistant, why an answer is not the one you expected.
Tip: for shared installations, the Open WebUI documentation recommends pinning a specific version instead of always following the latest one, and making a backup before every update. Ask the supplier whether that is how it works.
What, on the other hand, counts as new work?
Everything that adds something to the system. The distinction is simple: support maintains, development adds.
| Request | Support | New development |
|---|---|---|
| Updating installed software and models | Yes | |
| Restarting or restoring the system after a fault | Yes | |
| Checking that connections and automations work after an update | Yes | |
| Connecting a new data source (email, management system, chat) | Yes | |
| Adding users with a separate environment | Yes | |
| Creating a new archive or a new automation | Yes | |
| Adapting a connection because an external service has changed | To be agreed |
The last row is the one that causes the most discussion, and it is best to clear it up straight away. If a connection stops working because Google, Meta or the vendor of the management system has changed its rules, it is not a fault in the installation: it is an adaptation to a decision taken by others. It should be estimated and agreed before it is done.
What can a company do on its own?
A great deal, if there is someone with a little technical familiarity. Day-to-day management does not call for programming skills:
- Creating and removing accounts, and deciding who sees which archives, from the interface's admin panel.
- Adding documents: whatever is saved in the connected folders goes in automatically.
- Creating assistants with their own instructions, by department or by task.
- Checking that the backup has been made and how much space is left on the disk.
- Restarting the machine when needed.
Systems skills are needed, on the other hand, for program updates, replacing models, the private network and restoring from backup. An in-house technician can do these, or any supplier familiar with these tools, which are among the most widely used in the field.
What support does not cover
Even the best contract has boundaries, and it is only right to know them beforehand:
- Hardware faults. Computers and disks are covered by the manufacturer's warranty, not by software support. That is why you need a backup of the whole machine.
- Third parties' decisions. If an external service withdraws a feature or starts charging for it, no supplier can guarantee that the connection will stay as it was.
- The quality of the answers in absolute terms. An update can improve a model, but local models have limits of their own: on some tasks the big cloud services remain superior.
- Hours. A standard support service picks up requests within one working day, during office hours. Anyone who needs interventions at night or on public holidays must ask for it expressly, and it costs more.
These are points to have written into the contract, together with the number of scheduled interventions per year and the response times.
How it works with Dabryx
We install local AI on site, throughout Italy, and on handover the system is complete: it works without us. We leave you the documentation and the credentials, so you are free to manage it yourself or with whoever you prefer.
If you would like us to keep looking after it, there is optional annual support, for a modest fee: scheduled update and check interventions, assessment of new models, restoration in the event of a fault and answers to questions about use. Requests are picked up within one working day. Any addition is estimated separately, before starting.
The service is described on the Local AI for businesses page. To find out what would be needed in your case, write to us.
About the author: Dabryx
Dabryx is an Italian software house that installs local AI in companies, on site across Italy, on servers or Macs. The basic installation starts at €2,900, hardware not included, and consulting at €100 per hour. The office is in Rocca Priora, near Rome.
We also install local AI in Rome, Milan, Parma and Reggio Emilia, and we develop custom management systems and custom software.
Frequently asked questions
Does the local AI stop working if I do not renew support?
No. The system belongs to the company and does not depend on the supplier's licences or services: it keeps working. Without maintenance, over time, programs and models stay at the version installed.
How often does a local AI need updating?
There is no need to chase every version. Scheduled interventions over the course of the year are enough, in which the programs are updated, new models are assessed and the general state is checked, always after a backup.
Is adding a connection, for example to email, support?
No, it is new work: it adds a function that was not there before. It should be estimated and agreed separately. Checking that the connections already installed work after an update, on the other hand, is support.
Can another supplier take care of maintenance?
Yes. The programs used are open source and widely adopted, and with the credentials and handover documentation another technician can take over the system.
Want an estimate for your company?
Tell us how many people will use the AI and what for: we'll tell you which configuration you need.
Response times
Within 2 hours on business days (9:00–18:00) for clients with a support contract.
AI consulting
Basic installation from €2,900, consultancy from €100 per hour, on your company's server or Mac.
Where we work
Throughout Italy: we come on site to the client's premises and also support you remotely.