Skip to content
Buyer's guide

Managed on-premise AI: what "managed" should include

Updated
By
The MagnolIA team
On this page
7 min read

Short answer

On-premise AI is "managed" when the provider, not your staff, looks after the hardware, the models, updates, monitoring, support and security, with the machine installed in your office. The word has no official definition, so it is worth checking what each contract includes. This MagnolIA guide sets it out as twelve questions.

Back to contents

What does it mean for on-premise AI to be managed?

It means the AI runs on a machine inside your network, as with any on-premise AI, but the work of keeping it running does not fall on your people. "Managed" describes a division of labour, not a technology, and each provider draws the line in a different place.

Buy and operate
You buy the hardware and your staff install, configure and maintain everything. This is the model described in building your own AI server.
Software with support
A provider gives you the software and responds to incidents, but the hardware, the models and day-to-day operation remain yours.
Managed
The provider supplies the hardware or chooses it with you, installs it, maintains the models, watches the service and responds when it fails. Your people use the AI.

Between the second model and the third there are many shades. That is why the useful thing is to ask, area by area, who does what, and to get it in writing.

Back to contents

What should you ask a provider? A twelve-question checklist

They work for any provider. They follow the seven areas where a service can fall short: hardware, models, updates, monitoring, support, security and exit from the contract.

  1. Hardware: who owns it and who sizes it?

    Whether you buy it or it is included, who decides the capacity and on what usage data.

  2. Hardware: what happens if it breaks or falls short?

    Who repairs or replaces it, how quickly, and whether adding capacity means stopping the service or paying more.

  3. Models: who chooses them and under what licence?

    Which models are installed, who tests them and whether their licence allows your use.

  4. Models: who changes them and how do you find out?

    A new model answers differently. Ask how notice is given, whether you can try it first and whether you can go back to the previous one.

  5. Updates: when, and with what downtime?

    Who updates the software and the operating system, how often, and whether there are agreed windows.

  6. Monitoring: what is watched and what leaves the machine?

    What data travels to the provider, through which channel, and whether any of it is the content of conversations.

  7. Support: who answers and how quickly?

    Language, hours, channel and response times, in writing.

  8. Security: how does each user sign in and who administers?

    Individual identity, roles, corporate single sign-on and two-factor authentication for administration.

  9. Security: what is logged, and where?

    Per-user logging, how long it is kept, and whether it sits in your installation or the provider's.

  10. Security: how are backups made?

    What they include, how they are encrypted, where they are stored and when a restore was last tested.

  11. Cost: what does the fee include and what is charged separately?

    Hardware, software, models, support, integrations and expansions: what is in and what is not.

  12. Exit: what happens to the hardware and the data at the end?

    Whether you keep the machine or return it, how you get your data back and how whatever remains is erased.

Back to contents

What is the provider's role in data protection?

Having the machine in your office does not make the provider disappear: it is still a company you sign a contract with, and it may have some access to the machine. Article 28 of the GDPR governs processors, meaning whoever processes personal data on someone else's behalf. Whether a managed provider is one depends on what access it has and what it can see; that is for your data protection officer to judge.

That is why questions 6 and 9 carry more weight than they seem to: a provider that only receives the state of the machine is not the same as one that can read conversations. If you work for the Spanish public sector, add the questions in AI and Spain's ENS.

Back to contents

What does MagnolIA include in each area?

Only what MagnolIA's website states. Where the website is silent, the table says so, so that you can ask.

What MagnolIA's website states for each area of the checklist
AreaWhat is statedWhat the website does not detail
HardwareIt is hardware as a service: included in the fee. MagnolIA chooses it according to your usage, installs it in your office, maintains it and renews it. There are four tiers, and capacity is added without interrupting the service.Repair or replacement times.
ModelsOpen-weight models that MagnolIA selects, tests and updates, with profiles per task.Which specific models are installed at any given time.
UpdatesOperation, hardware maintenance and model updates are MagnolIA's responsibility.The update schedule and windows.
MonitoringContinuous. Only the state of the machine travels out: load, health and security. MagnolIA does not see prompts, documents or answers.Nothing relevant to this checklist.
SupportIncluded in the fee, in Spanish.Hours and response times.
SecurityA separate user for each person, role-based permissions, two-factor authentication for administration, TLS-encrypted entry, per-user usage log, a chained audit trail kept for two years in your installation, and AES-256 encrypted backups in the storage you choose.If you need a specific certification: MagnolIA does not advertise any under Spain's ENS.
CostA fixed monthly fee per capacity tier, which does not depend on consumption. It includes hardware, software, models, support and monitoring, integrations with your systems and added capacity.The figure: it is agreed with each company.
ExitAt the end of the contract you choose between keeping the hardware, renewing it or returning it. The data stays in your installation throughout the contract.How data is erased if you return the hardware.
The middle column summarises what MagnolIA's website states in November 2026. The third is not a gap in the service: it is what the website does not publish and is worth requesting in writing.

What the fee covers, and when it is not worth it, is in how much on-premise AI costs. The types of provider on the market are compared in the comparison of on-premise AI providers.

Back to contents

When is managed AI not the right choice?

  • You have a technical team that wants control. In a managed service, users have no technical access to the system; if you want to choose every component, build it yourselves.
  • You are exploring. To learn what a local model can do, a tool such as Ollama on one machine is enough.
  • Few people will use it, occasionally, with information that is not sensitive. A fixed fee may come out above what you spend on licences.
  • The work depends on a proprietary model that is only offered in the cloud.

If none of these applies to you and you would like the twelve answers with your own situation on the table, the next step is to request a pilot.

Back to contents

Frequently asked questions about managed on-premise AI

What is managed on-premise AI?

Managed on-premise AI is on-premise AI where the provider handles everything technical. At MagnolIA that means we choose the hardware, install it in your office, select and maintain the models, monitor the service and provide support. Your team does not administer servers or models: it uses the AI.

What should a managed on-premise AI service include?

A managed on-premise AI service should include the hardware and its installation, model selection and updates, the software for users, permissions and logging, monitoring, support, backups and an agreed exit from the contract. MagnolIA's fee includes the hardware, the software, the models, support, monitoring, integrations and extra capacity as your team grows.

What can the provider see in managed on-premise AI?

What the provider of a managed on-premise AI service can see depends on the service, and it is worth getting in writing: what data leaves the machine, through which channel and who has access to it. In MagnolIA's service, only monitoring of the machine's state (load, health and security) travels out; prompts, documents and answers stay in the customer's network.

How is managed on-premise AI different from buying an AI server?

When a company buys an AI server, it pays for the hardware and takes on installing it, choosing models, setting up users, updates and repairs. With managed on-premise AI the machine still sits in the office, but the provider takes on that work in exchange for a fee. At MagnolIA the hardware is included: hardware as a service.

What should you ask before signing up for managed on-premise AI?

Before signing up for managed on-premise AI, ask who owns the hardware and who repairs it, who chooses and updates the models, what the provider can see, what log is kept and where, how backups are made, what support there is and what happens to hardware and data at the end. MagnolIA's guide sets this out as twelve questions.

Sources

External figures on this page link to their original source.

  1. EUR-Lex. Regulation (EU) 2016/679, General Data Protection Regulation (GDPR). eur-lex.europa.eu/eli/reg/2016/679/oj

Ask us the twelve questions

Tell us how many people would use the AI and with what data, and we will answer them one by one with your case on the table.

Request a pilot