Skip to content

REF / DEPLOYMENT MODEL

Cloud or on-premise EDMS: how to actually decide

This question is usually settled by habit or by an assumption nobody has checked. It is worth twenty minutes, because it changes your cost shape, your AI capability and who carries the operational burden for the next five years.

REF / SIDE BY SIDE

The three models on the dimensions that matter

EDMS by Sibasi deploys in all three. We have no incentive to push you towards one, which is why this page is written the way it is.

Dimension Microsoft 365 cloud On-premises Hybrid or sovereign
Time to deploy Fastest. No infrastructure to procure, rack or harden. A Basic Deployment can be live in four to six weeks. Slower. Server procurement, sizing, hardening, backup and DR design all sit on the critical path. Cloud timelines for the collaboration tier; on-premises timelines for the classes held locally.
Where data sits The Microsoft cloud region selected for your tenant, or a dedicated Azure region where you need a specific country. Entirely on hardware you own, in a building you control. The strongest answer to a hard residency rule. Sensitive record classes stay local; everything else uses the cloud tier.
AI capability Full. Classification, extraction, OCR at scale and Copilot all depend on cloud AI services. Reduced. Core document and records management is unaffected, but cloud AI services are not available to a disconnected deployment. Available for the content that lives in the cloud tier. This is usually the deciding factor.
Cost shape Licences you already hold, plus the one-off deployment. No server refresh cycle. Deployment plus hardware, data-centre capacity, and a refresh every few years. Higher total, more control. Between the two, with the added cost of running and testing two environments.
Availability and DR Microsoft's own resilience and service commitments apply. Recovery objectives inherit from the platform. Yours to design, build, fund and test. A documented RPO and RTO is only real once a restore has actually been run. Two DR designs, and both need testing.
Ongoing effort Patching, capacity and platform upgrades are Microsoft's problem. Yours. Budget for the administrator time honestly rather than assuming existing staff absorb it. Highest of the three, because you carry the on-premises burden and the integration between tiers.

REF / DECISION

Five questions that settle it

Answer these honestly and the model usually picks itself. Most deployments that go wrong here got question one or question four wrong.

  1. 01

    Does a written law, regulation or policy require the data to stay in-country?

    If yes, and a dedicated Azure region in an acceptable jurisdiction does not satisfy it, you are on-premises or hybrid. If the requirement is assumed rather than written down, check it before it costs you the AI layer.

  2. 02

    Is AI classification and extraction central to the business case?

    If the case rests on eliminating manual indexing across a large backfile, a fully disconnected deployment removes the thing you are buying. Hybrid is usually the answer.

  3. 03

    Do you already run Microsoft 365?

    If your staff are already in Exchange Online and Teams, the collaboration data is already in the cloud. An on-premises document repository beside it is worth examining rather than assuming.

  4. 04

    Who will run the servers in year three?

    On-premises is a standing operational commitment: patching, capacity, backup verification, DR testing. If that team does not exist and is not being hired, the honest answer is cloud.

  5. 05

    What is your connectivity like at the edge?

    Branch and county offices on constrained links behave differently. In practice this argues for offline sync and low-bandwidth loading rather than for on-premises, but it must be designed for.

NOTE Most common outcome

Hybrid is chosen more often than either extreme

In practice most regulated institutions land on hybrid: the classes with a hard residency or sensitivity constraint stay on infrastructure they control, and everything else runs in the cloud where the AI and the collaboration tooling are. The trade-off is documented in the solution design before configuration starts, so nobody discovers it after signature. See security and compliance for how residency is handled, and implementation for where this decision falls in the delivery.

REF / NEXT STEP

Work through it with an architect

Bring your policy, your connectivity picture and your internal capacity. An hour here saves a deployment model you regret in year two.

Book an architecture session See pricing