Modernising IT in an architecture practice: BIM, AI and cloud without changing everything

IT is involved in almost every stage of a project. During design, architects produce drawings, models, options and visualisations. During the planning stages, they exchange information with civil engineers, building services specialists, contractors and clients. On site, they consult documents, track outstanding issues and write meeting minutes. And for years afterwards, they must be able to retain, retrieve and pass on information.

IT therefore amounts to more than computers and licences: it is a chain linking people, data, applications and business processes. If just one link works poorly, the whole project may slow down.

Example: a well-equipped but inefficient practice

Consider a fictional practice with 18 staff using Revit, AutoCAD, Microsoft 365 and a local file server. The workstations are recent, the server has spare capacity and the licences are up to date. Yet:

• teams waste time finding the right version of a file;
• remote staff suffer slow synchronisations;
• external engineers receive documents by email rather than through a structured sharing space;
• meeting minutes are written manually, then copied into several tools;
• backups run, but restores are never tested.

Should the computers be replaced, everything moved to the cloud or an AI solution purchased? Not before identifying the causes. Here, the workstations are not the problem: this practice mainly needs to organise its workflows and make its existing systems reliable.

Modernisation starts with a review of the current situation. This need not be a multi-week audit, but it should be a structured analysis of the problems affecting production, covering five areas.

The aim is a list of problems, with their frequency, consequences and likely causes. Modernisation can then follow business priorities rather than the latest market developments.

Adopting Revit, Archicad or a collaboration platform does not guarantee smooth projects. Collaboration quality also depends on how models are organised, responsibilities and information management.

The standards framework: the SN EN ISO 19650 series

In Switzerland, the reference for BIM information management is the SN EN ISO 19650 series. SIA technical specification 2051, long cited as the basis for the BIM method, was withdrawn at the end of 2024 precisely because this international series partly replaces it.

ISO 19650 defines the common data environment (CDE): the agreed source of project information, where each document exists once, with a known status (work in progress, shared, published, archived). A CDE is first a process; one or more software tools then implement it.

Revit and Archicad: the collaboration method matters as much as the server

For Revit, Autodesk distinguishes three approaches that are not interchangeable: file-based worksharing (local network only), Revit Server (multiple sites) and Revit Cloud Worksharing, included in the Autodesk BIM Collaborate Pro subscription. For Archicad, Teamwork collaboration relies on BIMcloud, a server application that must be sized as such.

We explain these constraints (latency, VPN, NAS and BIMcloud sizing) in our article Slow NAS in an architecture practice: diagnosing Revit, Archicad and storage.

The mistake to avoid: buying a platform without reviewing methods

A practice can have an excellent collaboration tool and still exchange important files by email, or keep several competing copies of the same documents. Before investing, define:

• who produces information, and who approves it;
• how revisions are managed and which version is authoritative;
• which documents may be shared with partners;
• how decisions are recorded.

BIM is more than a digital model: it is a working method and information governance.

Cloud offers real benefits: remote collaboration, multi-site access, scalability and less hardware to operate. It also creates dependencies: Internet connectivity, provider availability, subscriptions, access performance and data retrieval conditions. Local infrastructure, meanwhile, involves hardware, maintenance, backups and a replacement plan. The right choice depends on usage.

Three illustrative scenarios

These examples illustrate a way of thinking; the actual decision requires a diagnosis.

Cloud does not replace backup

Microsoft 365 follows a shared responsibility model: Microsoft provides service availability, but protecting and restoring data remains the customer's responsibility. Recycle bins, versions and retention policies serve compliance, not recovery after ransomware or an administration error.

A dedicated Microsoft 365 backup (Exchange, OneDrive, SharePoint, Teams), stored outside the tenant, therefore remains necessary. The same logic applies to BIM platforms: with Autodesk, documents stored in the cloud remain available for only 30 days after the subscription expires.

Sovereignty: what hosting in Switzerland really means

Since 2019, Microsoft has offered Microsoft 365 from two Swiss regions, in Geneva and Zurich, with data at rest stored in Switzerland. That is an advantage, not a complete guarantee of sovereignty: a provider subject to US law remains affected by the CLOUD Act, which can require it to disclose data hosted outside the United States.

The revised Federal Act on Data Protection (FADP), in force since 1 September 2023, also requires security breaches to be reported promptly to the Federal Data Protection and Information Commissioner (FDPIC). A practice must therefore know where its personal data is held, who can access it, which processors are involved and how to retrieve its data when changing provider.

The right architecture is not the one with the most cloud services, but the one that meets the practice's constraints with controlled risks and costs.

AI offers real opportunities for architecture, but not every use case has the same maturity or profitability. Administrative, documentary, creative and technical uses must be distinguished.

Before deploying document AI, you need to know which documents are authoritative.

Calculate returns across the complete process

Illustrative example: a practice writes 60 sets of minutes a month, taking 35 minutes each. With an AI assistant, the complete time, including review and correction, falls to 20 minutes.

• Saving: 15 minutes × 60 = 15 hours per month.
• At a fully loaded internal cost of CHF 75 per hour: CHF 1'125 per month, or around CHF 13'500 per year of theoretical capacity.

This is not yet a saving: licences, integration, training and maintenance must be deducted, then the practice must check what it actually does with the time released. Four indicators are enough to manage this: complete processing time, document quality, number of corrections and capacity actually reassigned.

AI that produces a document in thirty seconds but requires twenty minutes of correction must be assessed across the complete process.

A practice handles confidential drawings, information about sensitive buildings, personal data, contracts and documents belonging to its clients. None of this should be freely copied into just any AI assistant.

The number one risk: shadow AI

The most common danger is not an official AI project, but a colleague pasting a specification or contract into a personal ChatGPT account to save ten minutes. The practice then loses control: consumer terms of use, data retention and no access controls. Banning it without an alternative pushes the practice underground; providing an approved tool and clear rules is better.

Microsoft 365 Copilot: clean up permissions first

For a practice already using Microsoft 365, Copilot is often the most natural option. It only shows each user data for which they have at least read access. But that is precisely the trap: Microsoft points out that Copilot uses all data accessible to the user, including widely shared files whose existence they do not know about.

An HR folder or fee information shared too widely three years ago suddenly becomes discoverable with one question. Before enabling Copilot, Microsoft therefore recommends identifying and correcting oversharing in SharePoint and OneDrive (“Everyone” links, sites open to the entire organisation, inherited permissions).

Define an AI policy

A short internal policy should specify:

• approved tools, and the prohibition on personal accounts for practice data;
• permitted categories of data (public, internal, confidential client data);
• processes requiring human approval and who is responsible;
• rules for integrations connected to document spaces.

Before selecting a platform, also examine data processing, retention, processors, deletion and export options, and confidentiality commitments made to clients.

What about local or private AI?

Hosting a model in a controlled environment provides greater technical control, but brings its own costs: hardware, maintenance, security, computing capacity and model quality. It is not automatically the best option; the choice depends on the confidentiality required, expected performance and budget.

The aim: make AI useful without turning project information into an additional risk.

A practice's IT contains years of work. A server failure, accidental deletion, compromised account or ransomware can affect ongoing projects and their deliverables. Security must therefore be built in from the outset.

The five priority protections

What should an architecture practice back up?

Far more than RVT, PLN or DWG files: libraries, families, templates, textures, linked models, application settings, BIMcloud server data, contractual documents and Microsoft 365 content. Where data lives on an external collaboration platform, check what the provider actually backs up and what remains your responsibility.

What recovery time is acceptable?

A small practice may tolerate its archives being unavailable for a few days. Prolonged loss of access to active projects, however, can cost a submission or competition. These objectives are agreed with management, then translated into realistic technical measures.

Modernisation must do more than make projects easier to open: it must ensure the practice knows how to recover them.

Over the years, a practice's workstations diverge: software versions, Revit add-ins, graphics drivers, network settings, libraries, authentication methods and file locations. Some differences are justified; many simply reflect successive changes.

A controlled configuration makes installations reproducible, incompatibilities easier to diagnose and updates testable before deployment. Staff gain a consistent environment, and support interventions become faster.

Profiles rather than a single machine specification

This segmentation avoids two opposite mistakes: under-sizing production workstations and buying expensive machines for uses that do not need them.

Inventory-based replacement

Keep an inventory including each device's age, warranty, end of support and observed performance. A capable, maintained workstation stays in service; one that has become incompatible with required software is replaced or reassigned to a less demanding role.

The best hardware investment is not the one that replaces the most workstations, but the one that removes the limitations actually measured.

The amount depends on existing systems, staff numbers, applications and objectives. Rather than a shopping list of technologies, a useful budget is structured into categories that support trade-offs.

Compare over three to five years

A local server requires greater upfront investment; a cloud platform turns some costs into recurring subscriptions. An honest comparison must include licences, energy, hosting, maintenance, backup, migrations, support and administration time, then set these against the value created. A more expensive solution may be justified by productivity gains; an attractive technology may be no more than an expense if it solves no important problem.

Prioritisation example

Management has CHF 25'000 available over twelve months (an illustrative assumption, not a quotation). Three projects are being considered: replace ten computers, modernise storage, improve model collaboration.

The diagnosis shows that the workstations remain suitable and slowdowns mainly come from storage and an unsuitable sharing workflow. Investing around CHF 40'000 in ten new workstations would then be unjustifiable: addressing the measured causes and planning only the necessary replacements is more rational.

Successful modernisation is not about spending the entire budget, but using it where it produces the most useful results.

A technical improvement is not necessarily a business benefit. A new server may show better figures without any visible effect for staff; a BIM platform may be heavily used while retaining version problems; an AI assistant may generate many documents without reducing total working time. That is why a few simple indicators are useful.

An indicator is useful only with a baseline, a consistent measurement method and someone responsible for monitoring it. And twenty minutes released per colleague do not automatically create twenty minutes of revenue: management must decide how that capacity will be used.

An architecture practice does not need more technology, but infrastructure that truly supports design, collaboration and project management. At Sunitech, we support SMEs in Geneva, Lausanne and throughout French-speaking Switzerland in managing, securing and developing their IT.

• Audit and performance: workstations, servers, network, storage and slowdowns affecting production.
• BIM architecture and collaboration: choosing local, cloud or hybrid according to your Revit and Archicad workflows.
• Microsoft 365: identities, sharing, permissions, Copilot readiness and dedicated backup.
• Security and continuity: strong authentication, endpoint protection, tested backups, recovery procedures.
• AI and automation: identifying high-potential processes, a pilot and measurement of results.
• Governance: a roadmap aligned with your needs, budget and priorities.

When a subject involves modelling or BIM coordination, we work with your BIM manager and application specialists.

Are you considering modernising your practice's IT? Before replacing your servers, migrating your models or multiplying AI subscriptions, let us identify the investments that are actually useful. Discuss your project with Sunitech

BIM, AI and cloud are real opportunities, but adopting a technology is not an improvement in itself. A practice can have the latest software, expensive workstations and comprehensive cloud services while still wasting time in its processes. Conversely, a practice that standardises workstations, structures its information, makes backups reliable and improves exchanges can achieve substantial results without replacing everything.

The best IT system is not the one with the most cloud, AI or BIM software. It is the one that lets your architects spend more time designing, collaborating and building.

Before changing everything, start by measuring what really deserves to change.