AI for an architecture practice: which tasks can an agent handle?

Examples of agents we build for architecture practices

Each agent is custom-built on your tools; a few examples:

• The tender watch agent: reads the contract notices published each day and flags the ones that match your references.
• The bid agent: reads the tender rules, lists the documents to submit and gathers references, insurance certificates and team CVs.
• The references agent: finds your delivered projects by client, building type or budget, and prepares the reference sheets in the format requested.
• The written documents agent: drafts the project description, planning statement and specifications per trade package from the project's notes and comparable projects.
• The drawing revisions agent: reads the title blocks of drawings received, keeps the issue register and flags a contractor working from an outdated revision.
• The bid analysis agent: reads the bids received, compares them per trade package against the tender documents and drafts the analysis report in your template.
• The site minutes agent: works from your notes, photos and voice memos taken on site and writes the minutes in your template.
• The open actions agent: follows unresolved items from one meeting to the next and chases the responsible contractor before the due date.
• The snagging agent: keeps the list of snagging items at handover, with photos, and chases contractors until each one is cleared.
• The dashboard agent: prepares a monthly view of projects, current stages, fees invoiced and bids submitted.

Can it help respond to a public tender for design services?

You upload the contract notice and the tender rules, or the agent picks them up from your tender watch. It lists the documents you need to submit and gathers the ones you already have: comparable references, proof of insurance, team CVs. It drafts a first version of the approach statement from the ones your practice has already submitted.

The partners review and settle the team and fees before submission.

Can it find our project references?

Tenders ask for references of a similar kind, and they often sit in a former project architect's folders. Saména reads your project folders and finds a past project by client, building type or team, citing the source.

You then ask the way you would ask a colleague who had read everything:

• Which school buildings have we delivered in recent years?
• Who ran the library job on site, and which engineers did we work with?
• What construction budget did we state for this refurbishment?

Can it draft written documents, from sketch design to tender?

In French public projects, the architect's basic commission runs from sketch design (ESQ) to assistance at handover (AOR). At every stage, part of the written documents carries over from one project to the next: project description, planning application statement, specifications per trade package.

The agent drafts them from the project's notes, the drawings and comparable past projects, with each section pointing back to its source document, and spots what has changed from one stage to the next.

Can it analyse contractors' bids?

At the procurement stage (ACT), the agent reads the bids received and lays them out per trade package: amounts, alternatives, qualifications, departures from the tender documents. It drafts the bid analysis report in your template, and the architect adds their ranking. Choosing the contractors stays with the client.

Can it clean up site meeting minutes?

During construction (DET), meeting notes are taken in a hurry, photos pile up on the phone and the discussion carries on by email. The agent turns them into minutes in your template: attendees, progress per trade, action items, responsible contractors and due dates. On site, you just send it photos and a voice memo on WhatsApp.

From one meeting to the next, it lists the items still open and chases the contractors concerned; at handover, it follows the snagging items until they are cleared.

Where do our drawings and project files go?

Your files stay in SharePoint, OneDrive or Dropbox. Saména reads them in read-only mode and keeps no copy.

What Saména retains is encrypted with a key unique to your practice. Processing servers are in France, the database is in the European Union. You choose the model: Claude, models from OpenAI, or Mistral if you prefer a French model. It runs in Europe and keeps none of your data. If your requirements or a client's call for it, Saména can be installed on your own infrastructure, and everything can be deleted on request.

Where do we start?

It starts with a thirty-minute call: we look at how your practice works, from the bid to handover, and which tools you use.

Our engineers then build the agents around the way you work, connected to your project folders, your email and your management tools. A first agent is usually running two weeks after the call, in the Saména app, on WhatsApp or from Claude, ChatGPT or Copilot.

We then stay on to evolve your agents and add new ones. There is a one-time setup fee, then a monthly subscription; we share the amounts on the call.

Frequently asked

Does Saména already work with architecture practices?

Saména works today in consulting, environmental engineering and a network of car repair garages, and each agent is built around the way the company using it works. The first call is there to see what an agent would take over for you.

Does the agent produce drawings?

No, design stays with your architects. The agent reads your drawings for what is useful (revisions, title blocks, annotations) and takes on everything written around them: bids, project descriptions, specifications, bid analyses, minutes.

Can the agent connect to our project tools?

Yes, that is the point of custom work: we connect to the tools you already use, whether your project server, email, tender watch platform or practice management software. You don't change tools.

Can we use the agent from Claude or ChatGPT?

Yes. The agents are available in the Saména app, on WhatsApp, and also from Claude, ChatGPT or Copilot.