AI agents for production and quality in SMEs

Today

The site manager jots down meeting decisions on a phone or in a notebook, then writes up the minutes in the office in the evening. Reservations and deadlines get lost from one meeting to the next.

With AI

After the meeting, the site manager sends a voice note and photos in Teams. The agent transcribes them, carries over open items from the previous minutes in SharePoint, drafts the minutes by trade, creates tasks with owners and deadlines in Planner and prepares the e-mail to the parties in the drafts folder. The site manager corrects and sends it personally; the agent logs what it carried over and created.

Today

Non-conformities arrive by e-mail, by phone or on a shop-floor sheet, described unevenly. Root-cause analysis and follow-up of corrective actions fall behind.

With AI

As soon as a non-conformity is recorded, the agent finds the item, batch, supplier and affected deliveries in the ERP, looks for similar cases already handled, completes the record and opens the 8D template with the questions to ask. If a supplier is involved, it drafts the complaint and proposes containment actions in Planner. The quality manager approves the analysis and alone decides whether the batch is blocked or released; every step the agent takes is logged in the record.

Today

Each machine variant needs its own manual, data sheets and translations into German, French, Italian and English. Updates are made by hand and versions end up drifting apart.

With AI

We build an agent triggered by each approved engineering change in the ERP or CAD system: it identifies the manuals and data sheets affected for each variant, rewrites the relevant sections from your approved texts, updates the translations using your glossary and places a new revision for review in SharePoint, with the list of changes. The technical writer reviews each version and the responsible engineer releases the documentation; the agent’s code and documentation are handed over to you.

Today

Faced with a rare fault code, the mechanic searches technical bulletins and the vehicle’s history, or calls a more experienced colleague. What was found last time is not written down anywhere.

With AI

When a repair order is opened with a fault code, the agent checks the vehicle’s history and similar repairs in your garage management software, plus the technical bulletins your licences allow, then proposes a diagnostic plan with its sources. It checks stock for the likely parts and prepares the parts request to the supplier without sending it. The mechanic tests and decides on the repair; the workshop manager approves the order and anything affecting vehicle safety.

Today

The self-monitoring concept, cleaning schedules and checklists sit in a binder, rarely up to date and hard to explain to new staff.

With AI

In a workshop, the team learns to use an off-the-shelf assistant to update cleaning schedules, write short checklists for each station and translate them for staff who speak other languages. The manager or hygiene lead checks every document against cantonal requirements and your procedures before it is posted.

Today

Service technicians write their reports in the evening or at the end of the week, from memory. Details needed for invoicing and the next visit are missing.

With AI

The technician dictates in Teams what was found and done. A ready-made agent connected to your customer files puts the text into your report format (findings, work, parts, time, next steps), flags empty fields, files the report in the customer’s folder and creates the follow-up task for the next visit. The technician reviews before sending, and the service manager checks the billable items before any invoice.

Today

Workstation safety instructions exist, but they are long, seldom read and not always translated. Formalising the risk assessment for a new task takes time.

With AI

The assistant prepares a first draft of the risk assessment from the task description and SUVA checklists, then a short instruction for each workstation in several languages. The safety officer checks every measure on the shop floor and signs; nothing is posted or circulated without that approval.

Today

Breakdown history, operating procedures and manufacturers’ manuals are scattered across the CMMS, PDFs and the memory of long-serving staff. A night-time breakdown on a line or a forklift takes longer to fix than it should.

With AI

As soon as a breakdown is logged in the CMMS, the agent we build finds past work on the same machine and the relevant passages in the manuals, shows them to the technician with sources, checks spare-part stock in the ERP and prepares the work order, plus a purchase request if a part is missing. The maintenance manager approves the order and the purchase before they enter the CMMS and ERP; the agent logs every lookup and every proposal.