Swiss applied AI lab · Working worldwide

Your goals. Our AI. Wound into one.

unus takes on the hard problems that hold your numbers back, from a single process to a worldwide rollout. You bring the financial and operational targets. We bring the agents, the engineering and the people, and we stay until the numbers move.

We exist for the problems that usually get scoped out.

Hard problems, taken whole

The process nobody has managed to automate, the data nobody trusts, the target that slipped twice. That is where we start.

Two strands, one structure

Your people own the goals and know the business. Ours design, build and run the AI. Each engagement pairs one of your targets with the capability that moves it.

Measured in your numbers

Margin, cycle time, cost to serve, revenue. We report against the figures on your own dashboard, not against our deliverables.

Six things we build.

A team for you, and within you.

Forward deployed: inside your teams, on your systems, until the numbers move. One process is enough to start.

Three steps. No theatre.

1/

Start a conversation

Thirty minutes with an engineer, not a salesperson. Bring one process that costs you people's time.

2/

Two-week discovery sprint

On site with you, or remote. We map the data, the permissions and the risk class, and build a working slice.

3/

First agent in production in eight weeks

Evaluated, governed, observed. Your team owns it from day one, with our engineers in Riga, Tallinn, Kyiv and Bengaluru keeping it moving across time zones.

Golden-set evaluations before every release, human approval where the AI Act risk class requires it, every step traced. For your tech team →

Our global presenceBuilt by one team in Zürich, Riga, Tallinn, Kyiv and Bengaluru, with hubs in München and Wien.

Builders in the room, not account managers.

  • Product managersTurn a business target into a scope that can ship, and keep it honest.
  • DesignersMake agents and tools that your employees and customers actually want to use.
  • AI engineersBuild, evaluate and operate agents on your systems and your cloud.
  • Business consultantsRework the process around the technology, so the gain shows up in the P&L.

What this year's AI releases change, sector by sector.

As of November 2026: what shipped, what it changes for the business, and the source. Taken from vendors, the companies involved and regulators.

These are market briefings on public announcements. The companies named are not unus clients.

BankingSep 2026

Agentic KYC is in production.

Two colleagues reviewing financial documents at a desk

Deutsche Bank Private Bank put agents into source-of-wealth checks: they research and document, and a person reviews the assessment.

What it changes

Governance of agents and external models under DORA becomes the gate, ahead of high-risk credit rules from 2 Dec 2027.

InsuranceAug 2026

Claims agents move into the core platform.

An inspector filling in a form on a clipboard inside a car

Guidewire's Qusar release ships an agentic framework with prebuilt agents for claim summaries, policy changes and voice first notice of loss.

What it changes

Claims automation becomes configuration, not a project. Life and health pricing stays high-risk under the AI Act.

TelecomOct 2026

AI moves into the radio network.

Two technicians working high on a mobile network mast

Deutsche Telekom, Nokia and NVIDIA measured typical uplink gains of 10–15% from AI-based channel estimation in AI-RAN lab tests.

What it changes

Network operations go multi-agent across RAN, transport and core, and spectral efficiency becomes a software question.

ManufacturingOct 2026

Agents start acting on the line.

Engineers talking beside assembly rigs on a factory floor

Siemens showed agents at its Erlangen plant that recommend production changes, tested in a digital twin while the line keeps running.

What it changes

Automation engineering becomes an agent workload, and simulation becomes the safety net for anything touching production.

All six sector briefs, including energy and cross-industry →

Questions your risk team will ask.

Where does it run?

Wherever your rules say it must: in your own cloud, on a single-tenant setup we operate for you, or on our platform in Swiss and EU regions.

How do you stop an agent from doing something it shouldn't?

Agents can only do what they are allowed to do. Each one works through scoped permissions, its inputs are checked, and every release is security tested before it goes live. Where the risk class requires it, a person approves the key steps.

What happens to our data?

Personal data is removed before anything leaves your systems. We use zero data retention with the model providers that offer it, and EU-hosted models where they do not.

Will it stand up to an audit?

GDPR and the Swiss nDSG are built in from the start, and the EU AI Act duties are mapped for each use case. Our controls are aligned with ISO 27001, and certification is in progress.

Regions, encryption, the model gateway and certification status, in detail. For your tech team →

Bring one process. We'll bring the loop.