.png)
For business decision makers
Agentic work for business leaders.
Uliasti helps decision makers in Swiss and DACH companies turn what they learned about AI agents at vendor events into agents that run inside their own processes. Senior-led, Zurich-based, with a track record in regulated environments. The goal is a first agent live on one real process within a quarter.
How agentic work runs in your company
Trigger
A new invoice lands in the mailbox.
Agent
Reads it, checks it against the order in the ERP, prepares the booking.
Your systems
Books in the ERP, updates the CRM, files the document.
Your people
See only the exceptions and approve what matters.
Definition
What agentic work actually means.
Agentic work is work carried out by AI agents that complete multi-step tasks across your systems on their own, within limits you set, while your people handle exceptions and decisions. It differs from a chat assistant in three ways.
It acts, it does not only answer
A chatbot replies when asked. An agent is given a goal, breaks it into steps, and completes them: reads the invoice, checks it against the order, books it, flags what does not match.
It works across applications
Most AI assistants live inside one tool and reinforce the silo. Agentic work connects mail, documents, ERP, CRM and spreadsheets, which is where the real hours go.
It runs without being prompted
Agents start on a trigger: a new mail, a month-end, a changed record. People review the exceptions and sign off on the decisions that matter. The routine stops landing on their desk.
The offer
Three steps from vision to running agents.
The same arc as the event, keynote, customer cases, hands-on lab, but inside your company and ending in production.
Step 1 · Half a day
Executive working session
Like the hands-on lab at the event, but on your processes. We map the work your teams do every week, score it by frequency and by how much judgment it needs, and pick the one or two processes where an agent pays back fastest.
You leave with a shortlist, a first design and a cost estimate.
Step 2 · Within a quarter
First agent in production
We build the agent on that process, connect it to your systems and data, and put it live under supervision: every action logged, exceptions routed to a person, limits you can read and change.
A running agent, measured against the baseline from step 1.
Step 3 · Ongoing
Scale and keep it running
Once the first agent holds up, we add the next processes through the Agentic Factory, train your teams to work alongside agents, and own the maintenance so nothing stalls after go-live.
A portfolio of agents, one operating model, maintenance owned.
Deliverables
What a decision maker gets.
Concrete outputs, not a vision deck.
- A ranked map of your processes by automation potential, with the first two picked
- A business case per agent: baseline today, hours expected to be reallocated, cost to build and run
- A first agent live on one real process, with supervision and an audit trail
- A governance set: who approves what, what the agent may not do, how changes are released
- An adoption plan for the teams who will work with the agent
- A rollout path for the next agents, with us or with your own team
Your stack
Works with the platform you already run.
The event showed you one vendor's world. Your company runs on several. We build inside the tools your teams already use and in the ERP, CRM and banking systems behind them. We are an official Claude partner and neutral on everything else.
Google Workspace
Gemini Enterprise
Microsoft 365 and Copilot
Claude
SAP
Salesforce
HubSpot
ServiceNow
Oracle
Atlassian
Slack
NotionProof
Where this has had to hold up.
We lead with the conditions the work has survived.
%20(1).svg)
.svg)
.webp)
.svg)
.webp)
Questions decision makers ask
Agentic work, answered.
What is agentic work?
Agentic work is work carried out by AI agents that complete multi-step tasks across applications on their own, within limits set by the business, while people handle exceptions and final decisions. It goes beyond chat assistants, which answer questions inside one app, to agents that read, decide and act across several systems.
How is an AI agent different from a chatbot?
A chatbot answers when asked and stays inside one application. An agent is given a goal, breaks it into steps, uses tools and data across systems, and finishes the task or hands it to a person. For a business the difference is that agents remove work, while chatbots speed up work that still has to be done by someone.
What should a business leader do after attending an AI event?
Pick one process, not a strategy. Choose work that is frequent, mostly rule-based and measurable. Record how long it takes today. Then get a supervised agent running on it within a quarter and decide the next step on real numbers. Uliasti runs this as a half-day executive working session followed by a first production build.
How long does it take to get a first AI agent into production?
At Uliasti, a first supervised agent on one real process usually goes live within a quarter of the working session. The build itself takes weeks. Most of the time goes into connecting data, agreeing on limits and approvals, and testing on real cases before anyone relies on it.
Is agentic AI safe for regulated Swiss companies?
Yes, if it is built for it. Agents in regulated companies need an audit trail for every action, human approval for defined decisions, data that stays where the regulator allows, and a release process with versioning and rollback. Uliasti has built under these conditions for Swiss financial services and Treuhand clients.
Do we need Google Gemini, Microsoft Copilot or Claude to do this?
No single vendor is required. Uliasti builds on the platform your company already runs, whether that is Google Workspace with Gemini Enterprise, Microsoft 365 with Copilot, or Claude, and connects the agents to the systems behind it. The model matters less than the data, the limits and who owns the agent after launch.
What does a first AI agent cost?
The executive working session is a fixed price. The first agent is scoped after it with a fixed budget and date, and the business case from the session shows the expected payback before you commit. For reference ranges, see our article on what secure custom software costs in Switzerland.
What secure custom software costs in Switzerland →