Visual agent builder
Drag, drop and wire up nodes. No code needed, and Python is there when you want it.
AI agents for enterprise and the public sector
Connected to your systems. Your people make the decisions.
Designed to connect with systems such as these. Each connection is confirmed during mapping
Designed around people
Liba handles the system work around the team: intake, knowledge, review and handoff, while people stay in control of the decisions that matter.
The platform
Most tools help you build an agent and stop there. Liba also handles the review, the approval and the audit that come afterwards. The screens below are illustrative and use sample data.
Drag, drop and wire up nodes. No code needed, and Python is there when you want it.
An orchestrator delegates to specialists and merges the results.
Point it at your documents. Every answer links to its source page.
When the agent is unsure, the decision goes to a named reviewer instead of straight out the door.
Dev, staging and production, versioned. Publishing needs approval.
Commercial APIs, your own cloud, or open models running on your hardware.
How a project runs
Our engineers build the first agents together with your team, so the knowledge stays with you when we leave.
We map your systems and stand up secure connections.
We pick one painful process and build it together, with your people driving.
Roles, gates and retention set before production. Compliance designed in.
We train internal builders so departments ship their own agents.
Working with Liba
The platform runs the workflow. Our implementation team helps you get there, one process at a time.
We map the process with the people who run it: its information sources, systems, responsibilities and the points that need judgement. You receive it in writing.
A short scope document for the first implementation: its objective, inputs and outputs, systems, dates, and the measures it must meet to be accepted.
We build the workflow and connect the systems it needs, with your existing systems kept as the system of record.
We define where a person approves, how exceptions are routed, and what is recorded for every case.
Results are tested with your process owner, users are trained by role, and responsibility for running the workflow is handed over explicitly.
Exact scope, timing, deployment and ongoing support depend on the agreed engagement.
Applications, document checks and exceptions, with employees making the decisions. Includes a municipal case study.
Read about municipal workflows → OrganisationsHow an existing business process becomes an agreed AI implementation, with an anonymous example and what affects scope and timing.
Read about enterprise implementation →Government & local authorities
In most enterprise AI, the language your people actually work in is a translation layer bolted on late, and sovereignty costs extra. We built the platform the other way round.
Use cases
Each one is scoped, measured against your own baseline and accepted before anyone relies on it.
Grounded strictly in your published bylaws. Hands off to a human when unsure.
Files classified and checked against requirements before a planner sees them.
OCR and validation against your credit policy. Exceptions drafted as a memo.
Agreements compared clause by clause, with a deviation report and redlines at the end.
Reads the audit programme, pulls the evidence, records where it was found.
Draft tenders from your clause library, then score bids against the criteria.
Security & governance
Every step is saved, so you can always check what happened and why.
Nothing you process is used to train a model, and that is written into the contract.
AES 256 at rest, TLS 1.3 in transit, and encrypted credentials.
SAML SSO, SCIM, enforced MFA and access by role.
Answers respect the reader permissions you already set.
Every run is logged, versioned and can be replayed later.
Controls designed with ISO 27001, SOC 2, the GDPR and local data protection law in mind. Ask us for current certification status.
Engagement models
We would rather deliver one agent that survives contact with your auditors than fifty that never leave the pilot.
The one with the backlog, the paper, and the person who has been doing it by hand for eleven years. Tell us what the process is, where it gets stuck and which systems it touches, and we will tell you plainly whether an agent can take it on.