Workflow builders
The user must already know what to automate and then express it step by step in a formal interface.
Announced project / Enterprise AI
Automation that no longer waits to be specified.
PAERA observes how work happens, detects recurring decisions and proposes executable routines. The user decides what runs.
See how it worksPAERA
Pattern-Aware Enterprise Routine AI
Observed behaviour → proposed automationThe problem
Most automation begins with work the user has to do first.
Conversational AI
It answers a request, but usually leaves no durable routine that can run tomorrow without being asked again.
Process discovery
It can identify candidates in logs, but the result still needs specialist engineering before it becomes executable.
PAERA removes the specification step: it discovers the pattern, writes the candidate routine and asks for a decision.
How PAERA works
From ordinary work to a routine that is ready to run.
- 1
Observe
Action sequences across email, calendars, documents and connected business systems.
- 2
Detect
Recurring decisions and repeatable patterns within everyday work.
- 3
Propose
A candidate routine that is ready to review, not a blank workflow to build.
- 4
Decide
The user accepts, edits or dismisses the proposal.
- 5
Execute
Approved routines run with controls and a record of what actually happened.
A routine proposal as the user sees it
Demonstrated in live work
The complete chain already operates end to end.
In an early cohort, people worked on their own data and normal workloads while PAERA observed, proposed and executed routines under human control.
The counts show that the system works in a real environment. Broader external validation is the next step.
11users / two weeks of ordinary work
18routines proposed
8accepted as offered
6accepted after edits
4dismissed
Per-execution journal
Observed state and outcomes remain visible after every run.
The technology
Where discovering a routine becomes executing one.
Program synthesis
Deriving an executable program from observed behaviour instead of a written specification. Existing approaches work best in closed domains with a bounded set of actions.
Robotic process mining
Discovering automatable work from interaction logs. Existing approaches typically expect structured traces and produce a specification that still needs engineering.
PAERA connects the two.
Open behaviour across changing business systems, without marked routine boundaries.
Discovery to execution in one pipeline, producing a live candidate routine.
Evidence-grounded proposals traceable to the work patterns that produced them.
Bounded autonomy
It can propose on its own. It cannot act without boundaries.
Autonomy is useful only when people can inspect, approve and stop it. PAERA treats those controls as part of the system, not as optional settings.
Human approval
Outbound actions remain subject to explicit approval.
Evidence first
Every proposal is tied to the observed behaviour that produced it.
Per-run record
An execution journal records observed state, actions and outcomes.
Hard stop
Scheduled work remains stoppable through a persistent control.
Data boundaries
Organisational data is scoped to the relevant enterprise environment.
Reversible by design
Routines can be suspended, cancelled or withdrawn completely.
PAERA / NEXT VENTURE ACCEL