Automations
Automations proven on work your team already did
Once a workflow is mapped, Subliminal builds an automation for it, proves it on your own past cases, and puts it to work under your rules.
Work with usFour steps from a workflow to a live automation
Build
Assembled from the workflow’s recorded steps, inside a replica of your own apps.
Held-out test
Run on real past cases it never saw while it was built, and checked against what your team did.
Shadow
Run quietly beside your team on live work, until its answers match theirs.
Live
Live only on your team’s green light, inside the rules you set.
01 / 02 Develop
Turns a workflow into a proven automation
Subliminal builds an automation from the way your team already does the work, tests it on real past cases it has never seen, then runs it quietly beside your team until its answers match theirs.
- Built from your team’s recorded work
- Tested on cases it has never seen
- Compared with your team on live work before it acts
How an automation is proven
Your company’s own history is the benchmark. Each part of the test is there so that a pass means the work was really done right.
Real cases from your recordings
Subliminal finds real examples of the workflow in your team’s recorded work. Each one keeps what arrived, what the person looked up, what they did and how it ended, and points back to the recording it came from.
A replica of your apps
The apps the workflow runs in are rebuilt as a replica. Every example outside the held-back set is replayed through it, and the replica is complete when each one plays from start to finish as it did for your team.
The latest cases are held back
The most recent cases are kept out of everything the automation is built from, so it is tested on work it has never seen.
No answers handed over
The automation starts with only what your employee had when the case arrived. It never sees what your team did until it has answered.
Sealed before your team acts
In shadow, on live work, the automation’s answer is sealed before your team acts and compared afterwards. The seal proves the answer came first.
Judged on the outcome
An automation rarely takes the same path as a person, so each case is judged on where it ended up: what changed, what was sent and which rules were followed.
02 / 02 Deploy
Puts the automation to work, under your rules
It goes live only on your team’s green light, runs inside the rules you set, and hands anything unusual to a person. Every run is on the record, so you can see what it did and the hours it gave back.
- Run records kept in the region you choose
- Stops for a person where you say it must
- Every run can be watched and replayed
Live, under your rules
An automation’s page says what it may do. Your team sets it and can change it at any time.
What it may use
Every application the automation may open is listed, with whether it may read there, write there, or both.
When it starts
On a schedule, or when the work arrives: a statement that is ready, a ticket that comes in, a request from one of your own systems.
When it stops for a person
You decide what it must not settle alone, such as a payment over a set amount. The run waits, with what it proposes and the evidence, until a person approves it or tells it what to change.
Where it runs
Each automation runs on a machine of its own, with a time limit on every run. Its tasks, credentials and run records are kept in the region you choose.
Watch any run
Open a run to see its screen and every action it takes, as it happens or afterwards. Someone on your team can pause it, give it an instruction or take over.
Every change on the record
Each change to an automation is listed with who made it and when, and every run keeps its full record.
After go-live, every automation is still measured
What you see
Every automation in one place
Its status, how often it ran, how often it succeeded and how long a run takes.
Who stepped in
How many runs were handed to a person, and what was waiting for an answer.
The hours it gave back
Measured against the same work as it was recorded before go-live.
What happens when work changes
Automations are observed too
Their work is recorded in the same way as your team’s, so a mistake shows at the step where it was made.
A changed workflow is reopened
When an app or a rule the automation depends on changes, it hands the affected runs to your team, and the workflow is mapped and tested again.
New models, more work to automate
When a more capable AI model arrives, workflows that could not be automated before are checked again.
Case study

Automations went live on real tickets within three weeks of the first meeting.
Read the case study- Saved in avoided incremental costs
- $2m/year
- NPS score recovered
- +30 pts
- CX automation rate
- 85%
Questions
Each automation is tested on real past cases it has never seen, then run beside your team until its answers match theirs. It goes live only on your team’s approval, and hands anything unusual to a person.
A real example of the workflow, found in your team’s recorded work: what arrived, what the person looked up, what they did and how it ended. Only complete examples, where the outcome was seen, are used as tests.
No. The automation is built and tested in a replica of your apps. In shadow it reads live work as it arrives and records what it would do, while your team does the work. It changes nothing in your systems until your team gives the green light.
Only the applications its workflow uses. Each one is listed on the automation’s page with whether it may read or write there, and its credentials are stored in your region.
It stops and hands the work to a person. The run waits, with what it proposes and the evidence for it, until someone approves it or tells it what to change.
Subliminal provides workflow context, opportunity analysis, and automation capabilities. The implementation approach depends on the engagement.