ProcessRoot AI
From a text to a live workflow, with two human touches
Our own firm runs on the same operating system we install. This is the path a client takes through it, in order, with what the system records at each step.
Our phone line and inbox are being connected. This is the path a client takes the day they are.
The walk
Eight steps
You text us.
Your text becomes a lead in our operating system: a record for your business, a contact, a deal, and a work order for our Sales agent. In our recorded rehearsal that took three seconds.Our Sales agent reads it and drafts a reply.
It checks the lead against the fit rules the owner set and drafts an answer. The answer waits in the review window, thirty minutes by default, where a person at the firm can stop it. Then it goes out to you. The discovery call is booked the same way.A discovery call with one of our industry and AI experts. The first human touch.
A person runs the call and marks it delivered. The system refuses to let an agent do that step. The call is where the scope and the number come from.You accept the agreement online.
After the call the install is priced inside the band the owner set, and a statement of work is drafted from those lines. You get it by email with a one-time link and accept it on a page of ours. Your name, email, network address and the time go on the record, and from then on the text and terms of that agreement cannot be changed.The owner countersigns. The second human touch.
Only the owner, signed in, can countersign. The database refuses a signature from an agent or a service. The countersign opens your engagement, raises the install invoice, and starts the install clock.The install.
Connect, configure, evaluate, stage, acknowledge, go live. You connect your own accounts on a page made for you and answer the questions only you can answer. Before anything goes live the system checks four things: the workflow's test cases passed, your connections are in place, the settings are valid, and you said yes in writing. An agent cannot override those checks.Live.
The clock stops. The days from countersign to live are recorded against the install's target. Every run is recorded, and every run that can be measured is. Anything the workflow is not sure of goes to a person on your staff with the history attached, and that counts as the workflow working, not failing.Results on your portal.
You sign in with the address we invited. Each workflow appears in plain words with its number beside where you started, and a monthly report follows. A weekly note is written from those same numbers, and it waits in the review window like everything else.
The numbers
What the system measures, and what it measured last time
- First text to live workflow
- 808 seconds
- In our recorded rehearsal of the whole path on 2026-09-10, with the review window shortened to one minute for the walk. A rehearsal on our own stack, not a client. The model and the vendors were simulated; the queue, the guards and every record were real.
- Human touches on that path
- 2
- The discovery call and the countersign. Both are refused to software by the system itself.
- Review window
- 30 minutes
- The default wait before anything an agent sends goes out. No setting can make it zero.
- Clean runs before the wait goes away
- 20
- Per agent, per kind of action. One stop by a person resets the count.
- Install target
- 5 to 10 days
- Set per install in its manifest and measured from countersign to live on every client.
- Typed-in results
- None
- The numbers your portal reads are written only by the workflow's own runs; no screen and no person can type one in. A day with no measurement reads "not measured yet", never zero.
Your portal
What you see
Every result on your portal is measured by the workflow itself. The numbers your portal reads are written only by the workflow's own runs; no screen and no person can type one in. The monthly report is drafted by software from those numbers and is frozen the moment it is published. The one figure a person records is your starting point, and it is shown as exactly that.
Each workflow appears in the words of its manifest, with its number beside the number you started from. Where no starting number was recorded, it shows the number alone and says there is nothing to compare it with. It never shows an improvement over a starting point nobody measured.
Each month you get a report: hours back and money moved, computed only where there is a starting number, and a sentence saying why where there is not. A month with nothing to report is skipped and says so. It is never a page of zeros.
The report is frozen when it is published, so the figure you read is the figure that stays on the record.
What we never do
Four things you can hold us to
- We never guarantee an outcome. The install target is a target. A number nobody measured is not reported, and a starting point nobody recorded is never invented.
- We never put prices on this site. The number comes from the discovery call, inside a band the owner set, and it is on the agreement you accept.
- We never move your money. The collections install reads your balances and sends reminders. It never charges anyone, never threatens, and never writes to your ledger. Our own agents are barred from refunds and payouts by code.
- No agent can take a workflow live before you have said yes in writing. The one person who can override that gate is the owner, in writing, and the override is recorded in the audit trail where you can ask to see it. What the workflow then sends is what its manifest says, and anything it cannot answer goes to a person on your staff.