The numbersbecome our mission.
We built a tool that counts the work your company actually does. Every workflow. Every screen. Every hour. It hands you a ranked list with the numbers attached. Then you work the numbers — nothing else.
Why this is hard
The hard part isn't the technology.
Four things get in the way of using AI at all. None of them are anyone's fault.
- A
Everything looks like a candidate.
Every workflow could be automated. Every vendor says theirs is the one to buy. With no way to compare them, the safest move is no move.
- B
The people who know the work don't know the tools.
An inside sales rep can describe their job in perfect detail. They have never evaluated an AI product. They should not have to.
- C
The people who choose the tools don't do the work.
The call gets made two floors up, from a slide. The job it was meant to fix carries on unchanged.
- D
A pilot proves the tool, not the payback.
Six weeks later you have a demo that works. You still do not have the hours back.
So the question was never which tool. It was which work.
Why we built it
We got tired of guessing.
Every company knows it is losing time. Almost none can say where, in hours. So the loudest opinion wins the argument, and the most expensive problem stays invisible.
So our team built one thing. A way to put a number on every piece of work in the building.
Extract everything the company knows.
Query it until the work has numbers.
Work the numbers, in order.
It's really that simple.
— The Cirql team
01 — Extract
We take all of it.
Fifteen minutes, on their own phone, for every person in the company. Will-call counter to COO. No meetings. No consultants. No IT project.
Nobody is asked about AI. Not which tool to buy. Not whether a task could be automated. Not what a model can do. That is our job, not theirs.
They are asked one thing. Describe your own work, in your own words. What you do, in what order, with what, and what gets in the way.
This is not a survey. A survey collects opinions. We collect the work itself — and then we do the translating.
What comes back
Workflows
Every step, in the order it is really done.
Screenshots
The screens the work actually happens in.
Definitions
What your people call things, and what they mean by it.
Processes
The documented ones and the invented ones.
Systems
ERP, CRM, webstore, vendor portals, spreadsheets, email.
Volumes
How often, how long, how many people.
- No AI knowledge needed, and none asked for
- Their language, not ours — no jargon the floor doesn't use
- Any phone. No app, no login
- Voice or typing, whichever is easier on the floor
- Adapts to their role, branch, and region
- Anonymised by default
Median completion rate · 87%
Counter first
We start at the counter, not in the boardroom.
Most surveys ask everyone the same question. We don't. Each level answers, and those answers set the next level's questions.
- 01
The counter and the warehouse.
Inside sales, counter, pickers, drivers, shop techs
“Walk me through a real Tuesday.”
Ground truth.
- 02
Branch managers and buyers.
Branch managers, purchasing, dispatch
“Your branch said this fourteen times. Is that news?”
What is known and accepted.
- 03
Regional and functional leaders.
Sales, purchasing, and operations directors
“Who owns this process?”
Where it is stuck, and why.
- 04
The executive.
CEO, COO, President
“Here is what the company actually does.”
The plan, against reality.
02 — The map
It all lands on one map.
Every answer, artifact, and system becomes a node. People, tasks, tools, and the constraints between them.
Three people name the same wait — inside sales, the counter, the service shop. Their branch manager confirms it. The ERP and a vendor email thread are tangled in it. The map holds that shape, which is what makes it countable.
03 — The queries
Then we run the numbers.
A fixed set of queries runs across the map. Every one of them asks a counting question. How many people. How often. How long. How sure are we.
= Hours per year. For every finding. Every time.
Nothing is ranked by how loudly it was raised. It is ranked by the number.
Sample distributor · 11 branches · 214 respondents
Counted together: 7,570 hours a year, on six findings.
04 — The verdict
Every finding gets one of three.
You do not get a wall of charts. You get three piles, and every number sits in exactly one of them.
How a finding moves
14
Investigate
The number is big. The cause is not clear enough to bet on yet.
Next: one hour with the two people who named it.
6
Solidify
The cause is clear. Scope, owner, and cost still need pinning down.
Next: write the spec. Confirm the hours.
3
Act
Proven, quantified, owned. Nothing left to decide.
Next: build it. Measure against the baseline.
23 findings. Three you can start on Monday.
05 — The report
One page, already sorted. Every row carries a verdict, the evidence behind it, and the number. You can make the call in the meeting where you open it.
- Act3,120
Vendor quotes re-typed from email into the ERP
14 of 16 inside sales reps · 52 min a day
- Act1,850
Every non-stock quote waits on one buyer
All four branches · 20–45 min per quote
- Solidify1,240
Freight quoted eight different ways
6 of 8 branch managers want one standard
- Solidify610
Competitor cross-reference list has no owner
No owner named by anyone we asked
- Investigate470
Open orders checked twice, ERP and vendor portal
Three branches · cause unconfirmed
- Investigate280
New inside sales re-ask the same four questions
9 of 11 new hires · no written answer anywhere
What's in it
- Every finding, ranked by hours per year
- The verdict on each one — investigate, solidify, or act
- The evidence — who said it, how often, how long
- Where the hours sit, by role, branch, and region
- Your systems, and every place two of them overlap
- Full transcripts, anonymised
Exports to PDF · CSV
A run, start to finish
Ten minutes to launch.
No IT project. No seat licences. No integration with your ERP. One day end-to-end is common. Two weeks is the outside case.
- 01
Roster.
Upload your people. Tag branch, region, and role.
CSV · 30 seconds
- 02
Invite.
Everyone gets their own link by text or email. No logins to hand out.
Any phone · SMS or email
- 03
Interview.
Fifteen minutes, whenever they can. It adapts as they answer.
10–15 minutes · voice or type
- 04
Report.
Ranked by hours. Verdict on every row. Evidence attached.
Exports to PDF · CSV
Stop guessing. Start counting.
Ten minutes to launch. Your people answer on their phones, between runs. You get the work quantified, ranked, and sorted into three piles.
$2,500 for four runs over twelve months, unlimited respondents, one per quarter. Every run after the first is scored against the last, so you can watch the numbers move. The whole campaign is credited toward any Cirql build within 60 days.
Start a free 25-person pilot