What people say. What people do. What the systems show.
EMBR measures how work actually moves by comparing three kinds of evidence: what people say, what they do and what the systems show. Every finding is measured in hours per week, taken as a baseline before anything is built, and measured the same way again 90 days after launch.
Last reviewed 30 September 2026
What do we look at?
What people say. Interviews, one hour each, where the work happens.
What people do. Observation: the cheat sheets, the second screens, the WhatsApp groups.
What the systems show. Systems mapped from real screens, not the architecture diagram, and data used as pulled, unadjusted.
FIG. 01 · The Transect, four processesThe Transect: one sample per process, aligned at the surface. Order 9 hours a week, close 7, buy 4, pay 2. Illustrative.
Which evidence counts most?
Observed. We watched it happen and timed it. This outranks everything.
System data. Used as pulled, unadjusted.
Claimed. What people tell us, scored only when a second source agrees.
Where evidence classes disagree, the lower figure wins. Where there's a range, we use the low end. Where we can't measure, we count zero.
How is a finding measured?
In hours per week. Every layer of a process gets a year and a state: live (the workarounds people use to keep things moving), dormant (systems partly used) or dead (still in place, not used at all). Its depth is the hours people spend each week working around it or through it: workaround debt, measured layer by layer.
Every layer was a sensible decision at the time. It only hardened because nobody went back. We date layers; we don't judge them.
Each diagnostic ends with three to five ranked findings. Each has a depth in hours per week, a class (process fix, plain automation, or genuinely AI) and a build path.
Diagnose. We take the sample: what people say, what they do, what the systems show. Every layer is dated and measured. This is the baseline. The Transect.
Build. We clear the layers in the way and build one that reaches the core. First Build and The Forge.
Improve. We keep the build in use, stop new layers settling, and take the same measurement at 90 days. Tending.
How do you know it worked?
Same measurement, 90 days later. Every build carries its own success numbers, agreed before work starts. Ninety days after launch we take the sample again, the same way, and compare. The test is simple: did the spreadsheet come back?
Can the answer be "don't build"?
Yes. Every finding is classed honestly, and some need no build at all. We're an AI company. We'll still tell you when the answer isn't AI.
Tell us what you're working around, and we'll come back to you within two working days.
01You write two linesWhat you're working around. We read it the same day.
02A 30-minute callWithin two working days. We ask what the workaround costs you, and say which stage fits.
03A written scopeFixed fee, what you get, when. Nothing starts until you've signed it.
Tell us what you're working around
Engagements start at R50,000. Every fee is fixed and agreed in writing before work starts. No hourly billing.
01Two lines from you, we read it the same day.
02A 30-minute call, within two working days.
03A written scope, fixed fee, nothing starts until you sign.
Your route from the pricing page
Thank you. We'll come back to you within two working days.
