Replacing a six-month build with three weeks of configuration.
The internal tool nobody had staffed for two years shipped in a sprint, with the same roles and audit trail as everything else.
3 weeks to production
- Customer
- Gridworks
- Sector
- Electricity distribution network operation
- Scale
- 2,340 staff, 1.9 million connections
- Duration
- Three weeks from first workshop to production
4 phases, in the order they happened.
- Week 1
Data model and approval flow agreed with two switching engineers and the control room supervisor.
- Week 2
Forge build: six forms, four tables, three roles from Gate, revision rules written in Script.
- Week 3
Asset register connected through Link, two correction rounds, production release on the Friday.
- Week 7
A second tool (jointing crew allocation) built in four days by Gridworks' own team, without us.
What changed, measured.
Every figure came from Gridworks. Where one is a median or a sample it says so.
3 weeks
First workshop to production
Against a six-month estimate held on the roadmap since 2022.
340
Planned outages a year
Now scheduled, revised and approved in one place.
4 days
Second tool, built without us
Jointing crew allocation, delivered by Gridworks' own team in week seven.
1
Audit trail
The platform's own, accepted at incident review without a separate assurance case.
The tool nobody would fund.
Every planned outage at Gridworks needs a switching schedule: the order in which circuits are isolated and restored, the plant involved, and sign-off from three separate parties. All of it was tracked in a workbook on a shared drive, under a naming convention that two people fully understood.
A replacement had been on the roadmap since 2022 as a six-month build with a number attached. It never reached the top of the list, because it was always smaller than whatever sat above it and never small enough to slip in beside something else.
Meanwhile: 340 planned outages a year, an average of fourteen revisions each, and a file that only one person could edit at a time.
Why the six-month estimate was honest.
It is tempting to say the estimate was inflated. It was not. The build would genuinely have taken six months, because most of the six months was not the tool.
It was authentication against the corporate directory, a permission model matching the three approving parties, an audit trail an engineer could defend at an incident review, and an integration to the asset register so that a switching schedule references real plant rather than a typed description. Those four are the same on every internal tool this company will ever build, and rebuilding them per tool is where the six months goes.
Forge is worth having precisely because those four stop being the project.
Three weeks, and the same audit trail as everything else.
Week one was the data model and the approval flow, drawn on a whiteboard with the two switching engineers who write most of the schedules and the control room supervisor who signs them.
Week two was the build: six forms, four tables, three roles, and the revision rules in Script. Roles came from Gate, so each of the three approving parties was an existing group rather than a new list to maintain. Week three connected the asset register through Link and ran two rounds of correction with the same engineers.
It is not a beautiful piece of software. It replaced a workbook, and it is measurably better than the workbook.
The six-month number was never wrong. It was the price of building the boring four-fifths again. We paid that once, on the platform, and the tool itself took three weeks.
Three others, from three different starting points.
Different sector, different product, the same argument about where the data should live.
34% shorter sales cycle
CEDX Pulse
Rebuilding a revenue engine on one system of record
Northbeam · Freight forwarding and contract logistics
3,140 staff across eleven countries · Eleven weeks, kickoff to final regional cutover
0 standing privileges
CEDX Gate
Standing up zero-trust access across nine countries
Aperture Diagnostics · Diagnostic imaging and pathology services
6,800 staff across 41 sites in nine countries · Twelve days to the first production token, fourteen weeks to full coverage
2-day close, down from 14
CEDX Ledger
Closing the books on the second working day
Tidepool Marine · Aquaculture and seafood processing
1,480 staff, six sites, three reporting currencies · Sixteen weeks, with the first two-day close at the fourth month end
See what Builder does before you read another story.
Low-code apps on CEDX data. There is a full trial on builder.cedxsystems.com, and an engineer who has run a migration like this one will walk it with you.