Six systems. Still running.
Six systems, each one commissioned and paid for by the business that still runs it. Client names are withheld under confidentiality; the figures are not. None of these were sold as a way to cut people, and none were used that way — what gets them approved is that senior people stop assembling files and go back to judging what the files say.
Worth being straight about the dates: all six were built and delivered before Knot Labs was formed, by the person who would run your discovery, for employers and clients who cannot be named. That is why the names are missing and the numbers are not. It also means the engines underneath them are not a prototype — they have been in production for years, which is the whole reason a fixed price in weeks is possible.
Four monthly exports became one dashboard and a two-hour review
The problem. Each month an analyst exported four files, reconciled them in Excel, worked out support status, then rebuilt twelve charts. Three days of work — and the numbers still moved between drafts.
- Software that reads all four files and matches the records by ID, so nothing lines up wrongly
- Dashboards for equipment health, support status and renewal exposure, plus an Excel audit workbook giving the reason each item was flagged
- A deck export in the group's own template, with commentary drafted
Where it repeats. If your team reconciles several exports each month and then rebuilds the same charts, this is the same build with your columns and your thresholds. It is our fastest project type.
overdue items nobody had counted
Recreated view. The first three steps stopped existing.
The dashboard stopped waiting to be read, and started emailing the people responsible
The problem. The dashboard was accurate and unread. A coordinator chased the red rows by hand, and two renewals lapsed in a quarter that the dashboard had flagged eleven weeks earlier.
- We read their own asset system and their machine-monitoring system directly, as it happens, replacing four monthly exports
- Machine readings, with each machine judged against its own history, and renewal exposure forecast by quarter
- Fourteen alert rules routed to the engineer named on the asset record, with reminders, escalation at day 21, and a two-week silent run before anything was sent
Where it repeats. If the value depends on someone noticing something, the alerting layer is where the return actually is — two to three weeks of work on top of a dashboard you already have. The honest lesson here is that we should have proposed it in phase one.
across two quarters
14 rules live · 31 open actions
Weekly snapshots and five early-warning measures across a portfolio
The problem. Project health was reported monthly, by exception, on slides. Problems were usually true for six weeks before they reached a report, and nobody could answer "which projects are drifting, and since when?"
- One place that pulls project, finance and timesheet data together, saved automatically every week so the history builds up without anyone remembering to keep a copy
- Five early warnings: spending less than planned, cost per unit of work, running late, work quietly growing, and how happy the client sounds
- A chat box for asking questions about the whole portfolio in plain English
Where it repeats. The same pattern fits a sales pipeline, a service queue or a franchise network — anywhere "is this getting better or worse?" is asked every week.
One view of project finances, built from four systems nobody wanted to open
The problem. Cost against plan lived in a finance export, a timesheet, a CRM and a set of calendars. Reconciliation was monthly, and the version being presented was usually two weeks stale.
- Separate manager and project-lead views of the same underlying numbers
- Timesheet reconciliation against planned effort, with unbooked time highlighted
- Calendar and CRM data used to explain variances rather than just show them
Variance against plan, by programme. ₹6.4Cr booked · 88% of plan used · 14 variances surfaced
Where it repeats. If answering a routine question requires opening three or more systems, the consolidation itself is the product — before any AI is involved.
Two more, in short
A 34-point standard applied to every deck, in 90 seconds
Client decks were prepared by many people to an inconsistent standard, and a senior reviewer read them the night before. The standard is now encoded as 34 weighted checks, with the missing slides drafted in the firm's own template.
Defect reviews stopped starting with spreadsheet prep
A fortnightly review needed three hours of querying, copying and formatting. The workbook now builds itself in the meeting's exact format, straight from the ticket system. The first run surfaced 11 stale items and 6 with no owner.
And one we built for ourselves
Nobody commissioned this one. We built it because our own reporting work kept reintroducing the manual slide-making the dashboards had just removed, and it now ships inside every reporting build we deliver. That makes it worth showing — but it is not a delivered client outcome, so it is not counted in the six above or in the figures at the top of this page.
The deck generator that makes the rest repeatable
Even with dashboards in place, charts were being screenshotted into slides by hand. DeckLoom produces a full deck from your numbers in a defined template, with snapshot history so you can reopen what March actually said.
It is one of the three engines the systems above are assembled from. DeckLoom powers Automated Reporting. SignalWatch powers Licence & Asset Tracking and Alerts & Escalation. DocCite powers Document Processing and the AI Chatbot.
- A full deck from your numbers, in a defined template
- Export to PowerPoint, PDF and HTML
- Snapshot history, so a past version reopens as presented
Included in Automated Reporting — it is why a dashboard project ends with slides on your desk.
What we would look at in your business
If one of these sounds like a Tuesday, the matching row is the shortest route to understanding what we would actually do.
| If this sounds familiar | The closest thing we have built |
|---|---|
| Several exports reconciled by hand each month | The Three-Day Month-End — our fastest build |
| An accurate dashboard that nobody acts on | The Dashboard That Chases — we read the systems directly, then alert the named owner |
| Problems are noticed weeks after they start | The Early-Warning Board — a weekly saved picture, and five early warnings |
| One answer requires three systems, or a meeting needs hours of prep | One View of the Money and Review Without the Prep — everything in one place, pulled straight from the system that holds it |
| Document quality depends on who prepared it | The Ninety-Second Standard — your standard, applied to every document |
| Dashboards exist, but the slides are still made by hand | DeckLoom — Automated Reporting; it ships inside every reporting build |
| It all lives in one person's spreadsheet | None — that is a build-the-register job, and we would say so |
Which row was a Tuesday?
Four hours with the person who lives it, and you get the workflow mapped, the hours counted, and a written verdict you own.