Enterprise · Wolters Kluwer
~70% of a Full-Time Role, Handed to One Dashboard
At Wolters Kluwer, a 25-page manual procedure across nine systems ate roughly 70% of a full-time role. Mission Control runs it instead: the operator clicks Process, and the book contract walks itself through all six phases — with a human sign-off on anything irreversible. In daily use.
~70%
Of a full-time role
9 → 1
Systems → one dashboard
6
Phases automated
0
Irreversible actions without a human
What this means for you
If part of your operation lives in a step-by-step procedure someone follows by hand, it can live in a dashboard instead. This is what that looks like done carefully: the machine takes the repetition, a person keeps every judgment call.
The Problem
At Wolters Kluwer's legal publishing division, every new book runs through a 25-page standard operating procedure: create the contract, draft it from a template, file it for legal review, route it for e-signature, set the pricing, and hand it to publishing. Six distinct phases, touching nine separate systems.
All of it was done by hand, for every single title. It was slow, it was easy to get wrong — the same book shows up under a different name, ID, or format in each system — and it was roughly 70% of a full-time role.
The Solution
One web dashboard, backed by an automation engine that drives the real publishing systems — not a sandbox or an API that doesn't exist. The operator sees a "Needs Action" queue, clicks Process, and the engine walks the contract through all six phases.
Crucially, it doesn't try to be clever where it shouldn't. The handful of steps a human must own — irreversible e-signature sends, ambiguous name matches — are explicit pause points. The machine handles the repetitive 90% and asks before doing anything it can't take back.
New Book Intake
Contract Draft
Legal Filing
E-Signature
Price Setup
Publish-Ready
The Result
- ✓ A 6-phase, 9-system contract lifecycle run from a single dashboard
- ✓ Near-zero manual touches per contract on verified runs, outside the human decision points
- ✓ Delivered solo, alongside the day role — no project team, no procurement cycle, no six-month rollout
- ✓ A self-healing / one-click-recovery model so a non-technical operator can run it
- ✓ In daily use — it's how I run the contract and product-setup phases now, and the platform the wider automation work builds on
The Hard Parts — and How They Were Handled
1. No shared idea of "which book"
Each system identifies a book differently — one by title and authors, another by an internal ID and ISBN — and the titles drift between them. The automation kept matching the wrong record. The fix was a single Book identity model and a resolver that scores matches author-first, so it survives truncated grids and naming inconsistencies without a human re-checking every row.
2. Driving enterprise systems with no API
None of these tools offer a clean automation interface, so the engine operates them the way a person does — through the browser, via Playwright over CDP. That meant handling React, Material Design, and Adobe Spectrum interfaces whose element IDs regenerate on every load; the resolver leans on stable labels and text, not brittle IDs.
3. A locked-down corporate machine
The work laptop blocks package installs and has no developer tooling. Dependencies are vendored through version control so the whole platform runs inside the corporate environment without ever touching a blocked network path.
4. Safety over speed
Anything irreversible — sending a contract for signature, setting up a record that can't easily be undone — is a deliberate human decision point, never auto-fired. More guard rails, not fewer. The goal was a tool a non-technical colleague could run without fear, not a black box.
Why You Can Trust It
0/9
end-to-end phase test cases passing (incl. negative cases)
Behind those end-to-end runs sits a suite of 1,400+ automated tests (zero failures), plus nine local mock pages that reproduce the real publishing systems so any phase can be exercised without touching production.
And it's held up in production, not just in tests: when a data-integrity fix shipped, several book records that had been silently stuck in the wrong pipeline stage were corrected automatically the same day — the first live confirmation that the self-correcting sync behaves exactly as designed.
Tech Stack
| Layer | Technology |
|---|---|
| Automation Engine | Playwright driving the real systems via CDP (Edge/Chrome), Node.js |
| Backend | Node / Express — 16 REST endpoints, SSE log streaming |
| Frontend | Web dashboard — 5 tabs, KPI strip, ⌘K search, batch mode, per-book detail + undo |
| Documents | jszip (Word template fill, no COM), xlsx (Vista + pricing readers) |
| Identity Resolution | Custom Book model + fuzzy Resolver (SequenceMatcher + alias tables) |
| Fallback | Power Automate Desktop (.robin) flows retained as backup |
Let's Talk
I take on 3 new clients per month.
The businesses that move first win. Let's find where you're leaving 20+ hours a week on the table.
Book Your Free Audit● Capped at 3 new clients per month — the audit call is how you claim a slot