The same data is entered more than once
A figure arrives in an email or a PDF, gets typed into a spreadsheet, then typed again into the system that actually holds it. Every copy is a chance to introduce an error nobody can trace later.
Some processes are slow because the work is hard. Others are slow because information has to be moved, re-typed and checked by hand between systems that do not talk to each other. This is work on the second kind.
In most of the processes we are asked about, the judgement takes a few minutes and the handling takes a day.
A figure arrives in an email or a PDF, gets typed into a spreadsheet, then typed again into the system that actually holds it. Every copy is a chance to introduce an error nobody can trace later.
When a process has no record of what was checked and against what, the safe response is to check it again. The work is repeated because nothing proves it was done.
It works until they are on leave. The knowledge is real, but it is not written down anywhere the business can use.
It has grown formulas, tabs and conventions. Everyone depends on it, and nobody wants to be the one who changes it.
These overlap in practice. A single engagement usually touches two or three of them.
A defined sequence runs without a person moving the data between steps. Where a model is used, it is used for a specific judgement inside the sequence, and what it decided is recorded so it can be reviewed.
Two systems that both hold part of the truth get a reliable connection, with a stated direction of travel, a defined schedule and a record of what failed.
Information is read out of PDFs, forms and exports into a structured shape, with each value linked back to the page it came from so a person can verify it quickly.
A small application built for one team and one job, instead of a general platform that needs configuring. Usually a screen that replaces a spreadsheet and the routine around it.
Supplier files are compared against your own catalogue, line by line, so the changes you need to approve are separated from the rows that need a human decision.
Both were built around a process that someone was doing by hand. Neither is presented as a completed client engagement, because neither has been published as one.
Distributors receive a new supplier price list and have to work out, row by row, which of their own catalogue lines changed. This compares the two lists against the catalogue and produces a change file a buyer can review.
In development Data used: Synthetic distributor catalogue and supplier price lists. No client data.
An account manager re-reads a submission package to find what is missing or contradictory before it goes to a carrier. This reads the package and lists findings with a link back to the page each one came from.
Live product Data used: Shipped as PacketSlip. The demonstration here ran on synthetic ACORD 125 and 126 documents.
Every step names what you provide and what comes back. Where a step depends on something only you can supply, that is said before the step starts rather than discovered inside it.
Pick the task your team dreads most this month. Describe it in a few sentences, with rough volumes, and we will tell you whether it is a fit and what scoping would involve.
We would like to measure which pages are useful. Analytics only loads if you accept, and we do not use it to identify you.