What the workflow score means.

Take one order. Here's what it might look like in your systems:

That's one case. It has 6 steps across 5 systems, over 40 days, and a clear ending: the order was paid.

We find cases by matching the same order, invoice or ticket number across your systems, even when each tool writes it its own way. "#1234", "Order 1234" and "SO-1234" all count as the same order.

Each number is a range, not a single score. We show the typical case and the high end, where only 1 in 10 cases goes further. We also show how many cases we measured.

• Systems per case. How many tools one piece of work passes through. More tools means a harder, more real task.
• Steps per case. How many things happen from start to end.
• Time span. Days from the first step to the last. Long spans show work that has to be tracked over time.
• Handoffs. How many times the work passes between people or teams, like from sales to the warehouse.
• Linked records. The share of your records that join into a case across two or more systems.
• Clear endings. The share of cases that end in a result you can check: paid, refunded, won, lost, fixed or delivered. The ending is how a buyer can tell if an AI got the task right.
• Variety. How many different paths cases take, how often work gets redone, and how often something goes wrong. Real work is messy, and buyers want that.
• Breadth. How many parts of the business the data covers: sales, finance, operations, support, supply chain and HR.
• Usable cases. About how many cases have at least 3 steps, 2 systems and an ending. We split them into endings a computer can check, like paid or not, and ones a person has to judge.
• Links kept. The share of links between records that still work after cleaning. We aim to keep at least 95 of every 100.

Together, they add up to a short rating, with the numbers behind it. For example: "Workflow: Deep. 4 to 7 systems per case, about 11 steps, a 38-day span, 62% of cases with a clear ending, 97% of links kept."

An inbox on its own shows people talking. Add the order system and the books, and a buyer can see the whole job: the ask, the work, and whether it got paid.

That's what AI labs pay for. They want long, real tasks that cross many tools and end in a result. Piles of loose files are worth much less.

So each system you connect can do more than add volume. It can turn loose records into whole cases.

• Connect the system where the work ends. Payments often live in accounting. Fixed issues live in the help desk. Without the ending, a case is only half there.
• Connect the systems in between, like orders, shipping or scheduling.
• Include more history. Longer spans give more whole cases.
• Keep the email and chat that go with the work. We link them to cases by the order and ticket numbers inside them.
• Include every part of the business you're allowed to license, not just one team.

Your report names the one change that would help most. For example: connect your accounting system, so orders link to payments.

We work out the score on cleaned data, inside our locked-down system. Only totals come out. Never the records.

On a Confidential deal, buyers see broad ranges only. A detailed picture of how a company works, plus its industry and size, could point to the company. We never show a count under 11, and nothing in the score names your customers.

Before your data is connected, any score is an estimate from your own answers, and it says so. After that, it's measured.