Software rarely fails inside a tool. It fails in the gap between two of them.
Tap what you run. Get a seam count, the failure points that matter, and a card to show your team
Count the unofficial layer too. It is usually where the real risk is.
Nothing matches. Add it to the count with Other tools at the bottom of this step.
Other tools not on this list. Each one still connects to something. Add them to the count and the assessment will flag the unmapped seams.
Each known failure pattern between two kinds of tool carries a severity from 1 to 3. The severities are judgment from operating real stacks, not measurements. Argue with them.
Two tools from the same vendor suite are scored one step lower, because the vendor builds and maintains that connection. A pair already at the lowest severity drops out. Password, AI and automation tools are not discounted, since their risk is access and oversight, not the integration.
The total is multiplied by who finds out first: a customer counts most, an alert and a runbook least. Automation glue adds 15 percent. The score climbs quickly at first and flattens toward 100, so a bigger stack scores higher without ever passing it.
Possible seams is every pair among your tools, n times n minus 1, divided by 2. Named failure points are the ones this tool has a pattern for.
The Bureau audits the seams in a real stack: what connects, what is held together by hand, and what to fix first.
Book a seam audit