The slowest ship in your fleet is probably this one.

Ask which of the five readiness dimensions holds up activation most often and the honest answer, in our experience, is nearly always the same one. Not governance, though that gets the headlines. Not strategic intent, though that gets the workshop. Data and platform foundations is the least glamorous of the five, the one nobody wants to own in a steering pack, and the one that most often sets the date.

Output quality is a data quality argument long before it is a model argument. Ready means three things:

• The tenant configuration actually supports the feature you want to switch on, in full, rather than a scaled-back version of it.
• The data behind that feature is clean enough to trust, with retention defined for what the feature produces, because AI features generate new data as fast as they consume old data.
• There is a documented lawful basis for the processing you are about to switch on, sitting in writing somewhere a regulator could find it, rather than in someone’s memory of a meeting.

Workday’s own evaluation guide for CHROs opens with a version of this same argument, which is worth noticing given it comes from the platform vendor rather than a competitor or an analyst. Their first test asks where the AI lives relative to an organisation’s data: bolted on as a separate layer, which requires constant pipelines and adds governance gaps at every handoff, or embedded inside the system of record, acting on the organisation’s actual structure and permissions rather than a copy of them. Their sixth criterion goes further, arguing that a platform which only sees full-time employees cannot help a CHRO advise the board on total cost of workforce, because contractors and AI agents are now part of the picture and need the same data model as everyone else.

None of that is a coincidence. It reflects something true about how these platforms actually work, and it is exactly the dimension we see stall activation more than any other, including in organisations that have strong governance and a well-argued value case already in place.

The upside is that most of this work has to happen anyway, and it is considerably cheaper to do before activation than after. A tenant that cannot support a feature today rarely fixes itself with more AI. It needs the unglamorous work: configuration, cleansing, a retention policy someone wrote down and followed. That work will not show up in a demo, and it will not make anyone’s highlight reel, but it is the difference between a platform that is theoretically capable and one that is truly ready.

Data and platform foundations is the third of the five dimensions in our AI Readiness Pulse, and in our experience it is the dimension worth checking first, precisely because it is the one most likely to be your fleet’s slowest ship.