Editor’s note: To our subscribers who have been following the publication for any length of time, you may notice some changes to the look and feel of our site and newsletter.
The 93/7 split explains a lot of what I see inside large organisations. But "the technology is ready" is only partly true, and where it isn't true matters. In regulated industries the fundamentals are still unsolved: security, transparency, the ethics of the decision itself, and long-term provenance — where proof of a decision made by a machine has to be reproduced exactly, possibly years later, in front of a regulator or a court. The AI acts now arriving will test every one of these. The technology might work in a demo. Enterprise readiness is a different bar, and there is a seriously long way to go.
There is a second problem sitting underneath the investment imbalance. A large share of what is being done under the AI banner could have been done with existing, simpler technology. Workflow, rules, decent data plumbing. Not everything needs AI, and reaching for it by default is part of why so much of that 93% delivers nothing.
Then there is the receiving end of these systems. Take recruitment. AI screening is the biggest disaster going — capable people filtered out before a human ever sees them, no explanation, no appeal. If you want a proof point for why deployment without redesign fails, that is it. The organisation saved time, and the cost transferred to people who never agreed to the system, cannot see it, and cannot challenge it.
Reimagining how the organisation works has to include the people the system points at, not only the people operating it.
The 93/7 split explains a lot of what I see inside large organisations. But "the technology is ready" is only partly true, and where it isn't true matters. In regulated industries the fundamentals are still unsolved: security, transparency, the ethics of the decision itself, and long-term provenance — where proof of a decision made by a machine has to be reproduced exactly, possibly years later, in front of a regulator or a court. The AI acts now arriving will test every one of these. The technology might work in a demo. Enterprise readiness is a different bar, and there is a seriously long way to go.
There is a second problem sitting underneath the investment imbalance. A large share of what is being done under the AI banner could have been done with existing, simpler technology. Workflow, rules, decent data plumbing. Not everything needs AI, and reaching for it by default is part of why so much of that 93% delivers nothing.
Then there is the receiving end of these systems. Take recruitment. AI screening is the biggest disaster going — capable people filtered out before a human ever sees them, no explanation, no appeal. If you want a proof point for why deployment without redesign fails, that is it. The organisation saved time, and the cost transferred to people who never agreed to the system, cannot see it, and cannot challenge it.
Reimagining how the organisation works has to include the people the system points at, not only the people operating it.