Honestly: from my own experience, most of the gains organizations get from AI adoption have little to do with AI.
Along the way you have to fix permissions, fix data governance, win people over, cut the redundant steps. By the time all of that is done, it looks like you could have improved efficiency without the AI anyway.
I ran into this on an earlier project. After the rollout it looked like the client had saved 80% of their time, and later I found that 80% had little to do with AI. It was structuring the existing process, the SSOT, and the pruning.
At bottom, it’s the people problem, plus the existing-process problem.
That’s also why I think shadow AI is a sore spot at a lot of companies. Rather than letting it run loose, better to find a way to coexist with it.
Two related pieces: The 2.5 Stage of AI Adoption on where adoption stalls, and Designing Enterprise AI Training on how the training side gets arranged.
This week’s video: “The AI Collaboration Methodology Distilled From Nearly a Hundred Workshops, Courses and Enterprise Rollouts” https://youtu.be/ST_v4UdC-Wc
An Anonymized Case, Plus a New Video
(An anonymized case from a past client.) Short version: a lot of what you’d call an efficiency gain doesn’t actually come from AI — it comes from restructuring the process and pruning it. Don’t adopt AI just to adopt it, or you end up polishing a turd.
The new video is about exactly this: “Want to Adopt AI? Look at How Much AI Can Actually Help Across These Seven Steps First” https://youtu.be/KgimZ5n3mC0