A view of the week, written on the basis of the industry news collected on the news page that week.
In insurance, AI pilots are running, but few of them change how claims and underwriting actually get done. In banking, leaders report productivity gains while staff keep correcting what the systems produce, because the process underneath was never set up properly. A muddled workflow stays muddled, even when a model makes it faster. In both cases what ends up in the record is the start, not the effect.
The process comes before the tool
In my project work the order usually runs backwards: first the tool is chosen and bought, then the question starts about which steps it should actually replace. When a few hundred cases run through an agent and nobody can say whether it works better or worse now, that gap is the whole story. What gets automated is then a workflow nobody had written down before, and that only shows up once the case numbers grow. Whoever does not record the comparison with the earlier state cannot prove the difference later.
AI is an amplifier, not a repair shop
Research on software teams puts it in one line: AI amplifies what an organization already has. Where collaboration holds, throughput grows. Where it does not, the first thing to grow is the pile nobody can review any more, and the bottleneck moves to the end of the chain. That is exactly where the least is measured.
After the training comes the work, not the result
The bottleneck at the end of the chain shows up in learning too. A leadership program can be strong, get excellent feedback, and still change nothing about behavior, because after the last session nobody looks again. Leaders return to daily work with useful frameworks and a shared vocabulary, and at year end the sponsor asks why nothing has moved. The end of the training says nothing about the practice that follows. The same gap shows up in work with AI support: the report is finished faster, and it remains unclear what the person still had to think through themselves.
What I take from it
What stays with me from this week is a checklist, not a tool tip. And I decide before the start who will look again afterwards: where, at what point, and by what sign we will see the benefit fading. Without that point we notice the difference only once the benefit is long gone. How would you notice that the benefit of your last project has started to fade?
Berthold Plenter, Mindset Consulting. Making technical projects successful.
Related services: Technical project management · Leadership & team development · Change & transformation · Organisation & processes
This issue is also available in German: KI verstärkt, was schon da ist. Wer einen unklaren Prozess automatisiert, bekommt ihn schneller unklar..