- Therapy and emotional support rank as the most common individual use of generative AI for the second consecutive year, with more than a quarter of top use cases involving AI forming opinions or decisions on behalf of users.
- Flock Safety cut its ALPR data retention from 30 days to 7 days and will mandate audit tools for all law enforcement customers by January 1, 2027, following more than 50 agency cancellations in the first half of 2026.
- McKinsey research finds that while AI adoption is near-universal, only around one-third of organisations have scaled beyond pilot programmes, a gap the data frames as organisational, not technical.
Generative AI’s most popular personal use case, according to new research, is therapy and emotional support, a finding that sits uncomfortably alongside a surveillance company overhauling its data practices under pressure from law enforcement cancellations and enterprise adoption figures showing that most organisations are still stuck at the pilot stage.
Therapy Tops the Use Case List
Therapy and emotional support have ranked as the top use case for generative AI for the second consecutive year.
The same research surfaces what the authors call “thinkslop”: the tendency to delegate not just tasks but the reasoning that should precede them to AI. Participants described opening a chatbot before clearly defining what they needed, or accepting first outputs without verification. One participant reported growing difficulty constructing their own reasoning after relying on AI for nearly every piece of text. More than a quarter of the top use cases involved asking AI to form opinions or make decisions on behalf of the user. The researchers propose using AI as a sparring partner, to stress-test arguments rather than generate them, as a counterweight to the habit.
Flock Safety Rewrites Its Data Rules
Flock Safety announced on August 13, 2026, that it is cutting its default automated license plate reader (ALPR) data retention period from 30 days to 7 days. The announcement also introduced what the company calls “Evidence Mode,” which allows specific records to be preserved only for active investigations rather than held as a matter of course. By January 1, 2027, Flock will require all law enforcement customers to use an audit tool that flags abnormal searches and can impose proactive lockouts pending review; officers will need to provide a case code for every search, with emergency overrides automatically flagged for administrator review.
The pressure behind these changes is concrete. More than 50 agencies or jurisdictions cancelled or suspended Flock contracts in the first half of 2026, citing civil liberties concerns, according to reports. The policy changes reflect a calculation that proactive accountability measures are preferable to continued contract attrition. Whether mandatory audit tools and shorter retention windows satisfy critics who object to ALPR surveillance on principle, rather than on implementation grounds, is a question the new policies do not answer.
Agentic AI Hits Governance Friction
Agentic AI, systems that reason through multi-step tasks and execute actions without per-step human instruction, is seeing real enterprise uptake. The governance picture has not kept pace. Deploying systems that autonomously manage complex workflows raises accountability questions that differ in kind from those posed by conventional AI assistants: when an agent makes a consequential decision across a chain of steps, tracing responsibility back through that chain is harder than reviewing a single model output. The EU AI Act’s liability provisions are already drawing attention to exactly this gap. Enterprises that have not built audit trails into their agentic deployments before those systems reach production are likely to find retrofitting them far more expensive than designing for it from the start.
The Scaling Gap
McKinsey research puts the scaling problem plainly: AI adoption across enterprises is near-universal, but only around one-third of organisations report scaling it meaningfully beyond pilots.
The structural tension is between bottom-up adoption, which is fast and widespread, and top-down scaling programmes, which are slow and frequently stall. The McKinsey figures suggest most organisations are at a stage where AI is useful but not yet systematically valuable, and that the distance between those two states is less a technology problem than an organisational one. For boards tracking where investment is actually going, the gap between AI spend and measurable ROI has become a standing agenda item.
