AI RADAR · ARCHIVE
All posts from the AI Radar
News with our verdict and studies with a key figure, newest first.
61 posts
How many AI leaders report full control over AI investment?
18%
Full control over AI investment decisions
AI leadership rarely means full control over investment
AI agents go shopping. Shops decide what they may do
How often did executions of discovered tasks receive full credit?
50%
Full credit
Finding a task does not mean completing it
How many report an established strategy for the workforce effects of technology change?
16%
Established strategy in place
New technology often lacks an established workforce strategy
How many report spending at least 25 percent of their AI budget on avoidable work?
39%
At least 25 percent of the AI budget
Data problems consume a substantial share of many AI budgets
How many report AI spending significantly above expectations or budget?
59%
Significantly above expectation or budget
AI spending often exceeds plans
How many firms use AI?
52%
Large firms · 250+
Large firms use AI much more often
How many report fundamental workplace redesign for AI?
10%
Fundamentally redesigned for AI
AI tools do not yet make a redesigned workplace
How do employees mainly use time saved with AI?
28%
More work in the same time
Time saved with AI often goes into more work
How many need as long for review as for producing the work themselves?
37%
Review and revision of AI work
AI work can require extensive review
What do workers who use generative AI at work report?
81%
AI saves time
AI users report time savings more often than better quality
For which critical AI applications is provider switching prepared?
7%
of respondents report organisation-wide preparation
Only 7 percent plan AI provider switching across the organisation
How long is an uninterrupted focus session on average?
37.8 min
Heavy AI use
Heavy AI use is associated with shorter focus sessions
How does AI change sales compared with unavailable support?
16%
AI chatbot instead of unavailable support
AI sales advice raises revenue compared with unavailable support
Mixed learning groups broaden teachers’ AI use
AI narrows the performance gap between education groups
What share of points do IT apprentices achieve on architecture tasks?
65%
With AI
AI improves performance on more demanding IT tasks
What do leaders report about AI in their business?
81%
More speed
More speed is not deep transformation
What is expected before sharing AI-supported work?
64%
Read and verify
AI-supported work needs checking before handoff
How long did the first writing task take on average?
27.8 min
With planning support
More planning improves AI writing but takes time
What new work emerged around AI?
12%
Directing agents, setting standards and building systems for them
AI creates new tasks for directing agents
How did developers spend their time working with AI agents?
33%
Formulating prompts or checking AI responses
AI agents need instructions and checking
Did an AI incident reach the board or C-suite?
80%
Reached the board or C-suite in the past year
AI incidents reach the top of the organisation
Are contributions increasing due to generative AI?
59%
More incoming contributions due to generative AI
More AI contributions need enough review time
Has AI changed sales volume per employee?
83%
No impact over the past three years
83% report no productivity impact from AI yet
How deeply is AI integrated into the business?
18%
Fully integrated into core operations
Using AI does not mean embedding it in core operations
What do the surveyed developers report about AI?
87%
Deliver work faster
More speed with AI does not remove every bottleneck
How deeply is AI embedded in the surveyed organizations?
34%
Embedded in some business functions
Using AI does not yet mean embedding it in daily work
What share of conversations with Claude focused on instructions, documentation and oversight?
10%
Autumn 2025
At Anthropic, more Claude sessions focus on instructions and oversight
How many respondents report rules for stopping AI initiatives?
31%
For AI pilots before production
Few respondents report rules for stopping AI initiatives
How often does the AI answer questions about team chats correctly?
41%
Without this additional organisation
Organised team chats: more correct AI answers in the test
How reliably does AI complete payment tasks?
40%
All three attempts successful
AI can fail again on the next attempt
How long does it take to retrace an AI agent’s actions after an incident?
74%
say it takes hours or days to retrace an AI agent’s actions
Investigating an AI incident often takes hours or days
AI pace and fundamental change
64%
AI pace on track or ahead
AI rollout speed and changed work are different things
AI returns and organizational change
80%
Report positive AI returns
Financial returns are one part of AI success
AI agents in automotive development: use and control
68%
Use or test AI agents
Many test AI agents; few report mature controls
Who may reuse employees’ expertise?
77%
Allow AI to reuse knowledge
Employees less readily allow expertise to be shared with AI
People need to detect errors without AI too
An AI test does not say how long AI can work alone
Exaggerated claims about AI use or knowledge
48%
of surveyed US employees admit at least occasionally exaggerating AI use or knowledge
Reported AI use can be exaggerated
Growing demand for hands-on AI development
+51%
Building AI tools, automating workflows and AI engineering
Demand is growing for people who can build AI solutions
Longer AI texts need more checking
What would change without your AI?
Withdrawal from AI innovation work
Over 80%
of participating experts withdrew at the law firm studied
AI project work needs a place in everyday work
More AI training is only weakly linked to extra time saved
A polished AI draft does not establish expert judgement
A new AI model needs suitable workflows
Human review at similar AI accuracy
39%
GPT-5.4
Similar AI accuracy, different amounts of human review
AI value needs to remain after review and rework
Prepared documents make evidence easier to reach
90%
Preselected relevant documents
AI also needs to find the right information
AI assessments need an expert cross-check
Test the whole AI working system
Better work with AI does not automatically mean learning without it
Tracking AI costs and value together
12%
of surveyed executives say they consistently link AI costs and value across their company