AI — Not Always that Productive.

For months we’ve been hearing about how AI is going to replace jobs. Some companies have even started doing away with beginner positions and are letting AI do that work. That’s a whole set of worries for people trying to break into different professional fields of work like insurance.

Today’s worry? The amount of time it takes to fix AI mistakes and the difference between trained and untrained workers.

Yes, AI makes mistakes according to a survey from the AI orchestration platform, Zapier. Its AI Workshop survey of 1,100 companies considered serious AI users found some flaws in the system.

Emily Mabie — the Senior AI Automation Engineer at Zapier — said 92% of the companies surveyed contend AI improves their productivity. However, she also points out that workers are spending 4.5 hours a week — or a half a work day per week — correcting AI errors.

Sometimes they have to completely redo the work of AI.

The survey’s report calls it “AI workslop” that, at a first look, seems perfect. The second look finds it lacking and sometimes needing a complete redo. Mabie said 74% of companies report at least one negative outcome from low-quality AI.

“The productivity gains from AI are real. 92% of workers feel them. But so is the cleanup work,” Mabie said. “The companies seeing the best results aren’t the ones avoiding AI. They’re the ones who have invested in training, context, and orchestration tools that turn AI from a sloppy experiment into a managed process.”

• Just 2% say they don’t often have to revise an AI production
• 74% have seen negative outputs from low-quality AI
• 28% have seen work rejected by stakeholders
• 27% have had security issues from AI
• 25% have seen and heard customer complaints

Data from the workslop list is the biggest headache:

• 55% say data analysis and visualizations need the most cleanup
• 48% of writing tasks need fixing

The good news for those using AI, 97% work workers with AI orchestration tools say it improves their productivity.

Zapier’s survey makes a real distinction between workers who are trained to use AI and those that aren’t. Those without training are six times more likely to admit that AI makes them less productive.

• 94% of AI trained workers claim AI improves their productivity
• 69% of workers not trained in AI say it improves their productivity
• However, trained employees spending more time with AI spend more time fixing it

Different firms experience different outcomes:

• Finance and accounting teams see 85% of negative consequences from AI use
• Those firms spend 4.6 hours per week doing cleanup
• Engineering and IT firms spend about 5 hours a week doing cleanup
• However, the negative consequences are only 78%

Companies spending 5 or more hours per week on AI corrections report losing revenue, and worse, clients and deals.

“The solution isn’t fewer tools, it’s better infrastructure,” Mabie noted. “Orchestration, training, and proper context convert AI from a vague experiment into a managed process where the extra cleanup is the cost of doing more meaningful work faster, rather than the cost of pretending you are.”

Here are the full survey results: https://bit.ly/4a2DIYx

Source link: CityBiz — https://bit.ly/3LK4bku