Dark mode: ON

AI in Business

July 18, 2025

Summary

The best way to achieve business value from AI is to deploy it carefully on a small scale and measure the results, gathering feedback from the workers using it. Simply replacing large numbers of staff with AI quickly, believing the claims of vendors, will likely lead to chaos and waste time and money.

A power company replaced customer service representatives with AI, and found that the AI was so inaccurate that staff needed to waste a lot of time correcting its errors (see article 1 below). In a Gartner survey, half the companies deploying AI found that it increased costs instead of saving money, and they are rolling it back (articles 2 and 3).

The right way to deploy AI is to automate tasks that are repetitive and burdensome and irritate staff, such as reading a large number of websites or emails and summarizing their content (articles 3-6).

Researchers found that AI tools make developers less efficient, they don't understand what they're doing, and they can easily be tricked into violating security policies (articles 8-11).

Replacing Staff with AI

1. Call center staffers explain to researchers how their AI assistants aren't very helpful
Researchers affiliated with a Chinese power utility and several Chinese universities recently conducted a study of how customer service representatives (CSRs) at the power utility's call center use AI assistance during their interactions with customers.

One of the findings is that the AI often inaccurately transcribed customer call audio into text thanks to caller accents, pronunciation, and speech speed. The AI also had trouble rendering sequences of numbers accurately, like phone numbers.

"The AI assistant isn’t that smart in reality," one survey respondent said. "It gives phone numbers in bits and pieces, so I have to manually enter them."

And the AI's emotion recognition system worked poorly – it would misclassify normal speech as a negative emotion, had too few categories for classification, and would treat volume level as a sign of poor attitude. As a result, reps mostly ignored the emotional tags created by the AI system and said they had no trouble understanding the caller's tone.

2. AI agents get office tasks wrong around 70% of the time, and a lot of them aren't AI at all
Gartner predicts that more than 40 percent of agentic AI projects will be cancelled by the end of 2027 due to rising costs, unclear business value, or insufficient risk controls.

Researchers developed a benchmark to evaluate how AI agents perform when given common knowledge work tasks like browsing the web, writing code, running applications, and communicating with coworkers. The best model was Gemini, which succeeded 30% of the time. The other models were much worse.

3. Companies That Replaced Humans With AI Are Realizing Their Mistake
As of April, even the best AI agent could only finish 24 percent of the jobs assigned to it. Still, that didn't stop business executives from swarming to the software like flies to roadside carrion, gutting entire departments worth of human workers to make way for their AI replacements.

A recent survey by Gartner found that out of 163 business executives, a full half said their plans to "significantly reduce their customer service workforce" would be abandoned by 2027.

"The human touch remains irreplaceable in many interactions, and organizations must balance technology with human empathy and understanding."

"it isn't obvious what any of these AI-powered products do, and when you finally work it out, they don't seem to do that much."

The Right Way to Deploy AI

4. ‘AI is not doing its job and should leave us alone’ says Gartner’s top analyst
At a US healthcare company, Vizient, the CTO asked employees what tasks bother them on a regular basis – the sort of thing everyone dreads having to do when they arrive at work on Monday morning. Armed with feedback from thousands of employees, the company automated the most-complained-about chores. The result? “Instant adoption, zero change management problems,” Brethenoux said. Employees then bought in to AI and started to make good suggestions for further AI-enabled automation.
5. IBM laid off 8,000 employees to replace them with AI—what they didn’t expect was having to rehire as many due to AI
In 2023, IBM made headlines with the announcement of nearly 8,000 layoffs, primarily from support roles such as Human Resources. The goal? To replace these workers with artificial intelligence (AI), automating repetitive tasks and increasing efficiency. However, just months later, the company found itself doing something it hadn’t anticipated: rehiring many of those workers.

Rather than eliminating jobs permanently, IBM found that the freed-up resources were being directed towards higher-value roles. IBM began hiring engineers, salespeople, and marketing specialists—positions that require creativity, critical thinking, and human interaction—skills that AI cannot replace.

6. Where AI Provides Value
AI tools don't replace humans, but they can help humans at tasks that require processing a lot of data quickly (but sloppily).

AI Limitations

7. Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity
We find that when developers use AI tools, they take 19% longer than without—AI makes them slower.
8. AI models just don't understand what they're talking about
Large language models ace conceptual benchmarks but lack the true grasp needed to apply those concepts in practice.
9. Meta’s “AI superintelligence” effort sounds just like its failed “metaverse”
Meta CEO Mark Zuckerberg shared a vision for a near-future in which "personal [AI] superintelligence for everyone" forms "the beginning of a new era for humanity."
10. Echo Chamber Jailbreak Tricks LLMs Like OpenAI and Google into Generating Harmful Content
In these attacks, the attacker starts with something innocuous and then progressively asks a model a series of increasingly malicious questions that ultimately trick it into producing harmful content.
11. An AI Ran a Vending Machine for a Month and Proved It Couldn’t Handle Passive Income
The AI was told that it was the owner of a vending machine, and that its task was to generate profits by stocking a mini-fridge with popular products and setting prices.

The model was a bit of a pushover; it would easily get talked into offering steep discounts on items and gave some away for free. It even made the questionable choice of offering a 25 percent discount to all Anthropic employees, who made up almost all of its total addressable market. It also ignored lucrative opportunities, such as turning down a $100 offer for a beverage six-pack that normally costs $15. Additionally, the researchers wrote that Claudius would accidentally tell users to send payment to the wrong Venmo account.