The company decided to replace customer support with AI, but the plan quickly got out of hand. In the end, programmers were forced to “change their shoes” into call center operators.

Swedish giant Klarna, which allows you to pay for purchases in installments, initially boasted that it no longer needed live employees. After all, algorithms could handle the work of “700 full-time agents.” In 2024, the company’s CEO Sebastian Siemiatkowski claimed that the experiment with “artificial intelligence agents” allowed them not to hire new employees for a year. The firm presented this as a breakthrough: automation saved millions and helped avoid staff costs.

However, in May 2025, it became clear that the scheme was getting out of control. Problems with algorithms forced the company to urgently return people to support. But, as they say, breaking is not rebuilding. And, as it turned out, reassembling a department of 700 employees is not the same as laying them off. As a result, even those who had nothing to do with it ended up working in call centers: programmers, marketers, and anyone else who was available.

After the failure, Semiatkovsky changed his tone dramatically. If he used to dream of “completely replacing people with algorithms,” he now promises that Klarna will become “the best company where there will always be someone to talk to.” We had to back off from loud statements about AI, as we felt too painfully that The artificial intelligence industry is living on borrowed time, and it is not certain that it will be able to pay it back.

This story is not only relevant to Klarna itself. For example, former OpenAI employees conducted an experiment: they entrusted AI with responsibility for a kiosk. As a result, they lost money and AI assured me that he was a real person who urgently needed to attend a business meeting at the Simpsons’ address. In addition, according to recent polls, 95% of attempts to implement generative AI in companies fail. Less than half of American executives are confident that their companies will be able to successfully complete the automation process. Klarna faced this problem publicly and in a rather embarrassing way

Source: Futurism

CLEVELAND, Ohio — The Cleveland Clinic and an AI-powered healthcare data company are exploring if artificial intelligence is better than caregivers at quickly identifying patients who might benefit from enrollment in a clinical trial.

Recruiting enough patients for various research trials is an ongoing problem. About 80% of clinical trials don’t meet enrollment deadlines, and about half of all trials don’t enroll any patients.

The Clinic and Dyania Health will collaborate on an initiative that aims to accelerate clinical trial recruitment by using medically trained large language models, the Clinic recently announced.

Dyania Health is an AI-powered healthcare data company based in Jersey City, New Jersey, and Greece.

Dyania’s Synapsis AI uses medically trained AI models to interpret data such as clinical notes, medical records, imaging and pathology, and combine it with other information, such as organ function or age, to draw accurate medical conclusions, the Clinic said.

“The future of medicine depends on building research systems that are precise, efficient, fair, and deeply connected to patient care,” said Dr. Lara Jehi, chief research information officer at the Clinic. “Through our innovative work with Dyania Health, we are creating an AI-driven foundation that helps identify the right patients for the right trials at the right time.”

The Clinic has used Dyania Health’s Synapsis AI platform in two pilot programs involving cardiology and oncology.

During the oncology pilot program, a research team led by Dr. Aaron Gerds, deputy director for clinical research at the Clinic’s Cancer Institute, evaluated Synapsis AI against two experienced research nurses. Both the nurses and AI searched medical records for patients who met the enrollment criteria for a study on melanoma, a type of skin cancer.

On average, Synapsis AI identified an appropriate trial patient in 2.5 minutes with 96% accuracy in the oncology pilot program. By comparison, a nurse specializing in melanoma found a patient in 427 minutes with 95% accuracy. An oncology research nurse had 88% accuracy in finding a patient in 540 minutes, the Clinic said.

The oncology pilot program results were presented at the American Society of Clinical Oncology annual meeting, the Clinic said.

In the cardiology pilot program, AI screened patients for a Phase 3 trial for transthyretin amyloid cardiomyopathy, a rare and potentially fatal disease of the heart muscle.

The AI system analyzed more than 1.2 million patient records and reviewed 1,476 in one week. It correctly identified 30 eligible participants, while routine trial recruitment methods found 14 patients over 90 days, the Clinic said.

The AI platform also identified patients from several sites within the Clinic health system who were eligible for the cardiology trial, widening patient representation and community engagement.

Dr. Trejeeve Martyn, director of heart failure population health at the Clinic, presented results from the cardiology pilot program at a meeting of the American College of Cardiology, the Clinic said.

“Academic medical centers like Cleveland Clinic are home to some of the most advanced clinical research in the world, yet they often face significant challenges when trying to connect patients to trials – challenges rooted in complexity, time and fragmented data,” said Eirini Schlosser, founder and CEO of Dyania Health. “Through our collaboration with Cleveland Clinic, we are creating a new standard where AI enables faster connections between patients and potentially life-changing trials.”

The Clinic has invested in Dyania Health and may benefit financially from the technology’s sale, the health system said.

TLDR

• Seaport downgraded AMD to Neutral citing slower AI accelerator growth
• Trial customers haven’t converted to large orders for MI Series chips
• Microsoft and Meta reviewing AI budgets could limit future orders
• AMD using more discounts to drive sales, pressuring profit margins
• Wall Street maintains average $184.91 price target despite concerns

Advanced Micro Devices stock dropped after Seaport Research downgraded shares to Neutral from Buy. The downgrade centers on weaker progress in AMD’s AI accelerator business.

Analyst Jay Goldberg pointed to supply chain checks showing slower momentum. AMD struggles to convert early customers into major buyers for its AI chips.

The company showcased multiple customers at its recent AI event. However, most have only purchased trial systems for AMD’s MI Series accelerators.

Customer Adoption Challenges

AMD faces difficulty turning small trial orders into large-scale purchases. Goldberg expects meaningful conversions won’t happen until future product generations arrive.

Major customers Microsoft and Meta are reviewing their AI spending budgets. These reviews could limit near-term chip orders from two key potential buyers.

The budget uncertainty creates revenue visibility issues for AMD. Management may struggle to forecast AI division performance in coming quarters.

AMD has increased its use of customer discounts and support programs. The company hopes these incentives will accelerate AI chip adoption rates.

Margin Pressure Concerns

Heavy discounting strategies raise profit margin concerns. AMD may sacrifice profitability to gain market share in competitive AI markets.

Goldberg warned that discount programs combined with weak demand visibility could hurt financial performance. This creates a challenging balance between growth and profitability.

Despite current challenges, AMD remains well-positioned long-term in AI chips. The analyst noted the company stays competitive against industry leaders.

Wall Street Remains Optimistic

Most analysts maintain positive ratings on AMD stock. Out of 37 analysts, 26 rate it Buy while 11 suggest Hold.

The average price target of $184.91 implies 14% upside from current levels. Stifel targets $190 while Raymond James sees $200 potential.

Bank of America reiterated its Buy rating with a $200 target. This reflects continued confidence in AMD’s fundamental business strength.

No analysts currently recommend selling the stock. This unanimous support shows Wall Street believes in the company’s core strategy.

AMD trades at $161.79, down 8% over the past month. Year-to-date gains of 37% still outpace many tech peers.

The stock maintains strong performance despite recent headwinds. Revenue growth continues across most business segments outside AI concerns.

Supply chain data suggests third-quarter AI accelerator shipments may fall short of earlier projections.