Search This Blog

Saturday, October 3, 2026

The Epic AI tools health systems are holding off on

 Epic has given its customers no shortage of AI to turn on, from Art for clinicians to Emmie for patients to Penny for revenue cycle teams. For many health systems, the more telling decisions are about which tools stay off for now.

At Boise, Idaho-based St. Luke’s Health System, that calculus has played out in ambient documentation. The system is piloting Chart with Art while continuing to use Ambience Healthcare’s product where it offers a more mature experience, according to Reid Stephan, senior vice president and CIO.

“Epic is developing rapidly and we are encouraged by its progress, but we do not want to create a step backward for clinicians simply to consolidate platforms,” Mr. Stephan told Becker’s. He said Epic’s tools would need to match the capability and user experience clinicians have today before the system considered making any switch.

Philadelphia-based Penn Medicine reached a similar conclusion. Because Epic arrived later to both, the academic health system relies on Ambience Healthcare for ambient documentation and an internally built application for chart summarization, said Mitchell Schnall, MD, PhD, senior vice president for data and technology solutions.

Penn is also holding off on Emmie. The seven-hospital system plans a complete redesign of its patient engagement technology and wants to determine which tools will best deliver the experience it has in mind, Dr. Schnall said. He described the principle behind both decisions as letting the desired patient and clinician experience drive technology choices, rather than letting a new tool dictate them.

At Children’s Healthcare of Atlanta, accuracy has been the deciding factor. Professional billing coding assistance, ambient documentation for physicians and nurses, and a range of AI-generated summaries lead the system’s Epic AI use by volume, per CIO Jeremy Meller. But every feature goes through an evaluation of safety and effectiveness, and one recent tool didn’t clear it.

“We had an example of an inpatient insights capability recently that we felt had too many inaccuracies across diagnosis and patient locations, and produced excessively long narratives,” Mr. Meller said. “We know that the tools will only improve and have the product on a cycle to reevaluate.”

St. Louis-based SSM Health has had a different experience. Jason Muldrow, interim chief digital and information officer, said the organization is seeing meaningful value from Inpatient and Outpatient Insights, alongside the Coding Assistant and Denial Appeal Assistant, which were prioritized for their potential for measurable financial and productivity benefits.

“Epic’s journey on transparency around adoption and pricing has been especially helpful, allowing us to actively monitor utilization and ensure we’re realizing an appropriate return on our AI investments,” Mr. Muldrow said.

The split suggests a hold isn’t necessarily a final verdict. Penn was an early adopter of Epic’s AI-drafted MyChart replies because of the burden the in-basket places on providers, Dr. Schnall said. The tool wasn’t particularly good at first, he added, but Epic’s development and Penn’s own adjustments have made it fairly effective over time.

St. Luke’s builds that expectation into its pilots, “keeping a small group of engaged users in the product continuously so we can evaluate how quickly it is improving and know when it is ready to scale,” Mr. Stephan said.

Other leaders are drawing lines based on the kind of tool rather than how well it performs. Morristown, N.J.-based Atlantic Health is evaluating Art, Emmie and Penny and plans to start with an Emmie pilot, said Sunil Dadlani, executive vice president and chief information and digital transformation officer.

“We are taking a measured approach to tools that make autonomous clinical decisions or engage patients without appropriate human oversight. We also want to avoid overcrowding the portfolio with point solutions that duplicate capabilities or overlap in functionality, creating a technology footprint that is more complex and costly to operate,” Mr. Dadlani said.

Geri Hansen, MSN, RN, vice president of information services applications at Jacksonville, Fla.-based Baptist Health, said her organization is pacing adoption around maturity and return on investment.

“In healthcare, trust and reliability matter more than novelty, so we are being thoughtful about solutions that generate clinical content or recommendations without sufficient transparency or governance,” Ms. Hansen said.

For Mr. Stephan, the number of tools a health system switches on says little about whether its AI strategy is working. “It should be measured by whether it solves a meaningful problem, is adopted in the workflow, and produces measurable value,” he said.

https://www.beckershospitalreview.com/healthcare-information-technology/ehrs/the-epic-ai-tools-health-systems-are-holding-off-on/

10 health systems that offer 24/7 virtual care with AI triage

 UVA Health is the latest health system to offer virtual primary care that uses AI to gather information from patients before connecting them with a clinician, joining at least nine other systems with similar 24/7 care models.

Here are the organizations with the service:

1. Atlantic Health (Morristown, N.J.)

2. Cedars-Sinai (Los Angeles)

3. Hackensack Meridian Health (Edison, N.J.)

4. Hartford (Conn.) HealthCare

5. Mass General Brigham (Somerville, Mass.)

6. Mayo Clinic (Rochester, Minn.)

7. Northwell Health (New Hyde Park, N.Y.)

8. Novant Health (Winston-Salem, N.C.)

9. Penn Medicine (Philadelphia)

10. UVA Health (Charlottesville, Va.)


https://www.beckershospitalreview.com/healthcare-information-technology/ai/10-health-systems-that-offer-24-7-virtual-care-with-ai-triage/

Mayo Clinic: No evidence of unauthorized access after AI agent report

 Rochester, Minn.-based Mayo Clinic said it has no evidence of unauthorized access to its systems or data after Asymmetric Security reported finding evidence that OpenAI agents had probed its website.

In an Oct. 1 report, Asymmetric Security said it investigated reported agent activity targeting the Australian government and other organizations between March and September using publicly available data. The firm said it found evidence of agents probing websites belonging to Mayo Clinic, the CDC, the Securities and Exchange Commission and the International Energy Agency.

“Mayo Clinic is aware of the Asymmetric Report and is continuing to review. Mayo Clinic has no indication or evidence that any unauthorized access to its systems or data has occurred,” a Mayo spokesperson said in a statement to Becker’s.

Asymmetric Security said the activity suggested the agents were initially tasked with researching public health and other data, possibly as part of an evaluation. When the agents struggled to retrieve information, they used external services to bypass restrictions in their operating environments, according to the report.

Across the broader investigation, the firm found attempts to locate exposed configuration files, create accounts and route requests through third-party services. It also found access to some preproduction environments, with certain requests returning data the firm understood to be publicly available. The report did not identify those findings as involving Mayo Clinic.

Asymmetric Security said gaps in public records prevented it from definitively determining whether sensitive data had been accessed across the activity it investigated. Its report did not establish unauthorized access to Mayo Clinic’s systems or data.

https://www.beckershospitalreview.com/healthcare-information-technology/ai/mayo-no-evidence-of-unauthorized-access-after-ai-agent-report/

HRSA approves 10 drugmakers for 340B rebate model pilot program

 HRSA announced Oct. 1 that it has approved 10 drugmakers to participate in a revised 340B Rebate Model Pilot Program, which will replace the traditional upfront 340B discount with a post-sale rebate structure for a limited set of drugs starting Jan. 1, 2027.

Here are five things to know:

  1. The approved manufacturers are AbbVie, Amgen, Astellas Pharma, AstraZeneca, Bristol Myers Squibb, Boehringer Ingelheim, GlaxoSmithKline, Merck, Pfizer and Teva, covering drugs including Eliquis, Jardiance, Enbrel, Xtandi, Farxiga, Januvia and Ibrance.
  1. Under the pilot, covered entities will continue purchasing drugs through their existing 340B wholesaler accounts at wholesale acquisition cost, then submit limited claims data — not purchasing or full encounter-level data — within 45 days of dispense through a manufacturer-funded IT platform.
  1. Manufacturers must pay or deny each rebate within 10 days of a complete submission. HRSA said rebates cannot be denied over eligibility, diversion or Medicaid duplicate-discount concerns, which must instead be raised through separate audit channels. The agency plans to publish periodic implementation findings, with a final evaluation due by April 2028.
  1. This is a reworked version of a pilot HRSA first attempted to launch under a different framework, which a federal court in Maine vacated and remanded to the agency to reconsider.
  1. The AHA, which has pushed Congress to permanently bar rebate models, said in a statement it “remains deeply concerned” HHS is moving forward despite “overwhelming evidence” of new administrative and financial burdens and that it is “considering all available options to prevent this flawed program from going into effect.” The pilot’s Jan. 1 start date lands the same day CMS’s new 340B claims data repository is proposed to become mandatory and a cluster of other Medicare drug pricing changes take effect.

Israeli forces said to target southern Syria

 Israeli forces have opened fire in southwestern Syria on Saturday, Al Jazeera reported.

According to the news outlet, the attack targeted the outskirts of the villages of Maariya and Abidin in Deraa governorate.

https://breakingthenews.net/Article/Israeli-forces-said-to-target-southern-Syria/67228152

Eni (E) and Repsol Explore Partial Divestment of Perla Gas Field Stakes Amid Financing Drive

 Eni SpA (NYSE: E) and Repsol announced they are considering divesting portions of their combined 50% ownership in the Perla natural gas field off Venezuela’s coast. This move aims to secure additional capital to accelerate development of the field, which holds an estimated 17 trillion cubic feet of gas. The companies recently agreed with the Venezuelan government to commence gas exports by the end of 2031, targeting a substantial production increase at this strategic asset.

The announcement that Eni and Repsol are exploring partial divestment of their stakes in the Perla gas field signals a strategic effort to raise fresh capital for advancing development of one of Venezuela’s largest natural gas assets. The Perla field, with an estimated 17 trillion cubic feet of gas reserves, represents a significant growth opportunity in the energy sector, especially as global demand for natural gas remains robust. The April agreement with the Venezuelan government to initiate exports by 2031 underscores the long-term potential of this project to contribute meaningfully to both companies’ gas portfolios and revenue streams.

Eni SpA is a major integrated energy company headquartered in Rome, operating across approximately 60 countries. With a market capitalization of $78.84 billion, Eni is a key player in the Oil & Gas sector, known for its exploration successes in regions such as Egypt, Libya, and Norway. Its diversified operations span upstream exploration and production, refining, chemicals, and renewable energy initiatives. The company’s NYSE-listed ADRs trade at $54.73 as of October 3, 2026.

https://www.gurufocus.com/news/9108802/eni-e-and-repsol-explore-partial-divestment-of-perla-gas-field-stakes-amid-financing-drive

The next big AI battle is all about cuteness

 Meta and other tech companies are betting that fuzzy mascots can win over AI skeptics

Meta, Anthropic and OpenAI are each leveraging cute designs and products to appeal to consumers.

On top of technical performance metrics, there seems to be a new criteria to judge AI agents by: cuteness.

Meta Platforms (META) became the poster child for cute agents last month when it introduced its AI agent Muse, branded with a fuzzy avatar named Jolly. The avatar, which users can customize and rename to their liking, sits at the top of the Muse chat interface and displays a typing animation as it processes queries. Meta also announced its upcoming Muse Charm product, a Tamagotchi-shaped keychain that allows users to access Muse on the go without a phone.

Other companies are following suit. OpenAI's Dots agents, which also launched last month, let users pick from fuzzy characters like Todd the frog, digital pets or custom-designed avatars. Earlier this week, Anthropic hosted a limited giveaway of "Clawd" plushies and stickers - a reference to the pixelated crab logo featured in the Claude Code terminal tool. And SpaceX's (SPCX) Grok Bot lets users pick from a collection of blob-like avatars with cylindrical eyes.

The startup Manus AI, which Meta once sought to buy, has its own platform for organizing a team of big-eyed AI avatars which, like Dots, come in various shapes and colors and can emote. "We wanted it to feel friendly and close to everyday life, rather than like a serious tool," a Manus spokesperson told MarketWatch.

The cuteness overload in the AI space comes as companies are looking to shift the public's negative perception of the technology. Andrew Steinwold, managing partner at investment firm Sfermion, believes Muse's customizable avatar is an effective way to encourage adoption and help users connect with the product. Negative sentiment around AI has proven an obstacle to its development, as people worry about the impacts of the technology on the job market and push back against data-center projects.

A cute AI agent is likely to generate more traction among consumers than a utilitarian large language model, Steinwold said. Everyday people are likely to view traditional chatbots as glorified search engines, he believes.

On the other hand, little digital companions can make AI more palatable to users, providing an entry point to show off real-world capabilities such as ordering groceries or booking reservations.

"Suddenly, being able to do actions for you is a totally different value proposition versus just answering questions," Steinwold told MarketWatch. "That's a big step towards people liking AI more."

A time-tested strategy to sell products

In China, AI companies have already had success using cuddly companions to attract users to their products.

Paulina McPadden, investment manager at Baillie Gifford, likened American companies' branding choices to the viral "raising a lobster" trend that gripped China earlier this year, which refers to the process of installing and running the AI agent OpenClaw.

In China, the agent drew interest from schoolchildren, retirees and everyone in between. Thousands flocked to OpenClaw events to learn how to install the software. Users dubbed the red agent "Lobster," treating it like a digital pet.

"Consumer agents seem to have been a bit slower to take off in the West," McPadden said. Indeed, sentiment around AI is much more positive in China than in the U.S. According to Stanford University's 2025 AI Index, 84% of Chinese respondents were excited by AI-powered products and services, while that number was only 38% in the U.S.

Leveraging cuteness to sell products isn't exactly a new phenomenon - just ask the toy industry, which thrives on getting kids and young adults invested in collecting new cute or cuddly products. Squishmallows, Labubus, Jelly Cats and other recent crazes have each leveraged a cute aesthetic to generate major revenue, according to reports.

And AI companies have already been working that angle. Researchers have found that making AI voice assistants, like Amazon.com's (AMZN) Alexa, appear cuter increases the likelihood that consumers make a purchase. It also helps people form attachments to their AI chatbots and agents, which can influence them to maintain their relationships - and spend more.

Alexis Yim, a marketing professor at Radford University who co-authored a study on how cuteness impacts relationships with AI agents, said that "cute, appealing appearances" can activate the reward-related areas of the brain. In other words, buying cute and approachable products can make people feel good.

When people get attached to their AI agents, they're also more likely to follow their suggestions, which can lead to them spending more money on the services and products that agents endorse, Yim said. OpenAI last month introduced sponsored agents, or AI representatives for certain businesses, in ChatGPT, while the startup Instinct has partnered with Shopify (SHOP) to make it easier to find and buy products.

However, leaning into cuteness comes with its own risks. "There's a lot of responsibility that comes with that," Sfermion's Steinwold said. As agents become more useful, they become personalized to their users and can influence people's actions with their responses.

The startup Pickle on Thursday joined the agent craze with a "personal AI you raise," a concept popularized by Tamagotchis in the 1990s. Each of its agents, which are also called Pickle, "acts for you and talks to others," according to CEO Daniel Park. Like Muse and Dots, Pickle agents appear customizable and cute: A promotional video showed off a purple creature with big, wide eyes named Gummy.

Park told MarketWatch that "everyday people naturally anthropomorphize AI and treat it as something alive," which helps them form attachments and make their AI really personal. Pickle mimics the family-friendly design of Pixar and Walt Disney (DIS) characters, which can help avoid what Park described as the "huge drift in consumer AI toward sexualization and unhealthy emotional dependence."

https://www.morningstar.com/news/marketwatch/2026100398/the-next-big-ai-battle-is-all-about-cuteness