Enterprises want to take advantage of AI, but figuring out what applications stick is taking a lot of trial and error.

That lack of clarity is illustrated in the accelerating rate at which businesses are dropping AI projects: According to S&P Global Market Intelligence, the number of companies that scrapped AI initiatives grew from 17% in 2023 to 42% in 2024, with the average enterprise cutting 46% of AI proofs-of-concept.

There are a number of reasons that these projects don’t make it out of the “infancy phase,” said Bob Rogers, chief product and technology officer of Oii.ai and co-founder of BeeKeeper AI. For one, a lot of companies are in a discovery phase with AI, in which they’re “throwing spaghetti at the wall and seeing what sticks,” he said.

Many find that scaling projects outside of small environments or datasets can be quite tricky, said Rogers. “They hit some fairly serious snags with the scruffiness of real life data.”

Factors like cost, privacy and security issues similarly put pressure on CIOs to be stringent about which AI projects they scale and which ones they scrap, he said.

Plus, hiring AI teams to actually build and manage these projects is a challenge in and of itself, said Rogers.

That leads many to the “build or buy” equation, said Rogers. While building an internal model that works only on your enterprise’s data can make it more niche, the difficulty of scaling may lead these firms “more in the buy direction,” he said, choosing models or AI integrations offered by outside sources. “The maturity of offerings coming from outside is going to be better,” said Rogers.

Internal projects need to make “that leap of making security, reliability, integration with other tools industrial-strength,” Rogers said.
So how can enterprises have better AI sticktoitiveness? One way is to set benchmarks from the start of a project, said Rogers. If you don’t know what exact metrics your AI project can improve, then figuring out whether or not a project will see returns in the long run is all the more difficult.

“Benchmarking is a critical component,” said Rogers. “If you don’t have as part of your plan a benchmark of what it is that you’re really improving … if you hit some tough sledding, it’s really, really hard to keep the momentum.”

Having the foresight to account for security problems and the “weirdnesses” of real-life data can also help enterprises get their AI projects out of the sandbox, Rogers said. “None of the AI projects that are hypothetically thrown away are doomed to failure for sure,” he said.

“Quite often there’s actually an opportunity to move forward,” said Rogers. “But the wherewithal of an organization to see through that noise and get to the next level is very challenging.”

Apple may have added another piece to the puzzle of its rumored Apple Car.

The company filed a patent application for a “user tracking” seat headrest for better audio control. In the filing, Apple describes a headrest with “one or more speakers” mounted into it, to provide the occupant with better audio output. However, this tech essentially applies spatial audio to a headrest, a concept seen in a host of different Apple devices thus far.

This headrest takes into account the placement of “the occupant’s head and/or ears” by tracking their location and movement within an “enclosed environment,” relying on cameras and sensors to do so. The audio is then modified based on a user’s movement, as the “movement of the user’s head and/or ears can have a noticeable effect on the acoustic experience of the user,” Apple said.

Apple noted that these headrests can help provide personalized audio experiences “intended only for the occupant of the seat,” and offer an “ambience channel” to the user.

One other note to consider: While Apple didn’t explicitly say that this headrest is for a car seat, the company said that this tech can be implemented in a “stationary apparatus,” such as a conference room or other type of room, or a “moveable apparatus,” including vehicles like a train car, an airplane, an autonomous vehicle, a boat, a ship or a helicopter.

Apple has been working on the Apple Car, reportedly called Project Titan, since 2014. The company has filed for more than 200 patents for tech related to the not-yet-real vehicle, including for physical components like windows, seats and suspensions, as well as for high-tech security systems and autonomous capabilities. The company reportedly aims to launch the vehicle by 2026, and plans to price it at less than $100,000.
Given that Apple’s go-to is high-end (and higher price point) products, the Apple Car would likely be marketed as a premium or luxury vehicle, said Bob Bilbruck, CEO of consulting firm

Captjur. Once Apple Cars hit the roads, the first drivers will likely be the die-hard fans of the company, he said, similar to the initial audience for the launch of the Vision Pro.

“I would say it’s going to be a premium product, just like all the other products,” said Bilbruck. “I’m assuming you’ll see some pretty dedicated Apple people at first that will pay up for the features and functions they want, maybe more so than anything that Tesla, Ford, Chevy or Dodge brings out.”

For this tech specifically, Bilbruck noted that it could hit the market sooner than the Apple Car itself. The company could license this tech to other carmakers that want to offer top-of-the-line audio experiences, similar to how Apple CarPlay is integrated into hundreds of different models of vehicles. But at the end of the day, Apple likes to “own the ecosystems that feed into those products,” said Bilbruck, so whatever Apple-based vehicle features hit the proverbial shelves before launch would likely just be a peek at what’s to come.

With no cars on the road – or even an official announcement – whether the Apple Car will find success is yet to be seen. But the company does have a few things going for it, Bilbruck said, should it enter the automotive arena. For one, the company already has hundreds of physical storefronts to sell its vehicles, making deployment as easy as making space for a car in a showroom, he said.

But more importantly, the company has a consistent track record of putting out products that people trust, and a built-in audience of dedicated “early adopters” that “wouldn’t have any problem with taking a risk on a new product,” said Bilbruck.

“Some companies aren’t that dependable, but Apple really is,” he added. “Everything they bring out is really quite bulletproof from the get-go on a new product introduction.”

JPMorgan Chase is not a fan of Brittany Spears, apparently.
The financial institution is seeking to patent a system to automate detection, elimination and prevention of “toxic combinations for personal information data” — essentially two or more pieces of personal information that, when paired together, become highly confidential, such as a name and an account number.
“When certain types of PI [personal information] are combined, the classification of that information may increase, even though the individual parts of the combination may be public by themselves,” the company noted in its filing.
While this patent is fairly broad, JPMorgan describes a scanning tool that uses one or more “rulesets” configured to identify toxic combinations of personal information in a database or computer code, continuously scanning to flag anything that could be dangerous. If the system flags anything, it notifies the user who performed the scan via an email or other notification, and presents the results of the scan through a user interface.
Because different personal information has different levels of risk (for example, personal information may be classified as “public, internal, confidential, and highly confidential), the user also gets a “degree of confidentiality” of the toxic combinations in the dataset.
The company noted that this tool can perform scans on code or databases in both “development and deployment phases” of an application. The rulesets of this scanner are also continuously updated with the latest organizational or legal policies regarding different types of personal information, as “many applications and programs do not follow organization guidance on toxic combinations and unknowingly expose PI to the public through these toxic combinations.”
As a company that processes millions of transactions an hour and deals with millions of pieces of personal data, adding any new layer of cybersecurity isn’t a bad thing. Plus, if patented, JPMorgan could easily license this tech out to other financial institutions, adding financial cybersecurity to its software toolkit.
The issue of “toxic twins” is a prevalent one in the data security tech field. The fact that a financial institution like JPMorgan is working on tackling this issue is itself a benefit, said Ari Weil, VP of marketing at data security firm Cyera.
“If (JPMorgan) can get the technology and approach right, It’s going to help them and probably other banks that they would license the software to become a lot more private and a lot more secure overall,” Weil said.
But the problem is that the technology JPMorgan lays out is, frankly, quite vague. One hitch: There are a lot of potential toxic combinations to identify. A ruleset-based scanner, as this patent suggests, tends to look for individual pieces of personal information. “Trying to figure out combinations requires a lot more logic, and also a lot more combinations,” said Weil.
Because of this, a method that relies on simple rules and pattern matching may lead to a host of different problems, including user-end latency and false positives and negatives.
“People scrunch up their faces at AI and machine learning, but the fact of the matter is … a human defining all the rules is just not going to get all the combinations,” Weil said.
The patent itself doesn’t go too in depth about how exactly JPMorgan’s scanner will work, said Weil. But if the company decides to implement AI that can rapidly identify toxic pairs, the scanner may have a chance of working correctly.
“The pro [case] is that something like this is really needed,” said Weil. “The way that they choose to implement it is to-be-determined. The question with a patent like this is ultimately going to be where does this evolve to?’”