Comment As AI Pilots within Enterprises increasingly fail, OpenAI is pivoting to consumers. This suggests that AI is more likely sneak into the enterprise via users than to walk in through the door. But IT departments will have to deal it once it arrives.
A new technology can take over an organization in two different ways. Most B2B traditional products are sold through the front door. The CIO plays poker and eats steaks with salespeople. He places a large-ticket purchase order and then distributes the new kit to whomever the business determines needs it.
There is a variation of this process, such as the “land-and-expand” approach, where a department or team leader uses their budget for a small purchase they believe will help their employees to do their job. The product then spreads throughout the organization. Consider how Adobe dominated graphics departments or how Salesforce and CRMs attracted leaders who wanted their salespeople to stay in the organization (and ensure their contact lists stayed with the org, if they left). According to lore, Microsoft’s SQL Server was originally a departmental tool.
The point is that some tech comes into being because business leaders force everyone to use it.
There’s also the back-door method. In this case, the end-users use something they find useful in their personal life and demand that employers let them use it at work as well. Think of smartphones before iOS 4, Dropbox, Slack and the like. This was once called “the consumerization of IT in the enterprise” – and even had a whole publication and conference series dedicated to it. (RIP CITEworld (19459050].
It appears that OpenAI has been betting on the second option over the past month.
Firstly, the AI valuation leader hired Fidji SIMOto be its new CEO for applications. Simo was the Facebook app manager at Meta, the company that is now Instacart. Thomas Kurian was hired by Google in 2018 to penetrate the enterprise. He is a competent consumer executive but not the type of ex-Oracle under-boss you would hire to penetrate the enterprise.
This week, ChatGPT, a maker of ChatGPT, invested $6.5 billion in a company led by former Apple designer Jony Ive. It’s not a small sum for a business that is reportedly losing billions every year and is dependent on investors to run. They brought him on board. The Wall Street Journal reported that Ive and Altman, perhaps along with Airbnb CEO Brian Chesky as a recent Wired article suggested, are building an AI-enabled gadget they believe will ship 100 million units while adding $1 trillion to OpenAIās market value.
While we hope it is a delivery device, it’s likely to be something similar to the personal digital assistant featured on the 2014 film Her. It will look no doubt at least as beautiful as the original iPod.
Putting aside the sarcasm, this move is a good one. Who cares if AI is used by consumers to help plan a roadtrip, informal therapy sessions or astrological predictions? The stakes are small, and those who provide incorrect information derived from internet sources are not usually held legally responsible.
Successful AI implementations in enterprises are largely the result of employees saving time on mundane tasks. It is widely used by law firms to summarize documents (although legal eagles must be careful to verify the output in case it hallucinates specifics of citations). We’ve heard of similar use cases in the finance industry – it isn’t perfect for summarizing documents or pulling data, but it is faster and better than a 22-year old team. Google’s AI vision is clouded by business models hallucinations.
Programmers are also increasingly using AI in very specific tasks, such as writing repetitive code.
But the top-down AI directives aren’t working very well. IBM’s recent survey of 2,000 CEOs revealed that only a quarter had achieved the expected return on their AI pilots. Johnson and Johnsonreportedly shut down most pilots and is now focusing on select few that have proven to be valuable. The use of AI by large corporations is causing hilarious resultsfor programmers.
In regards to all those companies that claim AI allows them to layoff unneeded workers, many are now having second thoughts and hiring these workers back . (A cynic may suggest that the trope is a cover to reduce costs after executives over-hired in the COVID-19-driven technology boomlet.
Try to imagine trusting AI with tasks involving real money or life and death scenarios, such as in the banking or medical fields. In these cases, a success rate of 80 percent or even 99 percent is not enough. If an assistant at a large bank makes a mistake in a spreadsheet that costs ahref=””https://www.wsj.com/articles/BL-MBB-28421″”rel=””nofollow” “target=””_blank” “>$100,000,000 ]real heads will roll. If it’s an AI, then the only person to blame is the idiot at the top of the company who ordered everyone to use AI without providing strict guidelines for checking the results.
It is true that AI can help people perform their jobs. It is best to leave the details up to workers who will bring in the technology. It will be up to IT departments and vendors to determine all the ways that it is being abused.
This is how we see generative AI taking over business. Staff who have either allowed it to creep into their lives, or actively sought it out. No more steak dinners or golf games with the CIO. (r)
