What is AI best at now? Improving products you already own

In reality it is shaping up as more a product feature than a new product category. As recent announcements from Apple and Google show, it is proving most useful as a technology to soup up the gadgets and software we already use, rather than reset the world order.

That disconnect has big implications for how we are using artificial intelligence at work and in our personal lives. It’s also going to shape the landscape of winners and losers among startups and tech giants currently investing a deluge of cash in building this technology.

The chatbots that started the generative AI wave, with their impressive ability to mimic human expression, seemed to portend an era of all-knowing, oracular automatons that we’d interact with like the AIs in science fiction. These stand-alone bots—including OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude—have been the focus of hopes and fears, not to mention massive investment.

They may also someday be viewed as a transitory state on the way to generative AI becoming part of everything we do—like the way word processors used to be physical devices you’d buy, rather than just another piece of software on your computer.

Look at Apple’s announcement on Monday of new AI features to be woven throughout the operating system in its forthcoming iPhones. These ran from the entertaining—AI-generated custom emojis—to the practical—summaries of incoming texts and emails and enhanced intelligence for the Siri voice assistant. Apple also is partnering with OpenAI to embed ChatGPT in aspects of its new AI offerings.

And when Google announced its latest Pixel phones in August, it touted a set of new features to entice people to upgrade, including a voice assistant, phone call transcription, photo tricks and weather summaries.

Even Microsoft, which has been pushing paid generative AI-driven Copilot products, also has promised to integrate generative AI throughout Windows 11 and in the form of its Copilot software running on “AI PCs."

Less sci-fi, more useful

Chatbots, with their conversational and image-generation abilities, have been a uniquely compelling way to demonstrate the power of generative AI to millions of people. OpenAI says ChatGPT now has 200 million weekly active users.

AI features might not be as impressive as the stand-alone bots. But direct integration into the software and systems we already use allows these AIs not only to tell us how to do things, but to actually do them for us. The efficiency enhancements of these tools makes this sort of “agentic" AI one of the most exciting recent developments in artificial intelligence.

Dan Seifert, a Google strategy staffer, recently showed off a simple example. He demonstrated in posts on Threads how, with the click of a button on a new Pixel phone and a brief instruction, he could import all the important dates and times from a Word document that his child’s school had sent out into his Google calendar.

ChatGPT can accomplish something similar, but it is more complicated, and requires significantly more know-how and work on the part of a human, because stand-alone AIs of this sort aren’t deeply integrated into the software that already runs our world.

These AI features can benefit Apple, Google and other established tech companies if adding them spurs consumers to buy more gadgets and software—but that’s not necessarily at the level that will bring quick returns on their investment in AI infrastructure. And the AI-as-a-feature reality is likely to fuel even more failures and consolidation among stand-alone AI startups.

OpenAI is embracing this trend, even as it also continues to invest in its chatbots. Recently, when I asked an OpenAI spokesman whether he thought that in the future, the way most people would use AI, most of the time, was as a feature in other software rather than a stand-alone product, he began nodding so vigorously that I cut my question short.

“AI is a very powerful general-purpose capability, and you’ve got to meet users where they are," says the spokesman. “That’s why for example the Apple partnership we announced is so critical, if we want to imbue use cases with AI, instead of people remembering AI is a chatbot or an app."

Summary and synthesis are quickly becoming part of existing search offerings—for example in the AI-generated summaries atop some Google searches. Image generation can now be done on-device in the latest Apple and Google phones. And the automation of business processes like processing invoices—which could be how generative AI has the biggest impact on our lives and economy—is something that works best when an AI is part of existing software.

Concrete examples in an unlikely industry

Recently, I plunged down the surprisingly deep rabbit hole of how AI is used in one of the industries you’d least expect: Construction.

Executives at builders, software startups and other suppliers in the industry told me companies are using all kinds of AI-powered tools to speed up labor-intensive tasks in an industry desperate to increase efficiency. (McKinsey regularly notes that per-worker productivity in construction has stagnated for decades.)

AI is helping to estimate the cost of new projects, manage and track workers on-site, and detect issues with construction plans to avoid the common and costly headache of having to rebuild parts of a structure.

Procore, which sells construction-management software, has embedded AI as a feature in its platform to make it easier for workers to get answers to questions about how their company typically does things. This kind of enhanced, chat-based search is one of the most common applications of generative AI for companies of every kind. For example, it’s common in systems designed to help customer service reps—or even replace them.

Construction giant JLL has created a handful of generative AI-powered tools for its own use, says Bruce Beck, Chief Information Officer of enterprise and corporate systems at the company. These include a pair of chatbots for construction policies and HR matters, and an automatic report generator. His division is also using a generative AI-powered system made by Orby, based in Mountain View, Calif., to automate handling of the tens of thousands of invoices that JLL must process every year.

Use of the software could over time help JLL reduce its human workforce of invoice processors by about 300 people, says Beck.

These tools work a lot more like previous waves of AI—software humming away in the background, doing stuff for us—than the kinds of human assistant-like AIs we’ve been told are just around the corner. While that might not be as exciting as a future built on AI supergeniuses that we spend our days talking to and even taking direction from, it may actually be a lot more useful.

Write to Christopher Mims at christopher.mims@wsj.com