What is AI best at right now? Improving products you already have

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

This disconnect has big implications for how we use AI at work and in our personal lives. It will also shape the landscape of winners and losers among the startups and tech giants that are currently pouring tons of cash into building this technology.

Chatbots, which ushered in the generative AI wave, with their impressive ability to mimic human expression, seemed to herald an era of omniscient, oracular automatons with whom we would interact like science fiction AIs. These self-driving bots—including OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude—were the subject of hopes and fears, not to mention huge investments.

They could also one day be seen as a transitional state on the way to generative AI becoming part of everything we do – much like word processors were once physical devices you could buy, rather than just another piece of software on a computer.

Take a look at Apple’s announcement from Monday about the new AI features that will be woven throughout the operating system in upcoming iPhones. They ranged from the fun — AI-generated custom emoji — to the practical — summaries of incoming text messages and emails and improved intelligence for the Siri voice assistant. Apple is also working with OpenAI to embed ChatGPT into aspects of its new AI offerings.

When Google announced its latest Pixel phones in August, it touted a slew of new features designed to entice people to buy the new model, including a voice assistant, call transcription, photography tricks and weather forecasts.

Even Microsoft, which has been promoting its paid Copilot products based on generative AI, has also promised to integrate generative AI throughout Windows 11 and in the form of Copilot software running on “AI computers.”

Less science fiction, more utility

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

The AI ​​capabilities may not be as impressive as those of standalone bots. But direct integration with the software and systems we already use allows these AIs to not only tell us how to do something, but actually do it for us. The improvements in the efficiency of these tools make this type of “agency” AI one of the most exciting recent developments in AI.

Google Strategy Fellow Dan Seifert recently shared a simple example, showing in a Threads post how, with the click of a button on his new Pixel phone and a quick tutorial, he was able to import all the important dates and times from a Word document his child’s school had sent to his Google calendar.

ChatGPT could achieve something similar, but it’s more complicated and requires significantly more human expertise and work, since standalone AIs like this aren’t deeply integrated with the software that already runs our world.

These AI features could benefit Apple, Google, and other established tech companies if their addition encourages consumers to buy more gadgets and software—but not necessarily at a level that will yield quick returns on AI infrastructure investments. And the reality of AI-as-a-feature will likely fuel even more failure and consolidation among independent AI startups.

OpenAI is following this trend while continuing to invest in its chatbots. Recently, when I asked an OpenAI spokesperson if he thought that in the future most people would use AI primarily as a feature in other software rather than as a standalone product, he started nodding so vigorously that I stopped my question.

“AI is a very powerful general-purpose capability, and you have to meet users where they are,” the spokesperson says. “That’s why the partnership we announced with Apple, for example, is so critical to driving AI use cases, rather than people remembering AI as a chatbot or an app.”

Summarization and synthesis are quickly becoming part of existing search offerings—for example, in the AI-generated summaries at the top of some Google searches. Image generation can now be done on-device on the latest Apple and Google phones. And automating business processes like invoice processing—which may be where generative AI has the biggest impact on our lives and the economy—is something that works best when AI is part of existing software.

Concrete examples in an unlikely industry

I recently delved into a surprisingly deep secret about the use of artificial intelligence in one of the industries we least expect it from: construction.

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

AI helps estimate the costs of new projects, manage and track workers on site, and detect problems with construction plans to avoid the common and costly problem of having to rebuild parts of the structure.

Procore, which sells construction management software, has built AI into its platform to make it easier for workers to get answers to questions about how their company typically operates. This type of improved, chat-based search is one of the most common uses of generative AI in companies of all types. For example, it’s common in systems designed to assist—or even replace—customer service representatives.

Construction giant JLL has created several generative AI tools for its own use, says Bruce Beck, the company’s chief information officer for enterprise and corporate systems. They include a pair of chatbots for construction policy and human resources, and an automated report generator. His unit also uses a generative AI system from Mountain View, Calif.-based Orby to automate the handling of the tens of thousands of invoices JLL must process each year.

According to Beck, the software could help JLL reduce its invoice processing headcount by about 300 over time.

These tools work much more like previous waves of AI—software buzzing in the background, doing things for us—than the kinds of human-like AIs we’ve been told are just around the corner. While that might not be as exciting as a future built on AI super geniuses that we spend our days talking to and even receiving guidance from, it could be a lot more useful.

Write to Christopher Mims at [email protected]