Businesses using the latest generation of artificial intelligence have a surprising requirement: actual human intelligence.
Companies that succeed with modern generative AI tools—such as Microsoft ’s Copilot, Salesforce ’s Agentforce, or offerings from a raft of new startups —are discovering that in order to get real value from AI, they have to organize their data in ways they might not have before. And this isn’t a one-and-done effort. To keep their shiny new AIs up-to-date, the information they feed them must be kept constantly updated—creating more work for humans.
This could have big implications for the economy as a whole. What seems to be happening, at least among the half-dozen companies I spoke with, is something economists have observed in countless tech revolutions past. A new form of automation is simultaneously eradicating some jobs , and rapidly creating new ones .
Some of those new roles include writing, editing and organizing information. And not for other humans—but for AIs themselves.
Better inputs, better outputs
BACA Systems, a company just outside Detroit that makes big industrial robots for cutting stone countertops, is an illustrative case. Each robot is a little different from all the others, because the tech keeps evolving, and is being used in a particular and idiosyncratic factory.
Andrew Russo, head of IT at the company, says that when a customer’s robot goes down, time is of the essence. To shorten the time between when a customer reaches out and when they get their first response, BACA is working on an AI-powered customer-service chatbot, using an AI platform from Salesforce.
At first, the documents this chatbot drew from were 400-page manuals. If a customer needed a pointer to the original documentation behind the chatbot’s answer, just offering a giant manual wasn’t much help, says Russo. So the same experts and service reps who handle customer queries are part of the team revising those manuals, and breaking them up into smaller individual articles that are relevant to particular aspects of the robots. The resulting “knowledge base,” when fed into the company’s AI, means more accurate responses that point to particular articles. The end result is much more helpful and usable for both customer service reps and the customers who might be handed this information by the company’s chatbot.
Importantly, creating such a system isn’t about reducing head count, adds Russo. Instead, it gives human service reps time to study an individual customer’s case and follow up with additional help.
Companies I talked to used either Salesforce’s AI platform or Microsoft’s. Salesforce Chief Executive Marc Benioff has been proclaiming that Microsoft isn’t delivering real value for its customers .
“Microsoft has deceived our customers,” Benioff told me in a recent episode of The Wall Street Journal’s new podcast , “Bold Names.” He went on to say that Microsoft can’t come up with a list of customers transforming their business with Copilot.
But what I found in speaking with Microsoft’s customers was that their experience with the company’s AI systems wasn’t much different than that of Salesforce’s customers. Many of them are using AI in exactly the ways that Microsoft’s chief marketing officer for AI at work, Jared Spataro, told me they are, from speeding up how they handle email to accelerating the creation of marketing campaigns.
In other words, as AI is becoming commodified —more like, say, cloud storage. It’s now table stakes for tech giants that sell to businesses, rather than a differentiator. And a lot of companies are opting to use the AI offered by whichever company they’re already storing their data with. Unsurprisingly, given Microsoft’s market dominance, for a lot of companies that means going with its AI .
Every company needs a ‘knowledge base’
Every company I talked with mentioned that to get real value out of their shiny new generative AI systems—no matter the application—they needed to overhaul or double down on their strategy for feeding it the kind of data that today’s AI excels at processing—“unstructured” data.
About 90% of the data most companies have is this kind of data—not numerical, not in a spreadsheet, but in the form of documents, emails, manuals, customer-service chats, contracts and the like. And the real value of today’s generative AI for companies is in unlocking it.
Think of it as centralizing all the know-how that is typically spread out across the brains and storage accounts of all the humans in an organization. Making all of that information and knowledge available to everyone else in an organization has long been the dream of corporate IT—and generative AI can get companies one step closer to it.
At the Bank of Queensland , based in Brisbane, Australia, Chief Technology Officer Robert Wilson is using Microsoft’s AI tools for a wide variety of tasks. For example, on account of new regulation, the bank needed to search millions of contracts for unfair terms. This was only possible because his team could build a custom AI agent to accomplish the task, using Microsoft’s Copilot Studio.
Yet more examples of this phenomenon include Wiley, the publishing company, which found that organizing its customer-service data has become more important than ever, as it moves toward using AI-powered chatbots to handle most customer-service inquiries. It’s the same at London’s Heathrow Airport, where the folks in charge of digital communications found that in order to reduce hallucinations in their customer-service chatbot, they had to make sure their knowledge base was as up-to-date and accurate as possible.
At European fintech giant Finastra, AI has become an integral part of building marketing campaigns, which has sped up their production, says Joerg Klueckmann, head of corporate marketing at the company. For example, for a recent campaign about a new concept in finance, his team conducted interviews with more than 50 experts. AI helped summarize those interviews and highlight common themes. Then it helped create assets from the resulting insights. Every step of the way, humans and AI were essentially working hand in hand, even to produce the end products of the campaign, including everything from emails and videos to websites and ebooks. The entire process took about 2½ months. Without generative AI, it would have easily taken half a year, says Klueckmann.
These kinds of productivity gains were captured in an aphorism from a recent speech by Microsoft CEO Satya Nadella : “What Lean did for manufacturing, AI will do for knowledge work.”
Lean manufacturing, pioneered by Toyota in the 1950s and 1960s, has since become the operating principle behind countless other manufacturing systems, and even the e-commerce operations of Amazon .
With automation coming to knowledge work, it looks like the same forces that have reshaped blue-collar work have finally arrived for white-collar workers.
For more WSJ Technology analysis, reviews, advice and headlines, sign up for our weekly newsletter .
