What Do AI Companies Know About the Future of Work?

Summary

WSJ Tech News Briefing – "What Do AI Companies Know About the Future of Work?"

Date: June 30, 2026
Host: Imani Moiz, The Wall Street Journal
Episode Theme:
An exploration of how AI is transforming office work, the risks of sharing AI chatbot accounts, and what the biggest AI companies’ internal use of their own tools can teach us about the future of work.

---

Main Themes

• The Challenges of Sharing AI Chatbot Accounts: Why sharing access to AI chatbots is riskier and more complex than streaming accounts.
• Tech Companies ‘Dogfooding’ AI: How giants like Google and OpenAI internally use agentic AI workflows that hint at the coming workplace revolution.
• Workplace Impact and Friction: The immediate effects, challenges, and lessons from integrating AI deeply into business operations.

---

Segment 1: The Risks of Sharing AI Chatbot Accounts

(00:17–05:01)

Key Points

• Account Sharing is Spreading: Just as people shared Netflix logins, students and friends are now pooling subscriptions to premium AI chatbots (e.g., ChatGPT, Claude).

• Real-life Examples:

  • Many sharers are students in their early 20s trying to save money.
  • One user managing Crohn’s disease had private health details discovered by a study partner via shared account usage. (01:40–02:07)
  • Workflow disruptions occur—shared usage jumbles content, contaminating personalized outputs for things like study flashcards and cover letters.
    • "It would kind of draw on information from her roommate's study... and sort of taint the quality of the flashcards she was doing." — Natalie Kaufman (02:12)

• Quality Decreases Over Time:

  • Models get "confused" as training data from diverse users muddles intended outcomes. (02:59–03:26)

• What the Companies Say:

  • OpenAI specifically discourages password sharing, offers group chat features for collaboration—but insists on individual accounts for security and personalized AI experiences. (03:30–04:04)

• Privacy & Cybersecurity Concerns:

  • Sharing exposes intimate or sensitive information.
  • Risk isn’t limited to the intended sharing group—others could access data via chat logs or screenshots.
    • "Try not to outsource things like personal finance or medical related to a chatbot... it's a bit more private and intimate." — Natalie Kaufman (04:10-05:01)

---

Segment 2: AI ‘Agents’ Reshaping White-Collar Jobs

(06:00–10:44)

Key Points

• The New Workflows:

  • AI is moving beyond simple tasks, now handling multi-step, complex “agentic” processes—AI acts as the operator, humans as fact-checkers.
    • "The big thing right now is using agents to handle multi step tasks... the AI is just handling more sophisticated work that previously a person would have done." — Katie Binley (07:06)

• Real-World Implementation at Big Tech:

  • Google:
    • On the finance team, an AI agent validates invoices against contracts, flags discrepancies, and is projected to save $200 million annually by preventing overpayments. “And it seems to be doing quite a good job...” — Katie Binley (07:42)
  • OpenAI:
    • Legal team uses Codex to analyze new employee disclosures, draft replies—replacing some junior associate work.
      • "She's still hiring junior associates, but... they would be taking over some of the Codex review." (09:06)

• Why This Matters:

  • These AI companies are testbeds for new tech—what they do preview shifts likely for the broader workforce.
    • “Tech workers have long experimented with their own products and used them themselves before passing them on to everybody else. So, they call it dogfooding.” — Katie Binley (08:25)
    • "Dogfooding or drinking their own champagne. That's another one I've heard." — Bell Lynn (08:54)

• Job Impact So Far:

  • No direct layoffs attributed to AI yet. Instead, productivity rises while headcount stays flat.
  • Some roles are evolving—junior associates may supervise or extend AI output rather than do initial work themselves. (09:06–09:47)

• Friction and Challenges:

  • AI adoption can cause cross-team friction. E.g., sales using AI for contract reviews may irk legal teams.
  • AI success can create operational backlogs, prompting the need for new AI agents to handle downstream tasks.
    • “With Google... the agent's actually been working so well that they ended up with like a backlog of these flag discrepancies. And then they have a separate operations team that needs to go to take those issues to the suppliers… building another agent that’s going to start initiating the process.” — Katie Binley (09:47–10:44)

---

Segment 3: The Human Side of AI Transformation

(Interspersed throughout, emphasis at 06:00–06:27 and 11:12–11:52)

Key Points

• Beyond Tech: Organizational Change is Crucial:

  • “How are you redesigning the organization or the capabilities of the people that are actually doing the work and linking that to the technology? These other pieces are 70, 80% of the work to get that right.” — Dan Swan, McKinsey Senior Partner (06:00)

• Skills for the Future:

  • Employees need to master prompting AI tools, not just technical analysis.
    • "Whereas before you may need to be really good at analysis, now the skill might be how do you prompt these tools to get the right answers and outcomes." — Dan Swan (11:22)

---

Notable Quotes & Memorable Moments

• “It would kind of draw on information from her roommate's study... and sort of taint the quality of the flashcards she was doing.”
  — Natalie Kaufman (02:12)

• "It's not like sharing a Netflix account... It's a bit more private and intimate, especially for people who start using a chatbot as sort of like a therapist or just like a repository for all of their personal information."
  — Natalie Kaufman (04:10)

• "The big thing right now is using agents to handle multi step tasks... the AI is just handling more sophisticated work that previously a person would have done."
  — Katie Binley (07:06)

• "Tech workers have long experimented with their own products and used them themselves before passing them on to everybody else. So, they call it dogfooding."
  — Katie Binley (08:25)

• "We’re more talking about getting into the chassis of the organization... These other pieces are 70, 80% of the work to get that right."
  — Dan Swan (06:00)

• "The skill might be how do you prompt these tools to get the right answers and outcomes... when you see that all humming together, it can be really, really special."
  — Dan Swan (11:22)

---

Timestamps for Important Segments

• 00:17 — Introduction to the risks and realities of sharing AI chatbot accounts
• 01:32 — Natalie Kaufman discusses why people share chatbot accounts
• 02:12 — Workflow issues and privacy concerns from shared accounts
• 03:30 — How OpenAI and others address account sharing and group features
• 04:10 — Data security risks and the importance of cyber hygiene
• 06:00 — Discussion on how tech companies are ‘rewiring’ organizations with AI; McKinsey’s perspective
• 07:06 — Workplace is shifting: complex task automation with AI agents
• 07:42 — Case study: Google’s finance team and invoice validation agent
• 09:06 — No layoffs yet, but evolving job responsibilities
• 09:59 — Friction in implementation: team workflows, new backlogs, more agents needed
• 11:22 — Skills shift: prompting AI tools as key for future roles

---

Conclusion

This episode dives beneath the surface of flashy AI headlines, focusing on two urgent realities: the underestimated privacy and workflow risks of sharing AI chatbots, and a clear-eyed look at how leading tech companies are prototyping the future of white-collar work by embedding AI ever deeper into complex business processes. While layoffs haven’t taken hold, the nature of work and worker skills are already changing rapidly, with friction and adaptation challenges emerging. The advice: adopt cautiously, keep your data secure, and prepare to learn new skills for an AI-transformed workplace.

Transcript

[A]

Countless Companies invest in AI tools without tying them to tangible business outcomes. Join McKinsey later to learn how leaders rewire their organizations for sustained impact and value.

[B]

Welcome to Tech News briefing. It's Tuesday, June 30th. I'm Imani Moiz for the Wall Street Journal. Move over Netflix. Sharing your password has moved from streaming to chatbots. But while the practice might save you money, could lead to some unexpected headaches. We're getting into why sharing a chatbot may cause more problems than you bargained for. Then we peek inside the biggest tech companies to look at how the people behind your favorite AI tools are offloading their work to agents. We're unpacking what other office workers can learn from their experiments and where the tech is causing friction. But first, the cost of subscriptions for AI chatbots like Gemini, Claude, and ChatGPT can add up quickly. To save money, some people are sharing their passwords and their accounts. But the stakes around sharing a chatbot password are a bit higher than giving out details to your Netflix account. Instead of exposing your reality TV habit, sharing a chatbot can expose details about your work, health, and personal life. Natalie Kaufman, who reported on this for the Wall Street Journal, spoke with our producer Julie Chang about why you should think twice before pooling your logins.

[C]

You spoke to some people who are sharing passwords for their chatbot accounts. Can you tell us about some of them and why they were sharing a password?

[D]

So a lot of the people I interviewed are about 21, 22. You know, college students on a budget don't really want to pay for premium chatbots themselves. So a lot of them decided to share the wealth with their friends who are also studying with them. And one person I spoke actually uses it to manage his Crohn's disease. And his study partner found some intimate details about him when she was able to use the account. So that was unfortunate for him.

[E]

Yeah.

[C]

Other than that example you just gave us, what were some of the other issues that people experienced?

[D]

Yeah, a lot of it has to do with workflow disruption. So not everyone is studying the same thing at college. So one person I spoke with, she's a nursing student and her roommate does education studies. And so when the nursing student was tried trying to come up with some exam style questions to practice with the chatbot, it would kind of draw on information from her roommate's study. So, you know, it would put in jargon from whatever the roommate is doing in class and sort of taint the quality of the flashcards she was doing. So that was Kind of challenging. And then other people were using the chatbots to come up with text for cover letters, and it would just lump everyone together and make things a little bit jumbled.

[C]

Gotcha. So does sharing a password make the chatbot work less well?

[D]

Yeah, I would say so. At least the people I spoke with had that experience. A lot of the time, it starts out where everything is fine and no one's having any issues. And then as the chatbot is trained on all the different data over time, it kind of learns from that and starts just kind of getting confused. You know, your brain can only process so many things at one time. The same goes for, like, an algorithm. So it just kind of all gets mixed up and create some really bizarre outputs.

[C]

And what have the companies behind the chatbot set about sharing their products?

[D]

Yeah, so OpenAI, the ChatGPT maker, very much discourages people from sharing their accounts. They say someone else wants to use OpenAI's products, they can sign up for their own account. And then they also have a feature called group chats. So up to 20 people can use that tool at one time and do kind of like collaborative workflow, but everyone has their own separate account. So it's something that they say is not just like, a business necessity for them, like wanting people to sign up for their own accounts and pay, but also has to do with data security and making sure that the model is personalized to whoever's using it.

[C]

Speaking of data security, are there any cybersecurity or privacy risks to sharing a chatbot account?

[D]

Yeah, absolutely. Especially when you do start sharing things like medical or legal information or even family struggles. Other people can see that. And also some cybersecurity experts I spoke with said you don't even know who else outside of the small group of people you're sharing it with can see that, because people can take screenshots of the chat log. So it's just really important. The same way as you wouldn't do, you know, banking or trading stocks on public wi fi. Try not to outsource things like personal finance or medical related to a chatbot. I guess, like, the lesson is that it's not like sharing a Netflix account or another streaming service. It's a bit more private and intimate, especially for people who start using a chatbot as sort of like a therapist or just like a repository for all of their personal information. So it's just, you know, a matter of good cyber hygiene to keep things separate.

[B]

That was the Wall Street Journal's Natalie Kaufman speaking with our producer Julie Chang. A reminder that we want to hear from you. Have you ever seen a post on social media that you thought was real only to later realize that it was AI generated? What did you see? Why did you believe it? How did you feel afterwards? Shoot us an email to tnbsj.com or leave us a voicemail at 212-416-2236. That's 212-416-2236. And if you're a listener on Spotify, leave us a comment. We hope to hear from you coming up. The companies building AI are also among its biggest beta testers. We'll hear what tech giants like Google and OpenAI can teach the rest of us about the future of office jobs. That's after the break.

[A]

True AI value comes from rethinking workflows, not just layering on tech. Here's Dan Swan, senior partner at McKinsey.

[F]

When we talk about rewiring, we're more talking about getting into the chassis of the organization. And that means, yes, there's technology to it, but how are you redesigning the organization or the capabilities of the people that are actually doing the work and linking that to the technology? So for us, the technology, yes, is critically important. But these other pieces are 70, 80% of the work to get that right.

[B]

For many white collar workers, using AI means asking it to summarize a document or draft an email. But inside the companies creating AI tools, employees are starting to hand over more complicated tasks. These internal experiments are offering an early glimpse of how work itself could change as businesses become more comfortable relying on AI. WSJ reporter Katie Binley spoke with several employees about how tech companies are putting AI agents into action on the job and told our colleague Bell Lynn, a reporter for the Wall Street Journal Leadership Institute, about what she learned.

[E]

So Katie, how has the nature of white collar work changed because of these new AI tools?

[G]

The big thing right now is using agents to handle multi step tasks. So, you know, previously you might have seen people using AI to help write an email for them or some research. And now what you're seeing is this implementation of what are called agentic workflows where the AI is just handling more sophisticated work that previously a person would have done. And then now the person is basically a fact checker.

[E]

And give us a few examples of how employees at OpenAI, Anthropic or Google are using AI or agents in their everyday work.

[G]

Yeah, there's a bunch of interesting examples at Google on their finance team, they now have an agent who is doing, doing invoice validation work. And so the agent is reviewing what the invoice looks like as compared to the terms of a contract and then looking to see whether there are any discrepancies there. And then the people on the finance team are sort of going over the agent's output to see how good a job it did. And it seems to be doing quite a good job. They've estimated that it's going to save the company about $200 million a year on what would have otherwise gone to overpaying incorrect invoices.

[E]

And why are we looking at employees of these tech companies? Why are they a good indication of where white collar work is headed?

[G]

These are the companies making some of the most important tools that are then being marketed to the rest of the workforce. So it's an interesting window into what work could look like for a lot of white collar workers. And tech workers have long experimented with their own products and used them themselves before passing them on to everybody else. So this is. They call it dogfooding. So this is just a nice example

[E]

of dogfooding or drinking their own champagne. That's another one I've heard.

[G]

There you go. Oh, I like that better. I hadn't thought of that, but yeah.

[E]

Are any of the companies you talk to downsizing or eliminating some of their workers as a result of automating more work with AI?

[G]

So no one said, oh, we are now able to cut five jobs that we otherwise would have kept on this team. Instead, what Google described, for example, was that they're able to remain flat that while producing more. And then with OpenAI, for example, on their legal team, I spoke with an associate general counsel there who's been having Codex do the work that she said a junior associate otherwise would be doing. So that includes like analyzing disclosures from new employees and looking for conflicts and drafting replies. But this associate general counsel said she's still hiring junior associates, but that they would be taking over some of the Codex review that she's actually been handling herself.

[E]

And as we know, implementing new technology inside large companies is never that easy. So have any new challenges or issues come up since these AI companies started implementing AI agents?

[G]

Yeah, so I spoke with a guy who runs an AI advisory consulting firm, and he just said that you can end up with friction across teams. So, for example, if your sales team is now having AI review contracts, your legal team might not want them doing that or might take issue with it. And so he said it's very common. It's something that a lot of companies are talking about right now with Google, for example, with that invoice validation agent I was talking about earlier. He said that the agent's actually been working so well that they ended up with like a backlog of these flag discrepancies. And then they have a separate operations team that needs to go to take those issues to the suppliers. But so what they're ending up doing for that surprise, surprise is building another agent that's going to start initiating the process of going to the supplier when the operations team otherwise would have.

[B]

That was wha reporter Katie Binley speaking with our colleague from the Wall Street Journal Leadership Institute, Bell Lynn. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by Julie Chang with supervising producer Katie Ferguson. I'm Imani Moiz for the Wall Street Journal. We'll be back later this morning with TNB Tech Minute. Thanks for listening.

[A]

AI transformation is more than a tech initiative. It requires a broader rewiring of human skills and ways of working. Here's Dan Swan, senior partner at McKinsey

[F]

one of the biggest unlocks we see is the human being capabilities around the technology. Whereas before you may need to be really good at analysis, now the skill might be how do you prompt these tools to get the right answers and outcomes. And so for us, it's a really big priority to help companies to understand how do you need to change the operating model of your organization and change the underlying capabilities of the people that are interacting with the technology. And when you see that all humming together, it can be really, really special.

[A]

Discover how McKinsey helps organizations rewire to out compete with [e-mail] TechMoves this content

[D]

was created by Custom Content from WSJ,

[G]

a unit of the Wall Street Journal advertising department.