Latest / Now Shipping: A Mind the Product podcast / Microsoft's agent playbook, Altman's AI apocalypse reversal, and Anthropic IPO | Now Shipping

Transcript

Microsoft's agent playbook, Altman's AI apocalypse reversal, and Anthropic IPO | Now Shipping · Now Shipping: A Mind the Product podcast

• 0:01
  SPEAKER_00: I'm Mike Belcito, and this is Now Shipping, the weekly AI news show for product people.
• 0:08
  SPEAKER_00: Every week I'm bringing you three AI news stories that actually matter to you as a product person.
• 0:14
  SPEAKER_00: Not the hype, not the noise, just the stuff you need to keep up with.
• 0:17
  SPEAKER_00: And this is episode one, so I'm especially excited to get started.
• 0:21
  SPEAKER_00: Brought to you by the team at Mind the Product, this is Now Shipping.
• 0:26
  SPEAKER_00: All right, we've got a lot to get into today, so let me give you the quick preview.
• 0:30
  SPEAKER_00: First up, Microsoft gave the world a behind the scenes look at how they're deploying agentic AI across their entire engineering organization.
• 0:39
  SPEAKER_00: Is it something we all should be paying attention to?
• 0:42
  SPEAKER_00: And then after that, Sam Altman just did something pretty surprising.
• 0:46
  SPEAKER_00: He walked back his own predictions about AI wiping out jobs.
• 0:49
  SPEAKER_00: He said he was, quote, delighted to be wrong.
• 0:52
  SPEAKER_00: But here's the thing.
• 0:54
  SPEAKER_00: Is the timing a little suspicious?
• 0:56
  SPEAKER_00: And then finally, Anthropic just filed what could be one of the biggest tech IPOs in history.
• 1:02
  SPEAKER_00: We're talking about a valuation close to a trillion dollars.
• 1:05
  SPEAKER_00: And yes, this is a finance story, but there is a product story underneath, and we'll talk about why.
• 1:11
  SPEAKER_00: Three stories, let's get into it.
• 1:15
  SPEAKER_00: All right, story number one is all about Microsoft.
• 1:19
  SPEAKER_00: So this week, Microsoft's internal IT organization, they call it Microsoft Digital.
• 1:25
  SPEAKER_00: They put something out that I think a lot of people might uh kind of skim over.
• 1:30
  SPEAKER_00: Maybe they wouldn't have even paid attention to it.
• 1:33
  SPEAKER_00: But it all came out of Microsoft Build, their annual developer conference.
• 1:37
  SPEAKER_00: And on the surface, it kind of looks like maybe just a statement, some sort of corporate statement about how Microsoft uses its own software internally.
• 1:46
  SPEAKER_00: But there's a lot going on here.
• 1:47
  SPEAKER_00: I think it's worth digging into, and I'll set the scene.
• 1:50
  SPEAKER_00: So Microsoft Digital is essentially the IT department for one of the biggest companies in the world.
• 1:56
  SPEAKER_00: Just within Microsoft Digital, there's over 220,000 employees.
• 2:01
  SPEAKER_00: So for the past couple of years, Microsoft Digital's been on a mission to transform how the whole company, how all of Microsoft uses agentic AI, not just a co-pilot as a productivity tool, but I'm talking about actually deploying AI agents, you know, systems that can take actions, automate workflows, make decisions at enterprise scale across the entire Microsoft organization.
• 2:24
  SPEAKER_00: Now, this week they essentially announced their playbook, how they're actually doing it.
• 2:29
  SPEAKER_00: And the framing that they use is really interesting.
• 2:32
  SPEAKER_00: They call it becoming, quote, a frontier firm.
• 2:35
  SPEAKER_00: Now, what does that mean?
• 2:36
  SPEAKER_00: Well, to them, their definition is an organization where AI agents aren't just tools that you open when you need them.
• 2:43
  SPEAKER_00: They're embedded into the workflow.
• 2:45
  SPEAKER_00: They're your digital colleagues.
• 2:47
  SPEAKER_00: That's the phrase that they use.
• 2:48
  SPEAKER_00: And I'm hearing that phrase more and more.
• 2:50
  SPEAKER_00: The VP of Microsoft Digital, this guy Brian Fielder, he put it pretty directly.
• 2:55
  SPEAKER_00: He said, agents are, quote, the most significant technological change we've seen since the shift to the cloud, arguably since the advent of the internet.
• 3:04
  SPEAKER_00: That's a big claim, but I happen to agree with them on that.
• 3:07
  SPEAKER_00: But what did they actually build?
• 3:09
  SPEAKER_00: Okay, a few things stood out to me.
• 3:10
  SPEAKER_00: First, they built something called Agent 365.
• 3:14
  SPEAKER_00: This is an internal system for managing and governing AI agents across the company.
• 3:19
  SPEAKER_00: Think of it as like an app store meets a control panel for agents.
• 3:23
  SPEAKER_00: So you could discover agents, you could deploy agents, you can also govern who's using what and how.
• 3:30
  SPEAKER_00: Because apparently, when you start letting tens of thousands of employees, hundreds of thousands of employees build and use agents, things get complicated pretty darn fast.
• 3:40
  SPEAKER_00: They also built something called work IQ.
• 3:43
  SPEAKER_00: This is essentially a measurement layer.
• 3:45
  SPEAKER_00: It tracks how AI is actually being used across the company and whether it's actually generating real value.
• 3:52
  SPEAKER_00: And honestly, this is a problem that a lot of companies haven't really started thinking about yet.
• 3:57
  SPEAKER_00: So then there's the governance angle.
• 3:59
  SPEAKER_00: Now, this is the part that maybe will get the least attention from all this, but it may even be the most important.
• 4:05
  SPEAKER_00: They published an entire guide on governing AI agents at scale.
• 4:09
  SPEAKER_00: So that means security, sensitivity labels, access controls, what happens when an agent has access to data that it shouldn't, all of that.
• 4:18
  SPEAKER_00: Now, why does this all matter to you as a product person?
• 4:23
  SPEAKER_00: I think a few things.
• 4:24
  SPEAKER_00: So, first, if you're building enterprise software, definitely pay close attention because Microsoft is essentially telegraphing what your customers are about to need.
• 4:33
  SPEAKER_00: The companies buying your product right now are trying to figure out exactly what Microsoft has been working on, and at least according to them, what they've figured out.
• 4:41
  SPEAKER_00: How you move from AI experimentation to actual enterprise deployment.
• 4:46
  SPEAKER_00: And the questions they're going to bring you are the same ones that Microsoft spent the last two years trying to solve for.
• 4:53
  SPEAKER_00: So, how do you actually govern this?
• 4:55
  SPEAKER_00: How do you measure it?
• 4:56
  SPEAKER_00: How do you get developers to actually adopt it?
• 4:58
  SPEAKER_00: If your product doesn't have answers to these questions, there's definitely a gap there.
• 5:03
  SPEAKER_00: Now, the second thing, there's this concept baked into the story that I think every product team should absolutely steal.
• 5:10
  SPEAKER_00: Hopefully, you're already stealing it.
• 5:12
  SPEAKER_00: Microsoft calls themselves customer zero.
• 5:15
  SPEAKER_00: The idea is simple.
• 5:16
  SPEAKER_00: It means before you start selling the technology to the world, you deploy it internally.
• 5:21
  SPEAKER_00: You use it yourselves.
• 5:22
  SPEAKER_00: We've heard people, you know, call it dog fooding or drinking your own champagne or whatever you want to call it.
• 5:28
  SPEAKER_00: But it's so important when you do this, you end up coming off with lessons, failures, wins, and you could take that experience back into the product and your customers can benefit as a result.
• 5:40
  SPEAKER_00: So I'm gonna I'm gonna leave this story.
• 5:42
  SPEAKER_00: I'll kind of end this story by asking you something.
• 5:44
  SPEAKER_00: Is your team doing that today?
• 5:47
  SPEAKER_00: Like, I mean, really, like not just saying it, not just you know, bringing it up in sales calls, but are you actually doing that today?
• 5:54
  SPEAKER_00: Are you actually using your own product the exact way your customers use it?
• 5:58
  SPEAKER_00: Because Microsoft's whole argument is that customer zero isn't just some nice internal benefit.
• 6:03
  SPEAKER_00: It's how you build better products.
• 6:05
  SPEAKER_00: The feedback loop is shorter, the insights are real.
• 6:08
  SPEAKER_00: When you talk to a customer about a problem, you've actually lived it.
• 6:12
  SPEAKER_00: And maybe third thing here, third takeaway is that, and maybe this is the most forward-looking, that agents as digital colleagues.
• 6:19
  SPEAKER_00: I think that's going to be a big design question very soon.
• 6:22
  SPEAKER_00: Right now, product teams are adding AI features, sprinkling a little AI on it.
• 6:28
  SPEAKER_00: But very soon, I mean, we're all moving in this direction, it seems like, where it's again, it's the agentic future seems to be our future, right?
• 6:37
  SPEAKER_00: Next wave isn't about the features, it's about the agents that operate independently within a workflow.
• 6:43
  SPEAKER_00: And the design challenges that come with that, it means things get a lot different.
• 6:47
  SPEAKER_00: How does a user know when an agent is doing something on their behalf?
• 6:51
  SPEAKER_00: How do they actually trust it?
• 6:53
  SPEAKER_00: How do they override it?
• 6:55
  SPEAKER_00: Microsoft is working on all those sorts of issues right now, but at enterprise scale.
• 7:00
  SPEAKER_00: And maybe very soon you and your company will be doing the same thing.
• 7:06
  SPEAKER_00: So interesting story.
• 7:07
  SPEAKER_00: That's story one.
• 7:08
  SPEAKER_00: Let's move on to story two.
• 7:12
  SPEAKER_00: All right, story number two is all about Sam Altman, the CEO of OpenAI.
• 7:17
  SPEAKER_00: He did something kind of surprising just recently.
• 7:21
  SPEAKER_00: He walked back one of his most provocative predictions in an interview that he gave in Sydney, Australia.
• 7:28
  SPEAKER_00: What he said was he was, quote, delighted to be wrong about the impact that AI would have on jobs.
• 7:35
  SPEAKER_00: Now, why is this statement such a big deal?
• 7:38
  SPEAKER_00: Well, for the past couple years, Altman's been one of the loudest voices in the entire world warning us that AI is going to fundamentally change and disrupt the job market.
• 7:49
  SPEAKER_00: He said that AI would, quote, probably replace most jobs that people do today.
• 7:54
  SPEAKER_00: He said entire job categories would just be completely wiped out.
• 7:58
  SPEAKER_00: And it almost became like a ritual in the AI industry, right?
• 8:02
  SPEAKER_00: Like a company launches some powerful model and then they have to remind everybody about how disruptive it's going to be and change everything, including the job market.
• 8:12
  SPEAKER_00: But this week, he's saying that the displacement that he feared, it hasn't actually materialized.
• 8:19
  SPEAKER_00: And he went further than that.
• 8:20
  SPEAKER_00: He actually said that he tried an experiment on himself.
• 8:24
  SPEAKER_00: He delegated his own Slack messages and email to an AI agent, and he found himself coming back to actually respond to the Slack messages and respond to the emails himself.
• 8:36
  SPEAKER_00: This is the CEO of OpenAI.
• 8:39
  SPEAKER_00: Apparently, he didn't love handing things over to a bot.
• 8:42
  SPEAKER_00: Kind of interesting.
• 8:43
  SPEAKER_00: Now, the data does back him up on at least the broader point of it, right?
• 8:49
  SPEAKER_00: A Yale budget lab study from May found that AI is likely not the reason for any sort of weakening in the labor market.
• 8:58
  SPEAKER_00: Through March of this year, they found that there's been no meaningful change in unemployment for workers and jobs with the highest AI exposure.
• 9:06
  SPEAKER_00: So the apocalypse so far, maybe it hasn't arrived.
• 9:10
  SPEAKER_00: Now, is that actually true?
• 9:13
  SPEAKER_00: Let me dig in a little bit more.
• 9:14
  SPEAKER_00: Now, first, Altman is not the only one who's changed his tune.
• 9:19
  SPEAKER_00: Even this week, um, Anthropic CEO, Dario Modai, almost simultaneously made a pretty identical reversal.
• 9:26
  SPEAKER_00: He walked back predictions about AI's impact on jobs and took a much, I guess, calmer and measured tone.
• 9:33
  SPEAKER_00: And what kind of caught my eye is, you know, OpenAI, they are reportedly filing IPO paperwork.
• 9:40
  SPEAKER_00: They actually supposedly did that in May.
• 9:43
  SPEAKER_00: Um, Anthropic filed their S1 just this past week.
• 9:47
  SPEAKER_00: More on that in a little bit.
• 9:48
  SPEAKER_00: And then both CEOs just happened to soften their most alarming predictions the very same week.
• 9:55
  SPEAKER_00: I don't know.
• 9:56
  SPEAKER_00: Seems a little interesting.
• 9:58
  SPEAKER_00: Fortune magazine called it a quote, coordinated industry-wide walkback.
• 10:03
  SPEAKER_00: Maybe that's what it is.
• 10:04
  SPEAKER_00: It could be possible that they both genuinely looked at the data and updated their views.
• 10:10
  SPEAKER_00: I mean, that's what good thinkers do, right?
• 10:12
  SPEAKER_00: But what does it all mean for product people anyway?
• 10:14
  SPEAKER_00: Well, look, the past couple of years, the AI industry has had every incentive to make AI sound as transformative as possible.
• 10:22
  SPEAKER_00: Big wild predictions attract investment, talent, attention.
• 10:26
  SPEAKER_00: And now some of those people, they have a different incentive.
• 10:29
  SPEAKER_00: They're gonna go public.
• 10:31
  SPEAKER_00: So now they're attracting institutional investors.
• 10:34
  SPEAKER_00: They need to tell a story that doesn't scare regulators.
• 10:37
  SPEAKER_00: And so the rhetoric rhetoric might be shifting a little bit.
• 10:41
  SPEAKER_00: As a product person, does it even matter?
• 10:44
  SPEAKER_00: I mean, I think you need to see through both versions.
• 10:46
  SPEAKER_00: The AI will replace everything uh narrative.
• 10:50
  SPEAKER_00: I maybe that wasn't fully grounded in near-term reality.
• 10:54
  SPEAKER_00: But I'd say also like the actually it's fine, you know, nothing's happening here, you know, this is just tech as usual.
• 11:01
  SPEAKER_00: I mean, that narrative's not very helpful either.
• 11:03
  SPEAKER_00: The truth is probably somewhere in the middle for now, but it's moving and it's moving fast.
• 11:09
  SPEAKER_00: And your job as a product person is to build for what's actually happening, not for what the loudest people are actually saying in either direction, by the way.
• 11:17
  SPEAKER_00: So maybe you tech executives out there that are considering cutting 75% of your staff because you don't need them.
• 11:24
  SPEAKER_00: You could just do everything with AI agents.
• 11:26
  SPEAKER_00: Well, in the words of Lee Corso, not so fast.
• 11:30
  SPEAKER_00: The fact that AI hasn't caused mass unemployment, maybe that's a useful signal.
• 11:36
  SPEAKER_00: I mean, it means maybe that we're in this phase where AI is augmenting work more than it is replacing it for product teams, could be a big window, right?
• 11:45
  SPEAKER_00: I mean, this might be the moment where you can build products that make people dramatically more effective.
• 11:50
  SPEAKER_00: Not assume that AI is already doing the job for them.
• 11:54
  SPEAKER_00: The users that will love your product right now might be the ones that want to move faster, not the ones that want to disappear, right?
• 12:02
  SPEAKER_00: So I think that's important for us to think about how we continue to build our products.
• 12:07
  SPEAKER_00: I also want to point out that we see layoffs in the tech world.
• 12:10
  SPEAKER_00: We're hearing about them every single week.
• 12:12
  SPEAKER_00: Um, but then I'm noticing those companies are rehiring fast, but they're rehiring for different roles and they're looking for what they're calling AI native people.
• 12:22
  SPEAKER_00: So, what does that actually mean?
• 12:24
  SPEAKER_00: I think it's important to not rest on our laurels here.
• 12:27
  SPEAKER_00: You have to keep learning, you have to keep experimenting, especially with a gentic AI.
• 12:32
  SPEAKER_00: And if you haven't done that, because maybe it's like, well, my company's not there yet.
• 12:37
  SPEAKER_00: So that's fine.
• 12:38
  SPEAKER_00: I don't need to, you know, start to go so deep in that.
• 12:41
  SPEAKER_00: Well, I would rethink that.
• 12:43
  SPEAKER_00: I think you don't want to be a part of the group of product people that's being phased out of companies and going through a big layoff.
• 12:50
  SPEAKER_00: I think even if your company's not going through that transformation right now, you personally need to start doing that transformation.
• 12:57
  SPEAKER_00: Altman might be right that a total apocalypse isn't happening on his original timeline, but I'd be careful about concluding that it's not coming in some form.
• 13:07
  SPEAKER_00: So for me, I think the the useful sort of question that product people can ask isn't will AI replace jobs?
• 13:15
  SPEAKER_00: It's what does my product need to do before, during, and after that transition for when that time actually comes.
• 13:22
  SPEAKER_00: And what do I personally need to do before, during, and after that transition?
• 13:28
  SPEAKER_00: Again, you don't want to get left behind.
• 13:30
  SPEAKER_00: So that's story two.
• 13:32
  SPEAKER_00: One more story to go.
• 13:36
  SPEAKER_00: All right, story number three, all about Anthropic.
• 13:39
  SPEAKER_00: Yes, on June 1st.
• 13:41
  SPEAKER_00: I know we already teased it.
• 13:42
  SPEAKER_00: We're still gonna run through it.
• 13:43
  SPEAKER_00: On June 1st, Anthropic filed confidential paperwork with the SEC to run an initial public offering, an IPO.
• 13:50
  SPEAKER_00: And of course, Anthropic being the company behind Claude, one of the most widely used AI models in the world.
• 13:56
  SPEAKER_00: The numbers are pretty staggering.
• 13:58
  SPEAKER_00: The recent valuation for Anthropic was$965 billion, almost a trillion dollars, or probably over that for the IPO.
• 14:07
  SPEAKER_00: And for perspective, they had$47 billion.
• 14:12
  SPEAKER_00: That's their that's a revenue run rate now.
• 14:14
  SPEAKER_00: Last year, their revenue was$10 billion annually.
• 14:18
  SPEAKER_00: So they've nearly 5x their revenue in just 12 months' time.
• 14:22
  SPEAKER_00: Now, I do think it's interesting, just like we talked about in story two the same week that they're filing this paperwork, their CEO walks back his predictions about AI wiping out jobs.
• 14:33
  SPEAKER_00: Again, interesting.
• 14:34
  SPEAKER_00: You can make of that what you will.
• 14:36
  SPEAKER_00: But for product people, what does this whole IPO story actually mean?
• 14:40
  SPEAKER_00: Well, I think it matters, right?
• 14:43
  SPEAKER_00: Like right now, Anthropic operates with a lot of flexibility.
• 14:47
  SPEAKER_00: They can make decisions about pricing, model capabilities, API access, safety constraints based on what they think is right.
• 14:54
  SPEAKER_00: They have investors, yes, sure, but they're private investors.
• 14:58
  SPEAKER_00: The minute that they go public, things do change.
• 15:01
  SPEAKER_00: Now they have quarterly earnings calls.
• 15:03
  SPEAKER_00: Now they have to go to Wall Street and talk to analysts who are asking why has revenue growth slowed?
• 15:08
  SPEAKER_00: They have shareholders that want to see a path to profitability that's probably faster than what their original plan was, like when they were a private company.
• 15:16
  SPEAKER_00: All of that, you know, what does it mean for product people?
• 15:19
  SPEAKER_00: Well, if you're a product person that's building on top of Claude, your companies using the API, your building features on top of Anthrapex models, I think it matters a whole lot.
• 15:29
  SPEAKER_00: I mean, first off, on pricing, public companies under revenue pressure, what do they do?
• 15:34
  SPEAKER_00: They raise prices or they restructure their tiers in a way that make the economics work better for them and worse for customers.
• 15:41
  SPEAKER_00: It's just what happens.
• 15:42
  SPEAKER_00: So if your product's margins depend on a certain cost per API call, you might want to pay attention to this.
• 15:49
  SPEAKER_00: How about prioritization?
• 15:50
  SPEAKER_00: I mean, right now, Anthropic can choose to invest heavily in capabilities that matter to developers and builders, even if those capabilities don't directly drive revenue.
• 16:00
  SPEAKER_00: A public anthropic, they have to make harder trade-off decisions.
• 16:04
  SPEAKER_00: They might work on features that enterprise customers will pay premium prices for.
• 16:10
  SPEAKER_00: Maybe that's what ends up getting prioritized over the features that indie builders and uh, you know, product makers actually love.
• 16:17
  SPEAKER_00: And then there's the whole competitive dynamic.
• 16:19
  SPEAKER_00: I mean, anthropic file ahead of open AI, but we know open AI, they're going to do that before too long as well.
• 16:26
  SPEAKER_00: We already heard the rumblings as well.
• 16:28
  SPEAKER_00: So now you have two of the most powerful AI companies in the world at a race, not for model capabilities, but for public market dominance.
• 16:36
  SPEAKER_00: And when companies are racing for market position and investor confidence at the same time, they make aggressive moves.
• 16:43
  SPEAKER_00: There could be pricing wars.
• 16:44
  SPEAKER_00: There might be feature acceleration, there might be partnerships that end up locking in enterprise customers.
• 16:50
  SPEAKER_00: It's going to be really interesting to watch from a product perspective because some of those moves could create real opportunities, but some could create real risks if you've built deep dependencies on one single platform.
• 17:03
  SPEAKER_00: My question to you is which platforms are you actually betting on?
• 17:07
  SPEAKER_00: And tell me right in the comments.
• 17:09
  SPEAKER_00: I mean, I'm curious to know if you think the IPO will end up impacting how much you rely on one single platform.
• 17:16
  SPEAKER_00: I mean, again, we know both OpenAI and Anthropic, they're on this IPO track.
• 17:20
  SPEAKER_00: Does it matter to you?
• 17:22
  SPEAKER_00: I mean, it's interesting, is like a year ago, I feel like this kind of talk was just sort of like it was a distant, futuristic kind of question, but now not so much.
• 17:31
  SPEAKER_00: This is the conversation we're having right now.
• 17:34
  SPEAKER_00: Um, anthropic, open AI, they're not just these, you know, research labs anymore.
• 17:39
  SPEAKER_00: They are both major enterprise software businesses.
• 17:42
  SPEAKER_00: And I think it does have the potential to change the nature of what it means for product people.
• 17:48
  SPEAKER_00: I mean, Claude, it's it's not just a cool AI model, it is a platform, and platforms that go public at this scale, they have to serve their largest paying customers first.
• 17:58
  SPEAKER_00: So if you're a small to mid-sized product team, I'm not saying these platforms won't be relevant to you or that you won't be using them.
• 18:05
  SPEAKER_00: I'm sure you'll still be relying on them.
• 18:06
  SPEAKER_00: Just understanding where you fit in that ecosystem, I think that matters.
• 18:18
  SPEAKER_00: Well, that's a wrap on episode one of Now Shipping.
• 18:22
  SPEAKER_00: I hope you had a little bit of fun.
• 18:24
  SPEAKER_00: I hope you learned some things, and I hope it was valuable.
• 18:27
  SPEAKER_00: If it was, please subscribe, tell a friend, spread the word, and look, if you have suggestions on how to make it even better, leave a comment below.
• 18:34
  SPEAKER_00: I promise you I will read every single one of your comments, and I will be right back here next week with three more AI stories that matter to you as a product person.
• 18:45
  SPEAKER_00: Now, once again, my name is Mike Belcito, and this was all brought to you by the team at Mind the Product.
• 18:52
  SPEAKER_00: This is now shipping.