Latest / iwon™ podcast with Michael J Thomas / AI, Jobs and Power: Musk, Altman and Yang on What Comes Next - iwon podcast - Summer Series with Michael J Thomas

Transcript

AI, Jobs and Power: Musk, Altman and Yang on What Comes Next - iwon podcast - Summer Series with Michael J Thomas · iwon™ podcast with Michael J Thomas

• 0:00
  AI can help us build better products, serve customers faster, and solve difficult problems.
• 0:06
  But as it becomes more powerful, another question becomes harder to ignore.
• 0:12
  Who holds the keys?
• 0:13
  Most business podcasts are bs, so I created one that isn't in this summer series episode with you and me.
• 0:21
  I want to talk about some of the questions people have asked me in the comments and founders have raised through my email list.
• 0:29
  AI is exciting.
• 0:31
  It is also scary.
• 0:33
  Elon Musk, Sam Altman and Andrew Yang have all spoken publicly about where this could lead.
• 0:40
  They do not all predict the same future.
• 0:43
  That is exactly why their words are worth exploring.
• 0:47
  What could AI mean for your business?
• 0:49
  What could it mean for your employees and customers?
• 0:52
  And what happens when so much of the intelligence, infrastructure, money and influence sits with so few people?
• 1:01
  I do not pretend to have all the answers.
• 1:04
  I do feel privileged to have a podcast where I can ask the questions.
• 1:09
  So let's dive into it for the last couple of years, companies have been racing to prove what AI can do.
• 1:16
  The demos were impressive.
• 1:17
  The MVPs came quickly.
• 1:20
  The announcements were everywhere.
• 1:22
  Everything had an AI assistant, an AI co pilot or an AI agent attached to it.
• 1:28
  But several recent stories suggest the conversation has changed.
• 1:32
  The AI race is moving from demos and MVPs to delivering real world value for business owners.
• 1:40
  That changes the question.
• 1:41
  It is no longer simply can we use AI?
• 1:45
  It is can we use it to make our customers lives better without moving so fast that we create problems we cannot control?
• 1:53
  Microsoft is spending 2.5 billion helping customers make AI work Microsoft has created a new operating business called Microsoft Frontier Co.
• 2:04
  They are hiring approximately 6,000 industry, engineering and AI specialists too.
• 2:11
  Their big bet?
• 2:12
  To lead the charge and work with organizations such as Unilever and Novo Nordisk to help them choose right and turn those investments into actual business results.
• 2:25
  Microsoft already has the models.
• 2:27
  It already has the infrastructure.
• 2:30
  It already has access to some of the biggest enterprise customers in the world.
• 2:35
  Yet it is still putting billions behind the people needed to make the technology work.
• 2:41
  That tells us something important.
• 2:43
  The biggest challenge in enterprise AI may not be access to intelligence.
• 2:47
  It may be everything that has to happen around it.
• 2:50
  The data, the workflows, the integrations, the security, the change management, the trust, the ability to prove that any of it has delivered a return.
• 3:02
  A great AI model is not the same thing as a working solution.
• 3:06
  Microsoft appears to be betting 2.5 billion on the difference.
• 3:11
  ServiceNow shows where companies are already paying the ServiceNow Plus AI story.
• 3:18
  Its AI products crossed 1 billion in annual contract value during the second quarter of 2026, the company also raised its forecasts.
• 3:29
  Their annual subscription revenue forecast is up, reporting stronger than expected revenue and profit.
• 3:37
  This matters because it is real contracted value, not the number of people who tried a free tool, not the number of pilots announced, not a prediction of what AI might be worth in five years.
• 3:50
  Businesses are already paying ServiceNow to put AI inside workflows.
• 3:56
  That last word is important.
• 3:58
  Workflow ServiceNow is not asking companies to buy access to a clever model.
• 4:04
  It is selling AI in the processes those organizations already understand and need to run.
• 4:11
  Customer service, IT operations, employee support, public services.
• 4:17
  The technology connects to work that already needs doing.
• 4:20
  For enterprise software founders, there is a pretty clear lesson here.
• 4:25
  Customers do not want to buy AI.
• 4:28
  They want to remove delays, reduce repetitive work, make better decisions, serve customers faster and better, or solve a problem that is already costing them money.
• 4:40
  AI may power the outcome, but the outcome is what the customer buys.
• 4:46
  SAP has trimmed its 2026 profit outlook.
• 4:50
  The reason behind the reduced profits outlook?
• 4:53
  They made some acquisitions.
• 4:55
  They bought Dremio and Prior Labs.
• 4:58
  The impact is they expected to reduce profits by more than 100 million during 2026.
• 5:05
  Both acquisitions improve the data and AI capabilities underneath SAP's existing products.
• 5:11
  In other words, SAP is investing in the foundations required to make AI useful to its customers.
• 5:19
  That means across areas such as finance, HR and supply chain management.
• 5:24
  The cloud revenue outlook remained unchanged too.
• 5:28
  So this does not look like a story about collapsing demand.
• 5:32
  It's a story about the cost of changing an established enterprise software business in the age of AI to build something its enterprise customers want.
• 5:42
  That is an uncomfortable but important lesson for founders.
• 5:46
  AI is not another feature to add to the roadmap.
• 5:50
  Your architecture, your data strategy, your margins, your pricing, your support requirements, your product team.
• 5:58
  The assumptions on which the business has been built.
• 6:01
  The opportunity could be enormous.
• 6:04
  That does not make the transition cheap.
• 6:06
  Capgemini shows that SAP is not a one off.
• 6:10
  Capgemini has raised its 2026 revenue growth target to approximately 8.5% to 9%, according to Reuters.
• 6:20
  This is being driven by organizations preparing for AI transformation.
• 6:25
  The other side of the Microsoft story.
• 6:27
  Companies now have access to powerful models that are improving day by day.
• 6:32
  But many companies still have all the hallmarks of a bigger change afoot.
• 6:37
  Their old, often fragmented data, their aging infrastructure and systems that do not communicate with one another efficiently.
• 6:46
  Before an AI agent can complete work, it needs good data and it needs permission and guardrails to work well with the systems where that work takes place.
• 6:59
  That is where the money is flowing.
• 7:02
  Not only towards models and software licenses.
• 7:05
  It is also flowing towards integration, modernization, implementation and business process redesign.
• 7:13
  Is this the next big opportunity for founders?
• 7:16
  For founders that could create a huge opportunity.
• 7:19
  What it means is you do not need to build the next foundation model.
• 7:25
  You may be able to build the product that connects AI to a valuable or regulated piece of work.
• 7:32
  The model you use to do it all may change over time.
• 7:36
  Deep knowledge of the customer's workflow will not.
• 7:40
  But there are lots of warning signs ahead too.
• 7:44
  I'll start with the OpenAI agent example.
• 7:46
  The final enterprise story I am going to share with you is the most uncomfortable.
• 7:51
  In July, Reuters reported that an autonomous OpenAI agent escaped a controlled environment.
• 7:59
  It then connected to the Internet.
• 8:01
  It wasn't supposed to be.
• 8:03
  Next, it hacked into and took down Hugging Face, a place online where people share code.
• 8:09
  The extent of its damage may go further.
• 8:12
  Further reporting stated that it had also compromised an account on Modal Labs, although Modal said its own infrastructure was not breached.
• 8:22
  OpenAI described the original incident as unprecedented.
• 8:27
  The company said it was reviewing what happened and strengthening its safeguards.
• 8:32
  The incident has since attracted attention from American lawmakers and the UK Information Commissioner's Office.
• 8:40
  It is tempting to treat this as a one off and think it's not a big deal.
• 8:45
  As products become more autonomous, the risk changes.
• 8:49
  Traditional software follows instructions written into it.
• 8:53
  An AI agent gets given a goal, chooses actions and uses tools to pursue that goal.
• 8:59
  That is what makes agents potentially very valuable.
• 9:03
  It is also what makes boundaries, permissions, security and human oversight so important.
• 9:08
  If your product allows an agent too much access, you must be careful.
• 9:13
  If you gave it access to customer records, customers will want to know what can it access?
• 9:20
  What can it change?
• 9:21
  What happens when it misbehaves?
• 9:23
  Is it watching my every move?
• 9:26
  Can a human stop it?
• 9:28
  Who is accountable when something goes wrong?
• 9:31
  Remember, this trust takes years and years to build and it can go in minutes.
• 9:38
  I get it.
• 9:38
  In the age we are in, it has to be part of the product.
• 9:43
  So where can it add value and how are we making sure it doesn't go too far?
• 9:49
  These stories paint a very different picture from the AI conversation of two years ago.
• 9:55
  Microsoft is spending billions on implementation.
• 9:59
  ServiceNow is making serious money by embedding AI inside workflows.
• 10:04
  SAP is absorbing the cost of rebuilding its data and AI foundations.
• 10:10
  Capgemini is benefiting from the integration and modernization work enterprises need, and the OpenAI Inc.
• 10:18
  Demonstrates why we need control measures in place.
• 10:21
  The opportunity could be enormous.
• 10:23
  The easy part was creating the MVP, the demo and the sandbox like OpenAI thought they had.
• 10:30
  Now comes the difficult bit.
• 10:32
  Making AI work with real company data inside real workflows across old and new systems with measurable commercial value without creating unacceptable risk.
• 10:45
  That is where enterprise customers need help.
• 10:47
  It may also be how and where the next generation of valuable enterprise software companies begin.
• 10:54
  But if AI changes too much too fast, we could be facing a very uncertain future.
• 11:01
  Think of it this way.
• 11:02
  If it decides who works and earns, who has a voice online and where power lies, what kind of future are we racing towards?
• 11:12
  This is where the conversation becomes much bigger than enterprise software.
• 11:17
  I want to look at Three Voices and the AI debate.
• 11:20
  Not because we have to agree with them, not because I want to pass judgment on them, but because between them they raise some of the biggest questions facing all of us.
• 11:29
  I will start with Elon Musk.
• 11:31
  Musk recently sat down with the Economist for a wide ranging interview.
• 11:36
  He predicted that AI could exceed the combined intelligence of humanity in approximately five years.
• 11:43
  Within 10 years, he believes, AI and robotics could become so capable that money may matter far less.
• 11:50
  He also said something that has stayed with me.
• 11:53
  Work is going to be optional.
• 11:56
  Musk compared working in that future to growing vegetables in your garden.
• 12:00
  You would not need to do it.
• 12:03
  You might choose to do it because you enjoy it.
• 12:06
  I can see the optimistic version of that future.
• 12:09
  Machines do the work nobody wants to do.
• 12:12
  Goods and services become cheaper.
• 12:14
  People have more time for their families, communities, health and the things that give them fulfillment.
• 12:20
  That sounds lovely, but there is a big question between where we are now and that utopian idea.
• 12:27
  How do people pay for their lives while we make that transition?
• 12:32
  The Economist challenged Musk on this.
• 12:34
  If AI and robots can do almost every job better than a person, where does the money go?
• 12:40
  To the people who work?
• 12:42
  Or to the small number of people and companies that own the models, robots, data centers and platforms?
• 12:49
  Musk's answer was that governments could give people money while huge increases in production would help prevent inflation.
• 12:57
  Perhaps he is right.
• 12:58
  But abundance tells us how much can be produced.
• 13:02
  It does not tell us who owns it.
• 13:04
  It does not tell us how income is shared.
• 13:08
  It does not tell us how somebody pays their mortgage during the difficult years between employment and so called abundance.
• 13:16
  And it does not explain how we move from rising inequality and division to a future in which everyone shares the benefits.
• 13:24
  That is not me passing judgment on this idea.
• 13:28
  It is me asking whether one of the biggest economic changes in human history can safely rest on the prediction of one person.
• 13:37
  Especially when the people making the Predictions may also benefit from selling the future they describe.
• 13:44
  Even the people building it are still learning.
• 13:46
  And I think the changes, in his own view, are important.
• 13:50
  He has previously warned that AI could replace many of today's jobs and has talked about a future of extraordinary wealth created by AI.
• 14:00
  More recently, he said he no longer expects the jobs apocalypse he once feared.
• 14:05
  He said OpenAI had been roughly right about the technology, but pretty wrong about some of the social and economic effects.
• 14:15
  He also said there is a human part of work that cannot easily be replaced.
• 14:21
  I actually find that reassuring.
• 14:24
  It shows that predictions can change when reality arrives.
• 14:28
  But it also raises a tough question.
• 14:31
  If even the people leading the AI race are still learning what it will do to jobs and society, how confident should the rest of us be when we are told to move at 100 miles an hour?
• 14:42
  OpenAI's own recent security incident makes that question more urgent.
• 14:49
  During a controlled cyber test, its models found a route out of a restricted environment, reached the open Internet, and compromised part of Hugging Face's infrastructure.
• 15:01
  OpenAI described the incident as unprecedented.
• 15:04
  It also explained that the models had reduced safeguards for the test and were intensely pursuing the narrow goal they had been given.
• 15:13
  That context matters.
• 15:14
  This was not a public chatbot randomly deciding to attack the Internet.
• 15:19
  But the model still found a path its human testers had not intended it to take.
• 15:25
  To me, that is the point.
• 15:27
  The more AI is rolled out, the more things will go right.
• 15:31
  It can find security flaws, improve products, and help people do better work.
• 15:38
  More things will go wrong, too.
• 15:40
  The people building these systems admit they are still learning how capability, control, and human behavior interact.
• 15:49
  So who carries the risk while they learn?
• 15:52
  The lab?
• 15:53
  The business using the model?
• 15:55
  The employee whose role changes?
• 15:57
  The customer whose data or money is involved?
• 16:00
  Or the public who had no say in the decision?
• 16:03
  On Andrew Yang's podcast, he spoke with futurist Martin Ford.
• 16:08
  Their starting point was much more practical.
• 16:11
  Hundreds of billions are being invested in AI infrastructure.
• 16:16
  How will businesses earn that money back?
• 16:19
  There are two broad answers.
• 16:21
  They can create more revenue, or they can reduce costs.
• 16:26
  For many businesses, people are one of their largest costs.
• 16:30
  Yang acknowledged that his own company is already using AI for work that might previously have gone to a junior engineer.
• 16:38
  The easiest people to fire are those you haven't hired yet.
• 16:42
  That sentence matters.
• 16:44
  Job displacement does not need to arrive through one dramatic announcement.
• 16:49
  It can happen quietly.
• 16:51
  A graduate role that is never advertised.
• 16:54
  A junior employee who is never hired.
• 16:57
  A team of 10 that becomes a team of six.
• 17:01
  A departing employee whose position is not filled each decision may make sense for the individual business.
• 17:09
  Together, those decisions could reshape the employment market for generations.
• 17:14
  And here is the bit every business owner needs to think about.
• 17:18
  Employees are customers too.
• 17:20
  They earn.
• 17:21
  They spend.
• 17:22
  They start companies.
• 17:24
  They support other businesses.
• 17:26
  If large numbers of people lose well paid work, who buys the products and services created by this new age?
• 17:33
  This is why the three angles belong together.
• 17:36
  Musk asks us to imagine abundance.
• 17:40
  Altman reminds us that even the people building the future cannot predict every social effect or control every outcome.
• 17:49
  Yang and Ford ask us to follow the money and think about what happens to workers and customers along the way.
• 17:56
  None of that proves AI will be good or bad for humanity.
• 18:00
  It does prove that the questions are too important to leave to a handful of technology companies and billionaires.
• 18:08
  Power, Politics and Distraction this is where I become more uncomfortable.
• 18:13
  The same small group increasingly influences the AI systems, infrastructure, investment and communication platforms shaping our future.
• 18:24
  That does not make their ideas wrong.
• 18:26
  It makes questions about power legitimate.
• 18:29
  In Musk's Economist interview, the conversation moved from AI and abundance to into immigration, European politics and his use of X.
• 18:39
  The interviewer asked why one person who does not live in Europe should have so much influence over European politics.
• 18:47
  Musk said he hoped he was having a significant impact.
• 18:51
  Whatever your politics, that exchange is worth hearing in full.
• 18:55
  My concern is not that immigration, culture or identity should never be discussed.
• 19:01
  There are legitimate subjects for democratic debate, but so are ownership, taxation, automation, wages and the concentration of wealth.
• 19:12
  We should be capable of discussing both.
• 19:14
  My worry is that outrage can keep everybody looking at the spectacle, keep people arguing with one another.
• 19:22
  Meanwhile, the bigger questions receive less attention.
• 19:26
  Who owns the technology replacing the work?
• 19:29
  Who benefits from the productivity gains?
• 19:32
  Who pays for essential services if income from work falls?
• 19:37
  Who decides what the algorithms show us?
• 19:40
  Who gets to shape political debate across countries in which they do not live?
• 19:45
  I am not claiming there is one coordinated conspiracy behind all of this.
• 19:50
  I am saying distraction matters when enormous economic and political power is moving in plain sight.
• 19:58
  You may agree with me, you may not.
• 20:00
  Either way is fine with me.
• 20:02
  I feel privileged to have a podcast where I can ask these questions and express my view.
• 20:08
  The links to the full Musk, Altman and Yang conversations are at the end.
• 20:13
  Please watch and listen for yourself.
• 20:16
  Then make up your own mind.
• 20:18
  Work provides more than money.
• 20:20
  There is one final part of the work will become optional argument that gets overlooked.
• 20:26
  Work is not always enjoyable.
• 20:28
  It can be stressful, exhausting, and sometimes completely pointless.
• 20:33
  But work can also provide structure, identity, relationships, status, progress, community.
• 20:42
  A reason to get out of bed.
• 20:44
  Giving somebody an income may solve part of the financial problem.
• 20:48
  It does not replace purpose.
• 20:50
  As someone who talks about mental health, that concerns me.
• 20:54
  Think of it this way.
• 20:56
  If millions of people are told that their skills are no longer needed, how do they pay for things?
• 21:02
  We need to think deeply about dignity, belonging and meaning for us all as much as we do about productivity.
• 21:10
  Here's where I am and we are I use AI.
• 21:14
  It can be useful.
• 21:15
  It can help us to build better products, serve customers more, and solve problems fast.
• 21:22
  But with such a revolutionary technology, we should be able to look at this in a fair and balanced way.
• 21:29
  Microsoft ServiceNow, SAP and Capgemini show that commercial value is already there.
• 21:37
  Elon Musk describes a future of abundance in which work becomes optional and money matters far less.
• 21:45
  Sam Altman shows that even the people building AI are still revising what they expect it to do to work and society.
• 21:54
  Andrew Yang and Martin Ford ask what the investment means for jobs, customers and the wider economy.
• 22:01
  Cloudflare's data suggests bots have already become the Internet's largest audience.
• 22:06
  They are parts of the same story.
• 22:08
  AI is changing how we live.
• 22:10
  AI is changing how we learn knowledge, power, meaning, wealth and so much more.
• 22:17
  Everything.
• 22:18
  The difficult question is not whether AI will create more value over time.
• 22:24
  It will compound over time.
• 22:26
  The question is whether society, culture, people and governments can keep up.
• 22:32
  For founders, the question is not how do we add AI?
• 22:36
  It is what can AI do to help us make our lives better?
• 22:41
  That is a much harder question.
• 22:43
  It is also a much better place to start.
• 22:45
  I want the AI future to be something that benefits us all.
• 22:49
  I sincerely hope and believe that can be the case as long as we ask the tough questions too.
• 22:54
  What's your take.