Google Has More Data Than Almost Anyone. So Why Is It Bidding $10 Million On Old Emails?

Google has more data Google has more data than anyone else in the than anyone else in the than anyone else in the world. So why world. So why world. So why would Google offer would Google offer would Google offer $10 million for $10 million for $10 million for old old old Spirit Spirit Spirit Airlines work emails? Because of what Airlines work emails? Because of what Airlines work emails? Because of what they think they think they think your your your job consists of. How much job consists of. How much job consists of. How much does the work you does the work you does the work you do cost? If a deal do cost? If a deal like the one with like the one with like the one with Spirit Airlines goes through, Spirit Airlines goes through, the company selling the company selling the company selling the records will get paid the records will get paid , and the broker will definitely , and the broker will definitely , and the broker will definitely make a commission on it make a commission on it make a commission on it . The buyer, . The buyer, . The buyer, Google, hopes Google, hopes Google, hopes to create an agent to create an agent to create an agent capability that it capability that it capability that it can resell. can resell. can resell. And what does the person whose work And what does the person whose work And what does the person whose work helped create all this helped create all this helped create all this get get get ? She does not get the ? She does not get the ? She does not get the right to vote. right to vote. It turns out that It turns out that companies are paying companies are paying companies are paying millions and millions of millions and millions of millions and millions of dollars for work dollars for work dollars for work records that records that records that have accumulated over have accumulated over have accumulated over years. The reason is that the years. The reason is that the prevailing theory among AI labs is that prevailing theory among AI labs is that if they if they if they get these work get these work get these work records— records— records— Slack messages, Slack messages, Slack messages, Zoom call transcripts, Zoom call transcripts, Zoom call transcripts, documents—they documents—they documents—they can teach AI to can teach AI to can teach AI to do what you and I do do what you and I do do what you and I do , to do , to do , to do intelligent intelligent intelligent work. At least work. At least work. At least that's the idea. I think there are that's the idea. I think there are that's the idea. I think there are reasons for reasons for reasons for skepticism. And I skepticism. And I skepticism. And I think that at least you think that at least you think that at least you and I should and I should and I should know about it so that we know about it so that we know about it so that we can understand can understand can understand the value of our work, the the value of our work, the the value of our work, the value that we value that we value that we put in, and the value of the put in, and the value of the put in, and the value of the data that we data that we data that we create. By the way, create. By the way, create. By the way, why is this important? Because why is this important? Because why is this important? Because when we when we when we started 2026, started 2026, started 2026, we were mostly we were mostly we were mostly talking about what I talking about what I talking about what I call call call verified verified verified work. If the code work. If the code work. If the code runs, it's runs, it's runs, it's good. If it good. If it good. If it starts and starts and starts and works, that's great, it's a works, that's great, it's a works, that's great, it's a verified domain, and the

verified domain, and the verified domain, and the job is done. It's job is done. It's job is done. It's like a light switch. like a light switch. like a light switch. It is either on It is either on It is either on or off. It works or off. It works or off. It works or it doesn't work. But or it doesn't work. But or it doesn't work. But now we are talking about now we are talking about now we are talking about less verified less verified less verified areas. How do you areas. How do you areas. How do you know if a know if a know if a document you've document you've document you've written is good? written is good? written is good? Can you Can you Can you prove it? Did that prove it? Did that prove it? Did that particular sentence particular sentence particular sentence make it good? make it good? make it good? Is this something you Is this something you Is this something you changed in V5 changed in V5 changed in V5 compared to V4, or V37 compared to V4, or V37 compared to V4, or V37 compared to V36? You know compared to V36? You know , I used to get to , I used to get to , I used to get to numbers like that when I wrote PRDs numbers like that when I wrote PRDs numbers like that when I wrote PRDs for Amazon. They really for Amazon. They really for Amazon. They really are that are that are that tall. How do you tall. How do you tall. How do you know it's know it's know it's good? What makes it good? What makes it good? What makes it good? What makes this good? What makes this good? What makes this truly truly truly intellectual intellectual intellectual work? This is something work? This is something work? This is something we need to we need to we need to understand better, because understand better, because understand better, because in reality, a lot of in reality, a lot of in reality, a lot of the value that we humans the value that we humans create is not really create is not really create is not really captured by the captured by the captured by the types of data that types of data that types of data that these these these labs are buying. Now labs are buying. Now labs are buying. Now think about your think about your think about your own work. What own work. What own work. What part of the records part of the records part of the records explains where exactly you explains where exactly you explains where exactly you created value? created value? created value? How many records How many records How many records capture the effort capture the effort capture the effort spent on finding spent on finding spent on finding the right information the right information , getting permission, , getting permission, , getting permission, scheduling a meeting, scheduling a meeting, scheduling a meeting, and then the and then the and then the next meeting?

next meeting? next meeting? Someone who buys this Someone who buys this Someone who buys this story has to decide story has to decide what exactly the what exactly the what exactly the machine should learn. And machine should learn. And machine should learn. And this decision will be this decision will be this decision will be important for your important for your important for your work and mine, and you and I work and mine, and you and I work and mine, and you and I may not may not may not have much have much have much influence over it. What is influence over it. What is influence over it. What is work that matters work that matters work that matters ? This is the question ? This is the question ? This is the question that underlies the that underlies the that underlies the entire market, entire market, entire market, important to you and important to you and important to you and me for the next me for the next me for the next decade. If you decade. If you decade. If you spent lunch in Slack spent lunch in Slack spent lunch in Slack correcting an correcting an invoice, was the invoice, was the invoice, was the work in those work in those work in those messages? Was messages? Was messages? Was she looking for a she looking for a she looking for a mistake? Is it about mistake? Is it about mistake? Is it about getting someone with the getting someone with the getting someone with the authority authority authority to fix it to finally to fix it to finally to fix it to finally do it? And what is do it? And what is do it? And what is that listing worth to that listing worth to that listing worth to the buyer if the agent the buyer if the agent the buyer if the agent can't see can't see can't see the difference? It's a fight the difference? It's a fight the difference? It's a fight for what we do and for what we do and for what we do and how what we do how what we do how what we do counts. counts. counts. The answer depends on The answer depends on The answer depends on what what what the machine learns, what the machine learns, what the machine learns, what the buyer will pay, and whether the buyer will pay, and whether the buyer will pay, and whether your employer thinks they your employer thinks they your employer thinks they can replace you, can replace you, can replace you, or whether they buy a or whether they buy a or whether they buy a convincing imitation of convincing imitation of convincing imitation of your work, your work, your work, leaving you leaving you leaving you to sort out everything the to sort out everything the to sort out everything the agent missed, agent missed, agent missed, which I've heard stories about too which I've heard stories about too which I've heard stories about too . Imagine . Imagine . Imagine an invoice that an invoice that an invoice that doesn't match doesn't match doesn't match an order, and someone an order, and someone an order, and someone needs to resolve it needs to resolve it needs to resolve it before the month closes. An before the month closes. An before the month closes. An invoice in system A is invoice in system A is invoice in system A is here, an order in here, an order in here, an order in system B is there, the system B is there, the system B is there, the supplier supplier supplier sends a credit sends a credit sends a credit somewhere, but somewhere, but somewhere, but the reviewer hasn't the reviewer hasn't the reviewer hasn't been put in a copy, and been put in a copy, and been put in a copy, and somewhere in all this somewhere in all this somewhere in all this mess there is a mess there is a mess there is a disagreement about whether a disagreement about whether a partial delivery is considered partial delivery is considered complete. This is a completely complete. This is a completely complete. This is a completely realistic realistic realistic scenario. If you've scenario. If you've scenario. If you've ever dealt with ever dealt with billing, it's billing, it's utter chaos. If utter chaos. If utter chaos. If you know this company you know this company , you might , you might , you might remember that remember that remember that the supplier the supplier the supplier would combine two would combine two would combine two shipments on shipments on shipments on one invoice, or one invoice, or one invoice, or that an exception approved that an exception approved that an exception approved last month would be last month would be an error this month. For example, an error this month. For example, if you are in the world of if you are in the world of if you are in the world of accounts and you know

accounts and you know accounts and you know a client, right? a client, right? a client, right? Business needs to Business needs to Business needs to sort all this out, and it sort all this out, and it sort all this out, and it relies on your relies on your relies on your tacit knowledge to tacit knowledge to tacit knowledge to figure it out. He figure it out. He figure it out. He relies on you to relies on you to relies on you to authorize authorize authorize any corrections, any corrections, any corrections, etc. The messages etc. The messages etc. The messages you create in you create in you create in the process of this the process of this the process of this settlement settlement settlement matter because matter because matter because they help they help they help it happen, but that's it happen, but that's it happen, but that's not the work itself, is not the work itself, is not the work itself, is it? They don't it? They don't it? They don't bring money to the bring money to the bring money to the cash register. The record of this cash register. The record of this cash register. The record of this work could also work could also work could also include some of the include some of the include some of the effort it would effort it would effort it would take for take for take for the organization to the organization to the organization to allow this work allow this work allow this work to happen, but to happen, but to happen, but again, it is not work. again, it is not work. again, it is not work. Maybe someone Maybe someone Maybe someone needs a statement of needs a statement of needs a statement of account status, but that's account status, but that's account status, but that's not work. Maybe not work. Maybe not work. Maybe the meeting happened the meeting happened the meeting happened because the two systems because the two systems because the two systems didn't agree, but didn't agree, but didn't agree, but that's not work either. that's not work either. that's not work either. Now, some of Now, some of Now, some of this coordination this coordination this coordination was necessary. Some was necessary. Some was necessary. Some of this was the price of of this was the price of of this was the price of working within working within working within that particular that particular that particular company. And if you've company. And if you've company. And if you've never had a beer never had a beer never had a beer after work, after work, after work, arguing about arguing about arguing about your company, you your company, you your company, you probably haven't experienced probably haven't experienced probably haven't experienced true true true intellectual intellectual intellectual labor. Let me be blunt: labor. Let me be blunt: labor. Let me be blunt: happy hours and happy hours and happy hours and complaints are how complaints are how complaints are how office workers office workers office workers discuss important discuss important discuss important work moments. So, what work moments. So, what work moments. So, what we're really we're really we're really talking about is the difference talking about is the difference talking about is the difference between work and between work and between work and imitation. How do you imitation. How do you imitation. How do you demonstrate demonstrate demonstrate accessibility, accessibility, accessibility, show engagement show engagement , make sure your , make sure your , make sure your contribution is visible? These are contribution is visible? These are contribution is visible? These are things we things we things we talk about in big talk about in big talk about in big companies for companies for companies for career advancement, but they're not the career advancement, but they're not the career advancement, but they're not the same as the job itself.

same as the job itself. same as the job itself. And the archives put up And the archives put up And the archives put up for sale record for sale record for sale record all of this. The message all of this. The message all of this. The message that prevents a costly that prevents a costly that prevents a costly mistake and the mistake and the reminder message will be reminder message will be side by side, and the agent side by side, and the agent side by side, and the agent will read both. will read both. will read both. Let's say it Let's say it took 30 took 30 messages and three messages and three messages and three meetings to resolve the issue. This sounds meetings to resolve the issue. This sounds meetings to resolve the issue. This sounds quite plausible quite plausible . The valuable intervention . The valuable intervention . The valuable intervention was someone was someone was someone noticing that credit had not been noticing that credit had not been applied correctly somewhere in applied correctly somewhere in that mess. If that mess. If that mess. If the buyer perceives the the buyer perceives the the buyer perceives the entire entire entire data sequence as an example of data sequence as an example of data sequence as an example of competent competent competent accounting, what should the accounting, what should the accounting, what should the agent learn? How agent learn? How agent learn? How does an agent know what does an agent know what does an agent know what really matters? How really matters? How really matters? How can an agent learn can an agent learn can an agent learn anything other than imitating the anything other than imitating the anything other than imitating the work? There is another work? There is another work? There is another problem. The correspondence problem. The correspondence problem. The correspondence may end with may end with may end with the account being approved the account being approved the account being approved and the ticket closed, and the ticket closed, and the ticket closed, which which which looks like a success to the agent, looks like a success to the agent, looks like a success to the agent, but perhaps 6 but perhaps 6 but perhaps 6 weeks later a duplicate payment will be discovered weeks later a duplicate payment will be discovered weeks later a duplicate payment will be discovered . And . And . And if you think I'm if you think I'm if you think I'm theorizing, you haven't theorizing, you haven't theorizing, you haven't worked much in worked much in worked much in companies where companies where companies where real chaos reigns. Every real chaos reigns. Every real chaos reigns. Every example I example I example I give is a somewhat give is a somewhat give is a somewhat anonymized anonymized anonymized situation that I have situation that I have situation that I have personally experienced. This is a personally experienced. This is a personally experienced. This is a case where we case where we case where we underestimate the extent underestimate the extent underestimate the extent to which the data to which the data to which the data reflects the reflects the reflects the real mess in the real mess in the real mess in the company. And I know that company. And I know that company. And I know that buyers hope buyers hope buyers hope to extract to extract to extract a lot of a lot of a lot of valuable things from the working archives. In the valuable things from the working archives. In the valuable things from the working archives. In the August auction, August auction, August auction, Google won with a Google won with a Google won with a bid of $10 million bid of $10 million bid of $10 million , beating out , beating out , beating out Mercor, which offered $ Mercor, which offered $ Mercor, which offered $ 7.5 million.

7.5 million. 7.5 million. A hearing is due to take place A hearing is due to take place A hearing is due to take place on the proposed on the proposed on the proposed sale, so the deal is sale, so the deal is sale, so the deal is not yet complete. It not yet complete. It not yet complete. It is scheduled for is scheduled for is scheduled for October 14. And then someone October 14. And then someone October 14. And then someone has to decide: what will we has to decide: what will we has to decide: what will we teach the agent using the teach the agent using the teach the agent using the example of a example of a example of a bankrupt bankrupt bankrupt airline? Mercor is an airline? Mercor is an interesting potential interesting potential interesting potential buyer because buyer because buyer because it seeks to acquire both it seeks to acquire both it seeks to acquire both the records and the the records and the the records and the environments where environments where environments where agents are trained. agents are trained. agents are trained. Along with the Along with the Along with the Spirit bid, Mercor announced plans in July Spirit bid, Mercor announced plans in July to acquire Deep Tune, which to acquire Deep Tune, which builds learning builds learning builds learning platforms. They platforms. They platforms. They combine the software that the combine the software that the agent runs with and agent runs with and checking whether checking whether checking whether it successfully completed it successfully completed it successfully completed the task. Remember the task. Remember how I said at the beginning of the video how I said at the beginning of the video how I said at the beginning of the video that it is difficult to that it is difficult to that it is difficult to prove the effectiveness of prove the effectiveness of prove the effectiveness of intellectual intellectual intellectual work? Mercor work? Mercor work? Mercor is trying is trying is trying to solve this, and it's a to solve this, and it's a to solve this, and it's a really complex really complex really complex problem. Mercor problem. Mercor problem. Mercor claims that their claims that their claims that their experts create experts create experts create tasks and tasks and tasks and success criteria, and they success criteria, and they success criteria, and they posted paid posted paid posted paid jobs to jobs to jobs to set up agents set up agents set up agents and see if and see if and see if they were able to get the they were able to get the they were able to get the job done. People are actually job done. People are actually job done. People are actually paid to paid to paid to determine what determine what determine what counts as counts as counts as work done. For example, work done. For example, work done. For example, regarding accounts: is it regarding accounts: is it regarding accounts: is it enough to simply enough to simply enough to simply close the ticket? This is an close the ticket? This is an close the ticket? This is an example of what example of what example of what Mercor needs to deal with. Mercor needs to deal with.

These decisions These decisions transform transform transform the interpretation of the work by a the interpretation of the work by a the interpretation of the work by a specific expert specific expert specific expert into a goal that the agent into a goal that the agent into a goal that the agent must achieve during must achieve during must achieve during training. They also training. They also training. They also affect the evidence affect the evidence affect the evidence that the supplier that the supplier that the supplier can then can then can then present to your present to your present to your employer. When employer. When employer. When a vendor says an a vendor says an a vendor says an agent can handle agent can handle agent can handle your role, someone your role, someone your role, someone somewhere in that chain somewhere in that chain somewhere in that chain has already decided what exactly has already decided what exactly has already decided what exactly needs to be done for needs to be done for needs to be done for your job to be your job to be your job to be considered complete considered complete . You have reason . You have reason . You have reason to be concerned about this. This to be concerned about this. This to be concerned about this. This contains the answer to contains the answer to contains the answer to what work is for what work is for what work is for all of us. So, let all of us. So, let all of us. So, let 's go back 's go back 's go back to Spirit for a minute. to Spirit for a minute. to Spirit for a minute. Their August Their August Their August registry contained 100 registry contained 100 registry contained 100 million million million emails and, emails and, emails and, believe it or not, half a believe it or not, half a believe it or not, half a billion billion billion Teams entries, as well as code and Teams entries, as well as code and Teams entries, as well as code and business documents. business documents. business documents. Now imagine how many Now imagine how many Now imagine how many working days working days working days are recorded there. People who are recorded there. People who are recorded there. People who understand understand understand things, exchange things, exchange things, exchange information, information, information, substitute for substitute for substitute for each other, argue each other, argue each other, argue over problems, over problems, over problems, finally solve them, finally solve them, finally solve them, take sick leave take sick leave take sick leave when they are not sick. when they are not sick. when they are not sick. The employer owns The employer owns The employer owns the systems where the systems where the systems where all these records have accumulated all these records have accumulated all these records have accumulated . And let's not . And let's not . And let's not forget that forget that forget that Spirit eventually Spirit eventually Spirit eventually went bankrupt. So it's went bankrupt. So it's went bankrupt. So it's not even clear whether not even clear whether not even clear whether they were successful in they were successful in they were successful in doing the job in the doing the job in the doing the job in the long long long run.

run. But with that said, But with that said, the more we think the more we think the more we think about this issue, the more about this issue, the more about this issue, the more important it becomes to understand: the important it becomes to understand: the important it becomes to understand: the data space data space data space is filled with is filled with is filled with claims about claims about claims about work that are work that are work that are not sufficiently not sufficiently not sufficiently substantiated. There are substantiated. There are substantiated. There are many many many data brokers. Here's one that's data brokers. Here's one that's data brokers. Here's one that's just starting just starting just starting out called out called out called Grepped or grepped.ai, which Grepped or grepped.ai, which Grepped or grepped.ai, which offers to package offers to package offers to package small business data small business data small business data for anywhere from $20,000 to for anywhere from $20,000 to for anywhere from $20,000 to $5 million and $5 million and $5 million and simply resell simply resell simply resell it to labs. They it to labs. They it to labs. They need need need messages, code, messages, code, messages, code, documents, business documents, business records, etc. These are records, etc. These are records, etc. These are systems in which people systems in which people systems in which people spend their spend their spend their working lives, and they working lives, and they working lives, and they essentially become a essentially become a essentially become a product that can be product that can be product that can be sold to another sold to another sold to another buyer. And when I buyer. And when I buyer. And when I think about it, I ask think about it, I ask think about it, I ask myself: If myself: If myself: If the labs think the labs think that this chaotic that this chaotic that this chaotic imitation of work -- the imitation of work -- the records of how a records of how a records of how a task was done at task was done at task was done at this company -- is worth this company -- is worth this company -- is worth that much money, then aren't that much money, then aren't that much money, then aren't they worth more they worth more they worth more to the company itself to the company itself to the company itself to create to create to create value in the value in the value in the future? future? Think about it this Think about it this way. Imagine you are way. Imagine you are way. Imagine you are a startup and you can a startup and you can a startup and you can sell your data for a sell your data for a sell your data for a million dollars million dollars million dollars tomorrow. You could tomorrow. You could tomorrow. You could do it. Or you do it. Or you do it. Or you might think, “Wow, my might think, “Wow, my might think, “Wow, my work records are worth a work records are worth a work records are worth a million dollars. I million dollars. I million dollars. I could use this could use this to earn more to earn more to earn more because I know my market, because I know my market, because I know my market, my customers, my my customers, my my customers, my experience, I have a experience, I have a experience, I have a ready-made product. "I could have ready-made product. "I could have ready-made product. "I could have gotten gotten gotten more than a million."

more than a million." more than a million." And that's what I And that's what I And that's what I keep coming back to. keep coming back to. keep coming back to. The people who have the The people who have the The people who have the incentive to sell are incentive to sell are incentive to sell are business owners who business owners who business owners who are on the verge of are on the verge of are on the verge of bankruptcy and can't bankruptcy and can't bankruptcy and can't find another find another find another way to use way to use way to use this data. This is the data of this data. This is the data of this data. This is the data of the last hope. the last hope. the last hope. Remember the case of Remember the case of Remember the case of Spirit Airlines. In other Spirit Airlines. In other Spirit Airlines. In other words, because of the way words, because of the way words, because of the way this system is set up this system is set up this system is set up , the data that , the data that , the data that agents receive about agents receive about agents receive about how work is being done how work is being done how work is being done mostly mostly mostly comes from comes from failing companies. Think about this. Think about this. As we try to As we try to As we try to figure all this out figure all this out figure all this out , I think about , I think about , I think about keeping an eye on the keeping an eye on the keeping an eye on the money. People who money. People who money. People who are trying are trying are trying to figure out how to get to figure out how to get to figure out how to get agents to do agents to do agents to do work with work with work with this data have this data have this data have a huge incentive a huge incentive a huge incentive to get clean to get clean to get clean results that they results that they results that they can then can then can then resell and resell and resell and say, "You know what, say, "You know what, say, "You know what, labs? We have an labs? We have an agent training environment agent training environment that is truly that is truly that is truly suitable for suitable for suitable for product managers, product managers, product managers, engineers, or engineers, or anyone else.” And then anyone else.” And then anyone else.” And then they can they can they can allow allow allow labs labs labs to resell it. But to resell it. But to resell it. But I think we I think we I think we should insist should insist should insist on this, because on this, because on this, because ultimately we ultimately we ultimately we are trying to are trying to are trying to understand how to take the understand how to take the understand how to take the generalized complex generalized complex generalized complex field of activity that field of activity that field of activity that we call we call we call intellectual intellectual intellectual labor and reduce it to a labor and reduce it to a labor and reduce it to a set of set of set of optimization optimization optimization problems for individual problems for individual problems for individual parts of it. So we're parts of it. So we're parts of it. So we're taking all this taking all this taking all this Spirit Airlines data chaos and Spirit Airlines data chaos and Spirit Airlines data chaos and saying it's going to go saying it's going to go saying it's going to go digital.

digital. digital. We will force agents We will force agents We will force agents to work with them. We will to work with them. We will to work with them. We will give them different give them different give them different tasks, and they tasks, and they tasks, and they will optimize will optimize will optimize different aspects of the work. different aspects of the work. different aspects of the work. Maybe someone Maybe someone Maybe someone will optimize will optimize will optimize the accounts, like in our the accounts, like in our the accounts, like in our example. Someone else example. Someone else example. Someone else will optimize, I don't will optimize, I don't will optimize, I don't know, the distribution of seats. know, the distribution of seats. know, the distribution of seats. This is an airline. This is an airline. This is an airline. But when you do But when you do But when you do this kind of point-by-point this kind of point-by-point this kind of point-by-point optimization, you optimization, you optimization, you forget about many of forget about many of forget about many of the nuances that the nuances that the nuances that make make make intelligent work intelligent work intelligent work work. And I work. And I work. And I think that's one of the think that's one of the think that's one of the reasons why AI is still reasons why AI is still reasons why AI is still so difficult to demonstrate a so difficult to demonstrate a so difficult to demonstrate a lasting impact on the lasting impact on the lasting impact on the labor market. Work labor market. Work is a very complex thing, and is a very complex thing, and is a very complex thing, and the deeper we the deeper we the deeper we delve into delve into delve into data problems, data problems, data problems, the more we realize the more we realize the more we realize that things are not so simple. that things are not so simple. that things are not so simple. By the way, employees are By the way, employees are By the way, employees are not sitting idly by either not sitting idly by either not sitting idly by either . Employees . Employees . Employees can challenge can challenge can challenge the terms of agreements. the terms of agreements. Spirit Airlines' flight attendant union Spirit Airlines' flight attendant union has raised has raised has raised concerns about concerns about concerns about confidential confidential employee information and employee information and what can be inferred from what can be inferred from what can be inferred from how those how those how those records are linked. The union records are linked. The union records are linked. The union also also also questioned whether questioned whether questioned whether excluding entire excluding entire excluding entire systems sufficiently protects systems sufficiently protects systems sufficiently protects sensitive sensitive sensitive information that could information that could information that could be copied to be copied to be copied to other places, such as other places, such as email.

email. email. After all, remember, After all, remember, After all, remember, work is chaos. You work is chaos. You work is chaos. You may be may be may be obligated to put obligated to put obligated to put something in a certain entry, something in a certain entry, something in a certain entry, but this is the real world but this is the real world but this is the real world and you never and you never and you never think it will be think it will be think it will be resolved, so you resolved, so you resolved, so you put it in the wrong place. This is a put it in the wrong place. This is a put it in the wrong place. This is a concrete example of something I concrete example of something I concrete example of something I think think think we will see even more we will see even more we will see even more often. Employees often. Employees often. Employees will challenge will challenge will challenge where this where this where this data goes and who has the right to data goes and who has the right to data goes and who has the right to do so. And I'll tell do so. And I'll tell do so. And I'll tell you from my own you from my own you from my own experience, it's not just experience, it's not just experience, it's not just high-profile examples like high-profile examples like Spirit Airlines and their Spirit Airlines and their Spirit Airlines and their union that are union that are union that are publicly highlighting publicly highlighting publicly highlighting these issues. A lot of what's these issues. A lot of what's these issues. A lot of what's happening happening is individuals is individuals is individuals who are skeptical who are skeptical who are skeptical about about about how how how their data is being used, and they their data is being used, and they their data is being used, and they will privately will privately will privately resist the resist the resist the process one way or another process one way or another . And this doesn't necessarily . And this doesn't necessarily . And this doesn't necessarily have to happen just have to happen just have to happen just through the sale of data. I've through the sale of data. I've through the sale of data. I've seen this seen this seen this happen when happen when happen when people privately people privately people privately resist the implementation of resist the implementation of AI tools AI tools because it's presented because it's presented because it's presented as—and we as—and we as—and we go back to the go back to the go back to the meaning of work—this meaning of work—this meaning of work—this tool will do tool will do tool will do the work for you. And they the work for you. And they the work for you. And they think: no, I think: no, I think: no, I don't want that. This is not what don't want that. This is not what don't want that. This is not what I am looking for. This is not what I am looking for. This is not what I am looking for. This is not what I signed up for. I I signed up for. I I signed up for. I will resist will resist will resist this. And in private this. And in private this. And in private conversations, I’ve seen conversations, I’ve seen conversations, I’ve seen leaders leaders leaders estimate that up to estimate that up to estimate that up to a third of their teams are a third of their teams are a third of their teams are passively and quietly passively and quietly passively and quietly sabotaging and sabotaging and sabotaging and resisting these resisting these resisting these AI tools. I AI tools. I AI tools. I see this as a see this as a see this as a misunderstanding of the essence of the misunderstanding of the essence of the misunderstanding of the essence of the work. We've been talking all this work. We've been talking all this work. We've been talking all this time about time about time about how complex how complex how complex work is, and how it's work is, and how it's work is, and how it's pretty absurd pretty absurd pretty absurd to imagine that the sheer

to imagine that the sheer to imagine that the sheer volume of work, half a volume of work, half a volume of work, half a billion records in billion records in billion records in Teams, can really Teams, can really Teams, can really describe its value. describe its value. We need a better and more We need a better and more productive productive productive conversation about what conversation about what conversation about what work means. And work means. And work means. And it has to start with it has to start with it has to start with you and me. It must you and me. It must you and me. It must begin with our begin with our begin with our understanding that understanding that understanding that the value of what we the value of what we the value of what we produce is not produce is not produce is not a function of any a function of any a function of any single letter. This is not single letter. This is not single letter. This is not a function of any a function of any a function of any single document. single document. single document. This is not a feature of any This is not a feature of any individual individual individual message in Teams. It's message in Teams. It's message in Teams. It's really our really our really our ability to achieve ability to achieve ability to achieve something that will move the something that will move the something that will move the business forward. And business forward. And business forward. And in reality, in business, it in reality, in business, it in reality, in business, it all comes down to either all comes down to either all comes down to either increasing profits increasing profits increasing profits or reducing or reducing or reducing costs. These are your two costs. These are your two costs. These are your two options. And you options. And you options. And you are doing one of are doing one of are doing one of them. And if you can them. And if you can them. And if you can clearly describe how you clearly describe how you clearly describe how you do it, you can do it, you can do it, you can use use any AI you any AI you any AI you want to want to want to get to the goal faster. And I see get to the goal faster. And I see get to the goal faster. And I see many many many success stories in this area. success stories in this area. But if you find yourself But if you find yourself in a situation where in a situation where in a situation where management only tells you, management only tells you, management only tells you, “we’ll implement “we’ll implement “we’ll implement AI and then lay AI and then lay AI and then lay you off,” you won’t have you off,” you won’t have you off,” you won’t have much motivation to much motivation to much motivation to do anything. This is not the do anything. This is not the do anything. This is not the best situation best situation best situation for you in the for you in the for you in the long long long run. By the way, run. By the way, run. By the way, if you are in this if you are in this if you are in this situation, feel free to situation, feel free to situation, feel free to share this video with share this video with share this video with your manager. I would be your manager. I would be your manager. I would be happy happy happy to chat with them.

to chat with them. to chat with them. I think they should I think they should I think they should talk to talk to talk to me too. Because I am very me too. Because I am very me too. Because I am very open with open with open with managers. I tell managers. I tell managers. I tell them: listen, this is not a them: listen, this is not a them: listen, this is not a situation where simply situation where simply situation where simply cutting people cutting people cutting people is going to be productive in the is going to be productive in the is going to be productive in the long long long run. run. run. Intellectual Intellectual Intellectual work is not so easy work is not so easy work is not so easy to automate. to automate. to automate. To think that you can “ To think that you can “ feed” an agent a feed” an agent a feed” an agent a bunch of messages from bunch of messages from bunch of messages from Slack—I believe it is possible, Slack—I believe it is possible, Slack—I believe it is possible, but I will say it bluntly: but I will say it bluntly: but I will say it bluntly: having a long experience having a long experience having a long experience of working with large of working with large of working with large language models, even language models, even language models, even before the advent of ChatGPT, you before the advent of ChatGPT, you before the advent of ChatGPT, you will get a bunch of will get a bunch of will get a bunch of agents who agents who agents who skillfully send skillfully send skillfully send reports that there is reports that there is reports that there is nothing to report. nothing to report. nothing to report. They They They will simply imitate what they will simply imitate what they will simply imitate what they see in Slack and Teams, and we will see in Slack and Teams, and we will see in Slack and Teams, and we will end up with a simulation of end up with a simulation of end up with a simulation of work that is work that is work that is incredibly entertaining, incredibly entertaining, incredibly entertaining, quite expensive, quite expensive, quite expensive, and of little and of little and of little value. And even value. And even value. And even if you believe in if you believe in if you believe in agents, and you know agents, and you know agents, and you know I do, if I do, if I do, if you watch this channel, you watch this channel, you watch this channel, you have to understand: you have to understand: you have to understand: the value of an agent is not in the value of an agent is not in the value of an agent is not in doing the work, but in doing the work, but in doing the work, but in identifying where in the identifying where in the identifying where in the business there is business there is business there is repetitive, fairly repetitive, fairly repetitive, fairly boring work that we boring work that we boring work that we can take away from people can take away from people can take away from people and replace it with a system and replace it with a system and replace it with a system that will probably not look like a that will probably not look like a that will probably not look like a bunch of bunch of bunch of automation, automation, automation, handing off tasks from handing off tasks from handing off tasks from person to person and person to person and person to person and meetings, but will look meetings, but will look meetings, but will look like an agent that reads like an agent that reads like an agent that reads data, processes it, and data, processes it, and data, processes it, and outputs the result in a outputs the result in a outputs the result in a certain format without certain format without certain format without Slack, without Slack, without Slack, without email, or any email, or any email, or any other human other human other human means of communication, means of communication, means of communication, because it is simply not because it is simply not because it is simply not necessary. So it necessary. So it necessary. So it seems to me that we are seems to me that we are seems to me that we are going a little wrong here going a little wrong here going a little wrong here on many levels.

on many levels. on many levels. I think buyers don't I think buyers don't I think buyers don't understand what the understand what the understand what the work really is and work really is and work really is and just buy everything in just buy everything in just buy everything in a row because they a row because they a row because they need agents to need agents to improve at it. improve at it. I think the vendors I think the vendors I think the vendors probably don't probably don't probably don't realize realize realize how valuable their how valuable their how valuable their data is and sell it as a data is and sell it as a data is and sell it as a last resort. And it last resort. And it last resort. And it seems to me that seems to me that seems to me that employees are being told, employees are being told, employees are being told, we are being told, that this we are being told, that this we are being told, that this means the end of the means the end of the means the end of the job, while we, first of all job, while we, first of all , don't really , don't really , don't really understand what's understand what's understand what's going on, and going on, and secondly, we don't secondly, we don't secondly, we don't realize realize realize how complex the how complex the how complex the problem is that we problem is that we problem is that we solve every day, and how solve every day, and how solve every day, and how difficult it is to train difficult it is to train difficult it is to train agents to do it. agents to do it. Yes, I truly believe in Yes, I truly believe in intelligence. I know that intelligence. I know that intelligence. I know that artificial intelligence is artificial intelligence is artificial intelligence is useful, it's becoming useful, it's becoming useful, it's becoming more useful, but more useful, but more useful, but it doesn't necessarily it doesn't necessarily it doesn't necessarily handle the kind of handle the kind of handle the kind of intellectual intellectual intellectual work that requires a work that requires a work that requires a high level of high level of high level of judgment. I don't see judgment. I don't see judgment. I don't see scaling when it comes to having a scaling when it comes to having a scaling when it comes to having a deep deep deep understanding of a understanding of a understanding of a specific company's context specific company's context specific company's context and how to evaluate and how to evaluate and how to evaluate issues for a issues for a issues for a specific customer specific customer specific customer and solve them how that and solve them how that and solve them how that looks in a specific looks in a specific looks in a specific software software software environment. This environment. This environment. This specificity is something that is specificity is something that is specificity is something that is not easily achieved not easily achieved not easily achieved with with with current current current scaling laws. Because scaling laws. Because scaling laws. Because the laws the laws the laws of scaling are of scaling are of scaling are general. They should general. They should general. They should work in all the work in all the work in all the different companies where different companies where different companies where you work. And that's why you work. And that's why you work. And that's why people joke that people joke that people joke that Claude's presentation Claude's presentation Claude's presentation can be recognized from can be recognized from can be recognized from afar. Is this a good afar. Is this a good afar. Is this a good presentation? Well, it's presentation? Well, it's presentation? Well, it's beautifully designed. They beautifully designed. They beautifully designed. They solved the solved the solved the layout problem. This is layout problem. This is layout problem. This is great. But this is great. But this is great. But this is Claude's design, right? You Claude's design, right? You Claude's design, right? You have to make an

have to make an have to make an effort to change the effort to change the effort to change the design spectrum so that design spectrum so that design spectrum so that it matches it matches it matches your style and your style and your style and looks the way you looks the way you looks the way you want. And then want. And then want. And then you need to choose the you need to choose the you need to choose the right right right message. And you also message. And you also message. And you also need to establish the need to establish the need to establish the right mindset. And once again right mindset. And once again right mindset. And once again , we'll , we'll , we'll come back to this: come back to this: come back to this: the main reason for the main reason for a presentation is to a presentation is to get people get people get people on the same page in a on the same page in a on the same page in a conversation. And if you haven't conversation. And if you haven't conversation. And if you haven't put thought into it, and put thought into it, and put thought into it, and if you haven't built the if you haven't built the if you haven't built the right story, right story, right story, everything else doesn't everything else doesn't everything else doesn't matter. It's just matter. It's just matter. It's just Claude giving Claude giving Claude giving a presentation and producing a presentation and producing a presentation and producing mediocre results mediocre results . And I worry that we're . And I worry that we're . And I worry that we're letting agents letting agents letting agents think it's work, think it's work, think it's work, even though it's not work. even though it's not work. even though it's not work. The real work The real work The real work is in the coordination. is in the coordination. is in the coordination. Speaking of Speaking of Speaking of conclusions, I think conclusions, I think conclusions, I think employees should employees should employees should have a strong way of have a strong way of have a strong way of explaining what explaining what explaining what counts as success here and counts as success here and counts as success here and explaining which entries explaining which entries explaining which entries should be excluded. should be excluded. should be excluded. I think people should be I think people should be I think people should be able to able to able to ask how ask how ask how confidential records confidential records confidential records are used in are used in are used in situations like this. And situations like this. And situations like this. And if an employer or if an employer or if an employer or data buyer data buyer data buyer needs additional needs additional needs additional expertise to expertise to expertise to turn those turn those turn those records into educational records into educational records into educational materials, that materials, that materials, that expertise should be expertise should be expertise should be recognized and recognized and recognized and paid for as paid for as paid for as high-value labor. And high-value labor. And high-value labor. And if I worked for a if I worked for a if I worked for a company company company considering one of considering one of considering one of these deals, I would want these deals, I would want these deals, I would want to understand the scope of the to understand the scope of the to understand the scope of the proposed proposed proposed use. I would use. I would use. I would also choose a few also choose a few also choose a few typical tasks and typical tasks and typical tasks and explain what the explain what the explain what the correct correct correct result depends on. And I would result depends on. And I would result depends on. And I would make sure it make sure it make sure it actually works, because actually works, because actually works, because otherwise you end up in a otherwise you end up in a otherwise you end up in a situation where the data situation where the data situation where the data depreciates over time, depreciates over time, depreciates over time, the agent doesn't do the agent doesn't do the agent doesn't do anything useful, anything useful, anything useful, the company sold the company sold the company sold something for nothing, and then something for nothing, and then something for nothing, and then it's "burned out" and it's "burned out" and it's "burned out" and can't sell anything anymore can't sell anything anymore can't sell anything anymore . So, all of this boils down to the fact . So, all of this boils down to the fact . So, all of this boils down to the fact that you have to

that you have to that you have to feel feel feel confident in changing confident in changing confident in changing people's attitudes towards people's attitudes towards people's attitudes towards your work. How much your work. How much your work. How much human effort human effort human effort is left when it is is left when it is is left when it is impossible to prove what impossible to prove what impossible to prove what happened, but you happened, but you happened, but you know the job know the job know the job was done. A was done. A huge amount huge amount of work isn't of work isn't of work isn't clearly captured in clearly captured in clearly captured in Slack messages or Slack messages or Slack messages or email threads, but email threads, but email threads, but it got it got it got done anyway. You have to done anyway. You have to done anyway. You have to feel the courage feel the courage feel the courage to state this and to state this and to state this and say, “The work say, “The work say, “The work I do is far more I do is far more I do is far more valuable than what the valuable than what the valuable than what the agent can learn agent can learn agent can learn from my messages, and from my messages, and from my messages, and I can prove it because I can prove it because I can prove it because you see me you see me you see me increasing revenue, increasing revenue, increasing revenue, reducing costs, or reducing costs, or reducing costs, or increasing the real increasing the real increasing the real value of the business.” And value of the business.” And value of the business.” And before selling before selling before selling traces of knowledge to an traces of knowledge to an traces of knowledge to an external buyer, external buyer, external buyer, consider whether consider whether consider whether it will truly improve it will truly improve it will truly improve our company's performance and value our company's performance and value our company's performance and value in the world. in the world. in the world. Usually not. Usually not. Usually not. Consider whether it Consider whether it Consider whether it contains contains contains personal data of personal data of personal data of individual employees. It individual employees. It individual employees. It often is. And often is. And often is. And think about whether it's think about whether it's think about whether it's worth it. Consider worth it. Consider whether it is even worth whether it is even worth whether it is even worth teaching an agent to teaching an agent to teaching an agent to do this. Because I'm not do this. Because I'm not do this. Because I'm not sure that's the case.

sure that's the case. sure that's the case. I'm not sure we're I'm not sure we're I'm not sure we're really helping really helping really helping agents become agents become agents become better. Not to mention better. Not to mention better. Not to mention help; help; help; It seems like all of this is It seems like all of this is just an attempt to get just an attempt to get just an attempt to get agents to imitate agents to imitate agents to imitate the work. And I'm all for the work. And I'm all for agents making good agents making good PowerPoint presentations. We have certainly PowerPoint presentations. We have certainly made progress on made progress on made progress on this. I fully this. I fully this. I fully support them support them support them working well with working well with working well with Excel spreadsheets. We have Excel spreadsheets. We have Excel spreadsheets. We have certainly made certainly made certainly made progress on this. But progress on this. But progress on this. But this only matters if this only matters if this only matters if the worker the worker the worker using the using the using the agent is able agent is able agent is able to understand what is to understand what is to understand what is happening and happening and happening and create create create real real real work with it. The fact that work with it. The fact that work with it. The fact that a presentation has been a presentation has been a presentation has been created is not yet created is not yet created is not yet real work. I think real work. I think many people think many people think many people think differently, thinking that differently, thinking that differently, thinking that you can just make a you can just make a you can just make a request—and that's request—and that's request—and that's real work. So real work. So real work. So if we want to if we want to if we want to claim claim claim value as companies value as companies value as companies and as people, we and as people, we and as people, we need to do a better job need to do a better job need to do a better job of explaining, "This is not of explaining, "This is not of explaining, "This is not real work." These are real work." These are real work." These are different things, and that is why different things, and that is why different things, and that is why the price of a the price of a the price of a data archive does not reflect the data archive does not reflect the data archive does not reflect the whole essence of the matter. whole essence of the matter. whole essence of the matter. Yes, someone can Yes, someone can Yes, someone can make money today by make money today by make money today by selling a description of your selling a description of your selling a description of your work before that work before that work before that description can explain description can explain description can explain your contribution. They your contribution. They your contribution. They may, they probably may, they probably may, they probably shouldn't, and they shouldn't, and they shouldn't, and they may do so may do so may do so before it becomes the before it becomes the before it becomes the basis of your basis of your basis of your decision. They can decision. They can decision. They can do this when you are do this when you are do this when you are already grown up, and would already grown up, and would already grown up, and would make a completely make a completely make a completely different decision now because you are a different decision now because you are a different decision now because you are a better employee than you better employee than you better employee than you were then. The people were then. The people were then. The people making all these making all these making all these decisions, both on the decisions, both on the decisions, both on the buyer's and buyer's and buyer's and seller's side, need a seller's side, need a seller's side, need a good answer to the good answer to the good answer to the question we question we question we started with: "What is started with: "What is started with: "What is work?" And your

work?" And your work?" And your leaders should leaders should leaders should understand this too. They should understand this too. They should understand this too. They should realize that just because realize that just because realize that just because certain data indicates certain data indicates certain data indicates traces of work in an traces of work in an traces of work in an organization, it does not organization, it does not organization, it does not mean that it is mean that it is mean that it is work. There are entire work. There are entire work. There are entire companies, like companies, like companies, like Glean, that advertise to their CEOs: Glean, that advertise to their CEOs: Glean, that advertise to their CEOs: "You'll know what's "You'll know what's "You'll know what's going on in your going on in your going on in your company because we, Glean, are company because we, Glean, are company because we, Glean, are like a transparent window to the like a transparent window to the like a transparent window to the entire company. You entire company. You entire company. You see all the projects see all the projects see all the projects at once, you can just at once, you can just at once, you can just type in a query and we'll type in a query and we'll type in a query and we'll find it in Slack, Teams, find it in Slack, Teams, find it in Slack, Teams, mail, or anywhere mail, or anywhere mail, or anywhere else, and you'll also else, and you'll also else, and you'll also see the codebases see the codebases ." On the one hand, it's ." On the one hand, it's ." On the one hand, it's probably more probably more probably more visibility than visibility than visibility than many have had before, so many have had before, so many have had before, so that's great. I'm not saying that's great. I'm not saying that's great. I'm not saying that these companies don't that these companies don't that these companies don't solve real solve real solve real problems, but on the other problems, but on the other problems, but on the other hand, they are only as hand, they are only as hand, they are only as effective as the effective as the effective as the data data data inside them. If inside them. If inside them. If the data within the data within the data within an organization is messy an organization is messy , contradictory, , contradictory, , contradictory, ambiguous, or ambiguous, or ambiguous, or incomplete, as is often the case incomplete, as is often the case incomplete, as is often the case , then you will , then you will , then you will only get only get only get results from querying results from querying results from querying that data—incomplete, that data—incomplete, that data—incomplete, messy, and messy, and messy, and not telling the whole not telling the whole not telling the whole story. And so you will story. And so you will story. And so you will form an impression form an impression form an impression of the teams based on of the teams based on of the teams based on that. The best that. The best that. The best work I've seen, work I've seen, work I've seen, when we talk about when we talk about when we talk about what work is and what work is and what work is and how to work with how to work with how to work with agents, is work agents, is work agents, is work that allows agents that allows agents that allows agents to collaborate with to collaborate with to collaborate with humans by humans by humans by structuring data in a way structuring data in a way structuring data in a way that agents can that agents can that agents can consume and consume and consume and use. I

use. I use. I know it sounds know it sounds know it sounds boring, and boring, and boring, and it probably is, but I've seen it probably is, but I've seen it probably is, but I've seen phenomenal work from phenomenal work from product and product and engineering teams where people engineering teams where people engineering teams where people invest heavily in invest heavily in invest heavily in high-quality markdown files high-quality markdown files high-quality markdown files that agents can that agents can that agents can read and people can read and people can use to use to use to query and create query and create query and create meaningful products meaningful products that are then passed to that are then passed to that are then passed to QA teams for QA teams for QA teams for evaluation. There are evaluation. There are evaluation. There are many positive many positive many positive points here. But it points here. But it points here. But it all starts with all starts with all starts with very boring things. It very boring things. It all starts with all starts with helping the agent helping the agent helping the agent understand the source of understand the source of understand the source of truth in the system, truth in the system, truth in the system, giving people more giving people more giving people more time to think so time to think so time to think so they can they can they can prepare prepare prepare thoughtful documents thoughtful documents thoughtful documents and assessments that will and assessments that will and assessments that will help agents help agents help agents do their do their do their jobs better. This is work. jobs better. This is work. jobs better. This is work. This is value. This is This is value. This is This is value. This is not at all what they not at all what they not at all what they are trying to sell are trying to sell are trying to sell in such situations. in such situations. in such situations. So I want us to So I want us to So I want us to rethink this because rethink this because rethink this because I want agents to do a I want agents to do a I want agents to do a better job. I better job. I better job. I want them to want them to work productively with people, and work productively with people, and for us, the people, to be more for us, the people, to be more for us, the people, to be more aware of the aware of the aware of the value of our work.

value of our work. value of our work. We're not all cast members of the We're not all cast members of the We're not all cast members of the famous sitcom " famous sitcom " The Office" who imitate The Office" who imitate The Office" who imitate work for laughs. work for laughs. work for laughs. We actually We actually We actually make a make a make a living doing this, and we deserve to be taken living doing this, and we deserve to be taken