Latest / In The Loop / How to build an AI-native team & company: The six-principle playbook from Y-Combinator

Transcript

How to build an AI-native team & company: The six-principle playbook from Y-Combinator · In The Loop

• 0:00
  In this episode, I want to walk you through how to build a truly
• 0:04
  AI first team of business. This month, Y Combinator, one of
• 0:09
  the most famous start up accelerators and venture capital
• 0:12
  firms in the entire world who have been the place that Airbnb,
• 0:18
  Coinbase, Dropbox, Reddit, Stripe, basically most of the
• 0:22
  biggest businesses in the entire world have come from.
• 0:26
  And they share incredible insights every single quarter.
• 0:30
  And this month they shared a playbook for building what they
• 0:33
  see as an AI first company. And the contrast here is not
• 0:37
  just somebody or a team or a business that uses AI, but that
• 0:41
  is built on AI. They're the very heart of every
• 0:44
  new idea that is emerging in the world, based in San Francisco
• 0:48
  around the most innovative people and ideas you can
• 0:52
  possibly imagine. They showed 6 principles from
• 0:55
  what they're seeing with all the latest hot AI start-ups that
• 1:00
  they're working with. And I'm going to talk about a
• 1:02
  very publication of how this is actually playing out real time
• 1:05
  by Jack Dorsey, the ex founder of Twitter, who has a new
• 1:09
  company called Block. I'm going to explain the
• 1:12
  playbook. I'm going to talk about what
• 1:14
  that actually looks like when it's running properly and give
• 1:17
  you some ideas to bring to your next team or business meeting.
• 1:22
  This is in the loop with Jack Hot.
• 1:23
  I hope you enjoy the show. So let's start with a bit of
• 1:37
  background. So about 10 days ago, so about
• 1:41
  24th of April, Diana Hugh, a general partner at YC, he
• 1:46
  basically overseas their AI companies program published what
• 1:50
  was called the playbook for building an AI native company.
• 1:53
  And it blew up, of course, soon as it was shared.
• 1:57
  And the reason it spread so fast is that a, it's named something
• 2:00
  that a lot of people are watching happen without quite
• 2:03
  having the language for or the playbook for.
• 2:06
  And the fact that it covers team or company going through this AI
• 2:09
  journey. It was described as AI, as your
• 2:11
  company's operating system, open versus closed loop systems and
• 2:16
  companies and teams making a company or team fully queryable.
• 2:21
  And the rise of 1000 ex employee, team member and
• 2:24
  engineer. And essentially why this means
• 2:26
  that middle management disappears and why start-ups or
• 2:29
  smaller companies have a huge opportunity to win right now
• 2:32
  because of the natural challenges that big companies
• 2:35
  face. And what's quite interesting is
• 2:38
  a couple months earlier, Jack Dorsey, who has basically run a
• 2:41
  lot of these principles real time.
• 2:45
  So Block, Jack Dorsey's company, for anyone not familiar is, is a
• 2:49
  fintech company. So you might have seen Square,
• 2:51
  the payment provider systems when you go to buy a coffee or
• 2:54
  whatever. And it's basically a
• 2:56
  peer-to-peer money app used by about 57 million Americans and
• 3:00
  companies. And in February of 2026, he
• 3:03
  decided to cut 40% of the entire block workforce in a single
• 3:07
  letter, not because it was struggling.
• 3:09
  The business actually profit was up, you know, nearly 25% over
• 3:13
  the last quarter, but actually because the architecture of the
• 3:16
  company he believed no longer matched what the technology of
• 3:19
  LLMS now makes possible. And so when she published this
• 3:23
  playbook, you know, two months later, the block story was a
• 3:26
  really fantastic case study of this playing out in real time.
• 3:29
  And I'm not placing a value judgement or whether that is
• 3:32
  good or bad right now. I'm just going to talk about
• 3:34
  what an AI first company and team really looks like.
• 3:55
  So let's get into the playbook. So there's 8 principles and and
• 3:58
  principle number one is that AI is the operating system, not
• 4:01
  just a tool. So most companies right now are
• 4:04
  doing something quite similar. They're rolling out, you know,
• 4:06
  enterprise licences, putting AI assistance on the sidebar of
• 4:09
  every application. They've got internal prompt
• 4:12
  engineering workshops. That's not AI as an operating
• 4:15
  system that is making existing processes a little bit faster
• 4:18
  without actually changing anything.
• 4:20
  And Dorsey frame this as the turbocharger on a horse cart.
• 4:24
  The horse is still doing the work.
• 4:25
  You just made the horse a little bit less tired.
• 4:28
  And the distinction drawn in this principle here is that your
• 4:31
  company should not be able to function without AI in the same
• 4:34
  way it can't function without Microsoft Teams or e-mail or
• 4:37
  Slack. Not AI makes us work faster.
• 4:40
  AI is actually how everything gets done.
• 4:43
  And the test is quite simple. I think could your most
• 4:46
  important daily team workflows run for a week if you switched
• 4:50
  off all AI tools? If the answer is yes and the
• 4:53
  work would carry on more or less normal but slower, then you're
• 4:56
  just using AI as a tool and you're not running it as an
• 4:59
  operating system. I'll give you a couple of good
• 5:01
  examples. We've quite literally built
• 5:03
  internal tools for managing all systems that literally building
• 5:06
  them using public APIs from Twitter, for example.
• 5:09
  We have every process that is built around AI now and we
• 5:14
  haven't got enough people to suddenly run all of those
• 5:16
  anymore. We we would literally be ceasing
• 5:18
  to function as a team. For example, every task of
• 5:22
  scheduled task on one system feeds into the next system to
• 5:25
  run the next process. Every team takes those outputs
• 5:28
  and puts it into the next process.
• 5:29
  We we can't function anymore. Block shared a really
• 5:32
  interesting example. Jack Dorsey's company.
• 5:35
  Essentially they've got something called Manager Bot.
• 5:37
  It sits inside Square and monitors each merchants
• 5:40
  inventory levels, sales velocity, their staffing, and
• 5:43
  other signals are upcoming events and weather patterns.
• 5:46
  And it doesn't wait for a merchant to notice a problem or
• 5:48
  ask for help. It literally surfaces an action
• 5:51
  before the merchant even realizes there's a problem.
• 5:54
  You can't just switch that off on a Friday and start again on a
• 5:56
  on a Monday. It's not a feature, it's now a
• 5:58
  layer of their entire system. Principle number 2 is closed
• 6:03
  loops, not open loops. So an open loop is basically
• 6:06
  where a decision goes out, something happens, and the
• 6:08
  result never really comes back in.
• 6:10
  In AI, that is a pillar of success and effectiveness.
• 6:15
  You know, most companies are running almost entirely on open
• 6:18
  loops and they have no idea whether things are working or
• 6:20
  not, or if the agents are making the right decisions or they're
• 6:24
  not able to scale these systems because that closed loop doesn't
• 6:28
  exist. And I've talked about This is
• 6:30
  why, you know, MD file structures that we talked about
• 6:33
  in the last episode exist where an agent logs its decision and,
• 6:36
  and stuff like that. But that's just not scalable, as
• 6:38
  I've mentioned before, because these decisions need to be
• 6:41
  logged at scale and each agent needs to pick up from that
• 6:43
  decision. And most companies are basically
• 6:46
  running entirely on open leaps and have zero idea about what's
• 6:49
  working as a result. And that's how companies have
• 6:52
  always worked. They've relied on humans to be
• 6:54
  the closed leap to retain that bit of information about whether
• 6:57
  something was successful, put in a spreadsheet or a dashboard
• 7:00
  maybe, that a human may check every few months or weeks.
• 7:04
  A good example is you run a marketing campaign, you see the
• 7:06
  total conversion numbers. You do not know really which
• 7:10
  specific touch points drove that conversion.
• 7:12
  Some tools give you that insight.
• 7:14
  You don't have an idea of maybe the sequence of interactions,
• 7:17
  the variance of messaging that worked, and the information
• 7:20
  always exists but doesn't root back into how you design that
• 7:23
  next campaign in a systematic way.
• 7:26
  And a lot of the time, that's because you take insight, you
• 7:28
  take data and then you draw insights from it to shape your
• 7:31
  next things or your next ideas. But for agents, that doesn't
• 7:34
  work because an agent is not a human and you can't trust it in
• 7:37
  the same way. It's fallible.
• 7:39
  Whereas a closed loop system is a process where the outcome of
• 7:42
  every single decision roots back into a structured database
• 7:45
  really, so that an agent can pick up that decision and make
• 7:49
  subsequent decisions automatically.
• 7:51
  Again, you're scaling intelligence.
• 7:53
  A good example of this is what we're building with mindset,
• 7:56
  which is called Memex. Essentially, you plan work in a
• 8:00
  document. That document has all the why
• 8:02
  and the how this is going to be delivered, which creates a set
• 8:05
  of decisions to make. Humans and agents make those
• 8:08
  decisions, and then suddenly all of that is split into tasks.
• 8:11
  You then assign that to all the agents and people and they go
• 8:13
  off and deliver that work. Every single piece of task and
• 8:17
  item within that task that's delivered by a human or an agent
• 8:20
  is automatically updated into that task.
• 8:24
  So it's literally constantly scaling all the decisions that
• 8:26
  were made and as a result, every agent or person that picks up
• 8:30
  from that next, say, task or project or whatever it might be
• 8:33
  in the future has access to all decisions at scale.
• 8:36
  You don't need to provide a context.
• 8:38
  You don't need to worry if this is going to do the wrong thing
• 8:41
  because it has access to those decisions.
• 8:43
  As a result, you can suddenly run at scale AI systems that can
• 8:48
  take huge decisions and actions because it has such amazing
• 8:50
  context. That's a closed loop system.
• 8:53
  So as an action for you this week, pick a process and then
• 8:57
  close it. What's the most repetitive task?
• 9:00
  What's the AI process or human process and where is that
• 9:02
  getting locked? You could take an MD structure
• 9:05
  format like we talked about last week that can work.
• 9:08
  That has some problems at scale. You have what we call decision
• 9:11
  drift, where not every decision gets logged in the proper way.
• 9:14
  So the agent of the human misses key context and decisions.
• 9:18
  But it's a good start. If you have called Co work,
• 9:20
  describe the process, ask it to design what a closed loop system
• 9:23
  would look like, where it gets measured, where the output goes,
• 9:26
  where the next agent picks up from.
• 9:28
  So when you think about this, think about asking the agent to
• 9:31
  create closed loop systems, and then every agent or system or
• 9:35
  scheduled task that you create should then read from that
• 9:37
  place. So principle #3 is making your
• 9:40
  company fully queriable. So think about what happens in
• 9:43
  most companies when, say, someone new joins at a senior
• 9:46
  level. They spend months in meetings
• 9:49
  talking to people, getting chucked into things.
• 9:51
  So they get to grips with everything.
• 9:52
  They have to understand what decisions have been made.
• 9:54
  They have to figure out institutional knowledge.
• 9:57
  It's not indexed anywhere. It cannot be queried easily.
• 10:00
  They just have to ask people all the time.
• 10:03
  This is what Hugh calls an Inquiry Able Company, and it
• 10:06
  describes most organizations that existed before 2024.
• 10:09
  So what this means in practice, for example, is instead of
• 10:11
  emailing A colleague to ask why a decision was made six months
• 10:14
  ago or last week, you just open a search interface and ask it
• 10:18
  and you get all the information, the context, the emails,
• 10:20
  everything instantly. A good example of this is we
• 10:24
  have a system for managing every project and plan and code that
• 10:28
  was written at the moment. It was in an It was initially in
• 10:32
  an MD file structure that was very advanced, but we've now
• 10:35
  moved out to Memex. Every decision, every
• 10:37
  information, it's all stored in there.
• 10:39
  So it's instantly queryable from a product delivery perspective
• 10:43
  and team members in the product process can query and out.
• 10:47
  When a developer goes, OK, I want to deliver any project,
• 10:51
  they don't have to ask anyone about architecture or decisions
• 10:53
  or projects from a week, six weeks, a month ago.
• 10:56
  They literally just go to that system and start making it from
• 10:59
  there. Many companies have already, you
• 11:00
  know, started doing this. They've got AI note taking,
• 11:03
  Notion for say documentation, Slack instead of e-mail.
• 11:06
  But that's all very basic stuff. And This is why I think one of
• 11:10
  the most important spaces, and I'm not just saying this because
• 11:12
  this is what Memex does a lot of the time as a company that's
• 11:16
  I'll probably describe at the forefront of a lot of these new
• 11:19
  changes. We spot things that are really,
• 11:22
  really important quite early. And in the next 1218 months, the
• 11:26
  most important things is going to be a decision layer, A
• 11:28
  decision register, not a meeting summary that has loads of
• 11:32
  unstructured text and information, An actual organized
• 11:36
  database of every decision that was made, connected to each
• 11:39
  project and person that delivered it, and importantly,
• 11:42
  every agent. Because again, the the new
• 11:45
  paradigm here is that you no longer design processes and
• 11:49
  systems for human to human collaboration.
• 11:51
  It's all about agent to agent to human collaboration.
• 11:55
  Humans are a coordinator now, not just doers.
• 11:58
  So as an action, think about where decisions are getting
• 12:01
  logged. Think about how you can make
• 12:03
  everything queryable so you can always say to somebody who asks
• 12:06
  a question, have you asked X system?
• 12:11
  So then the culture internally should be 1 of initiative.
• 12:15
  Because once you remove those barriers, in my opinion, you
• 12:17
  also remove excuses. So you encourage and enforce
• 12:20
  people to take initiative and go and make decisions.
• 12:24
  So principle #4 is the idea of 1000 X team member, engineer,
• 12:28
  employee and what that means for how you know, hire.
• 12:32
  So you know, for software companies, we'll just take the
• 12:34
  engineer as the first example. Software companies for 30 years
• 12:38
  operated in a band of roughly 200,000 to $400,000 of revenue
• 12:43
  per employee. And that range is held
• 12:45
  throughout most of the package software era.
• 12:48
  So when you buy a perpetual license for 12 months all the
• 12:51
  way through SAS through to mobile, it was a bit of a
• 12:54
  structural ceiling on what a person plus a computer could
• 12:58
  actually just produce. But that ceiling is now well and
• 13:02
  truly gone. You know, Cursor, the AI coding
• 13:05
  tool, hit $2 billion of revenue and generates $6.1 million per
• 13:11
  employee. So that historical benchmark is,
• 13:14
  is is gone. It's 20 times the traditional
• 13:18
  revenue per employee ceiling that we used to have.
• 13:21
  And these aren't productivity improvements, they are a
• 13:23
  different level of output per person and function.
• 13:26
  And that thousand X team member employee engineer concept
• 13:30
  therefore actually means one person doing what used to take
• 13:33
  an entire team. I'll give you a good example.
• 13:36
  We for this product launch built an internal Twitter tool,
• 13:41
  created a research report that scanned over 20,000 public
• 13:44
  GitHub repos. So loads of different public
• 13:46
  repos for code to analyse the drift between the MD file
• 13:50
  structure and what's actually being written in code to bring
• 13:55
  to life what we see is this big problem of decisions getting
• 13:58
  made by agents but not updating your rule set for agents.
• 14:01
  Using the tools, we then create an entire strategy, a content
• 14:04
  strategy based on real Twitter conversations that were
• 14:08
  happening in vocabulary and language to make it super
• 14:10
  applicable to people and really understand the problem set.
• 14:14
  We then created and set up MCPS into Google, into Google
• 14:18
  Analytics and Google Ads and to run experiments on our
• 14:22
  proposition. So what is the best way of
• 14:24
  describing this product and problem set that resonates to
• 14:27
  people? And we did it all from just
• 14:29
  talking to AI and setting up MCP servers.
• 14:32
  And then we pull in automatic reporting from different places
• 14:35
  into reports and models, which create new insights, which
• 14:38
  creates a new document as an MD doc, which we send straight to
• 14:43
  our AI coding. So core code in in our MIDE
• 14:46
  which spins up a new landing page to test new hypotheses.
• 14:50
  All of this, all of this was done by literally probably the
• 14:54
  equivalent of 1 1/2 persons working part time.
• 14:58
  That would have been probably months of work.
• 15:01
  And we literally were doing that in literally over a bank holiday
• 15:04
  weekend. So the level of pace change and
• 15:07
  things you can achieve is insane.
• 15:09
  I think the hiring implication here is quite uncomfortable
• 15:13
  because most hiring today still uses some version of challenges
• 15:16
  or basic questions. If you take the engineers,
• 15:19
  you're given a challenge. So write this function, debug
• 15:22
  something, build this feature in a few hours, and then you test
• 15:25
  it. But in a world where agents do
• 15:27
  all of this, do the execution, you're screening for the wrong
• 15:30
  thing. The skill that matters now is AI
• 15:33
  management is being able to precisely describe what should
• 15:37
  be done for you. Can this person define what good
• 15:41
  enough looks like? Can they have the taste and
• 15:43
  judgement to make decisions? Do they have the initiative to
• 15:46
  get an idea and execute and move forward?
• 15:49
  Can they take the output and verify whether that was done
• 15:52
  properly and have the judgement to decide what to do next?
• 15:57
  These skills are hard to test for, but the interview doesn't
• 15:59
  have to be complicated. Instead of these challenges or
• 16:02
  basic questions, give people a, let's say, a product brief and
• 16:05
  ask them to write a specification.
• 16:07
  Not to build anything, but to define what it should do, what
• 16:10
  edge cases it needs to handle, the process that you're going to
• 16:13
  go through and then see what happens.
• 16:15
  Just because someone's a really good person in their job doesn't
• 16:18
  mean they're going to be a fantastic candidate in this new
• 16:21
  AI first world. And it's why education needs to
• 16:24
  rapidly change. The way we teach people to think
• 16:28
  about the world needs to change fast.
• 16:30
  So yeah, next time you go for hiring, think about how you can
• 16:34
  attract the right person and actually learn whether the right
• 16:38
  candidate through a different type of process principle #5 is
• 16:42
  token Max, not headcount Max. Now, again, I covered this in a
• 16:45
  recent episode. I, I have mixed feelings with
• 16:47
  this. I, I mentioned it a lot, so I'm
• 16:49
  not going to go into it, but it was a really popular episode.
• 16:51
  I recommend you go listen to it. But the principle that they were
• 16:54
  talking about is that most companies are still treating
• 16:57
  inference. So spending on tokens, which is
• 16:59
  what happens when you produce words through a language model,
• 17:03
  is essentially constrained. Like finance saying we don't
• 17:06
  want to spend too much here. We have to approve tickets to
• 17:09
  increase spend per person, which basically means that if you want
• 17:13
  to run many agents across many larger data sets and bigger
• 17:16
  tasks, you're going to burn through lots of tokens and spend
• 17:18
  a lot. And finance is constraining
• 17:20
  that. And the principle here from this
• 17:22
  book, which as I said, I land in a bit of a mixed feeling here,
• 17:25
  is just to burn as many tokens as humanly possible, not get
• 17:29
  more headcount. And if according to Y
• 17:32
  Combinator, if that doesn't make the CFOA little bit
• 17:34
  uncomfortable, they're probably not pushing it far enough.
• 17:37
  Now, I do think this encourages the wrong behaviours and you
• 17:40
  need to be teaching people how to be efficient with token spend
• 17:44
  and context usage when using AI. But if you're trying to get the
• 17:47
  team to really start adopting this, it can be a good place to
• 17:49
  start just to get people widely using it and then teach them how
• 17:53
  to be more efficient. So as an action, maybe whole the
• 17:56
  last three months of AI tool spend across all the different
• 17:59
  subscriptions, use core Co work to paste it alongside your
• 18:03
  headcount and ask it to calculate current spend per
• 18:06
  person per month on AI and then judge whether that's high
• 18:09
  enough. Take that into the next finance
• 18:11
  conversation and reframe it as maybe not increasing overhead of
• 18:15
  new people, but actually increasing spend.
• 18:17
  The goal really here is simple. It's not about cutting headcount
• 18:20
  here in my view, it's about getting people to stop asking
• 18:23
  permission to use AI and just to start using it.
• 18:27
  Principle number six is the idea of software factories.
• 18:30
  So most teams today still work in the same way as they did in
• 18:33
  2015. A ticket arrives, a person picks
• 18:36
  it up. So in your project management
• 18:37
  tool or in Jira or whatever it might be, an engineer or a team
• 18:41
  member picks it up. The write code or they do the
• 18:43
  task, they put it in review state to wait for somebody to
• 18:46
  look at it, and then it ships. AI is, you know, in a sidebar
• 18:50
  helping with certain parts of that, but the process is broadly
• 18:54
  unchanged. Now, what Hugh described is
• 18:57
  something very different. A-Team that does not write code
• 18:59
  at all, that does not perform the tasks themselves at all.
• 19:03
  Mindset's been like this for quite a long time.
• 19:05
  Almost no code is really written by humans anymore.
• 19:08
  We really care about decision making though, and planning and
• 19:12
  thinking about what the AI produces, because otherwise
• 19:14
  you're just vibe coding crap. So it's not about not caring
• 19:18
  about the input or the output, but it's trying to no longer
• 19:22
  constrain people to having to do the work themselves.
• 19:25
  Now, Dan Shapiro built a taxonomy basically of AI
• 19:29
  assistant coding, and the term comes from manufacturing.
• 19:32
  So a facility where robots work in complete darkness because
• 19:35
  they don't need to see. And a good example here is the
• 19:37
  code compiles in the dark and humans just show up to maintain
• 19:40
  the specification, verify the output and decide what comes
• 19:43
  next. So if people think their job is
• 19:45
  now writing the thing, coding the thing, this new thinking
• 19:50
  tells them that their job definition is about telling
• 19:53
  something what to do, making sure it makes the right
• 19:56
  decisions, and verifying the output.
• 19:59
  So precise specification is now a skill that's critical, and as
• 20:03
  a result, therefore you need specifications that are
• 20:05
  versioned and maintained like Coach used to be.
• 20:07
  If the spec changes, you should be able to see the differential.
• 20:10
  And again, this is exactly what Memex is solving for.
• 20:13
  If someone amends a test scenario, there should be a
• 20:15
  record of why. If the agent made a decision
• 20:18
  that impacts the plan, it should have recorded why.
• 20:20
  Principle 7. This is the intelligence layer
• 20:23
  replaces the organizational chart.
• 20:25
  So most people, when they hear middle management is going to
• 20:28
  disappear, they think it's about cutting costs.
• 20:30
  Well, it it just isn't. This principle is actually about
• 20:34
  the specific job of a middle manager, which if you watch
• 20:37
  carefully, is mostly rooting what a senior leader or leader
• 20:41
  of a company wants to the right person, or aggregating statuses
• 20:45
  from teams relaying contacts to, you know, people above them.
• 20:50
  Passing the decisions back down in the right way, making sure
• 20:53
  the left hand knows what the right hand is doing.
• 20:55
  Basically them having the right judgement of what to do or what
• 20:59
  to build when things go wrong. Well, that is now a much smaller
• 21:02
  fraction of the role than the title actually implies.
• 21:05
  Now what the intelligence layer replaces is the routing, not the
• 21:08
  judgement. When every person in the company
• 21:11
  can query the same set of information, see the same
• 21:14
  decisions, understand the same context without a human having
• 21:18
  to pass that information along well, therefore the layers that
• 21:22
  exist only to pass information and just not important and
• 21:26
  therefore the organizational chart now is no longer the same.
• 21:32
  It was previously a workaround for the fact that information
• 21:34
  couldn't really travel at scale. A good example here is now we
• 21:38
  have MCP service for all of our primary systems, for sales
• 21:41
  engineering, for product, for customer service, and every
• 21:45
  single one of those accessed by every leader in the company.
• 21:48
  You can get a scheduled daily report of everything that
• 21:51
  happened automatically. So therefore, why would I go and
• 21:54
  ask a middle management about what happened, about progress,
• 21:56
  about performance? I don't need that.
• 21:59
  As a result, the people who are now middle management are about
• 22:02
  deciding what the most important actions are to take, aligning
• 22:05
  the teams and their agents to take them, and making sure that
• 22:09
  thing happens on time. So as a result, in many ways the
• 22:13
  org chart flattens and I think leadership becomes about
• 22:16
  supporting people and getting them to think about the right
• 22:19
  ideas and the right decisions to make, not just passing
• 22:22
  information along the process. 7 principles are extremely
• 22:25
  concrete, really interesting by general partners who work with
• 22:29
  all of the most exciting AI start-ups in the world.
• 22:33
  And two months earlier, Jack Dorsey had run most of these at
• 22:36
  his company block, and actually cut 40% of his workforce in the
• 22:39
  process. So there are the seven
• 22:57
  principles and I think it's it's a good time to conclude and I'm
• 23:00
  going to conclude with a few different actions I think you
• 23:03
  should do. You don't need to implement them
• 23:05
  all at once, but I think you can start by a running three
• 23:09
  questions to test the team. What is the minimum number of
• 23:13
  people to keep this service running at full capacity?
• 23:16
  What is the minimum to stay compliant and secure and what is
• 23:20
  the minimum to grow and meet your commitments?
• 23:22
  The first two give you the floor and the third gives you the
• 23:26
  ceiling. For this stage, most leaders
• 23:28
  will know how many people work in their company or teams, but
• 23:31
  very few calculate the floor. This isn't supposed to be a
• 23:35
  redundancy exercise. It's supposed to be an audit of
• 23:38
  the structure moving forward and the number that comes back.
• 23:41
  I think will tell you more about how the organization runs today
• 23:45
  more than anything else you can do in a single week.
• 23:48
  Second, set a visible AI budget and just remove the approval
• 23:52
  friction. Pull what people spend at the
• 23:55
  moment per person per month on AI tools.
• 23:57
  Open clock Co work. Paste it in alongside the
• 23:59
  headcount and ask it to benchmark against what the
• 24:01
  leading companies are spending. Set a team budget higher than
• 24:05
  your current spend, but then work with them to make sure
• 24:07
  they're efficiently spending. Make sure that people just don't
• 24:11
  ask permission #3 close a loop on one of the most important
• 24:16
  repetitive tasks. Pick a task that someone on your
• 24:19
  team does all the time in the same way and use cloud Co work
• 24:23
  to design what a closed loop system looks like for that task.
• 24:26
  What gets measured, where the output goes, what changes next
• 24:28
  time, what the next agent process will read from.
• 24:32
  So is it going to be connected to that same decision layer?
• 24:35
  And then measure quality of outcomes from day one, not just,
• 24:38
  you know, how fast it runs and #4 run a software factory
• 24:42
  pattern on at least one internal process.
• 24:45
  So something that touches no customers and no live system.
• 24:48
  Just get a person to write the spec, One person writes the
• 24:51
  criteria for success. And get core code or core Co
• 24:54
  work or whatever it might be to just do the work.
• 24:58
  Review the specification when the output is wrong, not just
• 25:00
  the output itself. And then get it to the log its
• 25:03
  decisions in a structured way, whether it's an MD file
• 25:06
  structure, whether it's pushing it into a spreadsheet.
• 25:09
  So yeah, there are the four things I think you need to start
• 25:12
  now. This is a 6-7 week project, you
• 25:16
  can really transform your team and organization as a result.
• 25:19
  And really I think many of these principles, although quite scary
• 25:24
  for the future of employment quite frankly, and the fact that
• 25:27
  we really need to push for change for education and
• 25:30
  opportunities, for apprenticeships and pre
• 25:33
  employment training. But I think this is how
• 25:35
  organisations are going to run in the future.
• 25:38
  Anyway, I hope you find that interesting.
• 25:40
  Thank you for listening and I'll see you next week.