Latest / Identity at the Center / #390 - Identity Management for Agentic AI with Tobin South

Transcript

#390 - Identity Management for Agentic AI with Tobin South · Identity at the Center

• 0:04
  This is identity at the center. Welcome to the Identity at the
• 0:11
  Center podcast. I'm Jeff, and that's Jim.
• 0:12
  Hey, Jim. OK, Jeff, how are you?
• 0:15
  Oh, not so bad yourself. Good.
• 0:17
  Guess what? I did something today that you
• 0:19
  could never do. I mean, there's so many things I
• 0:23
  can't do. Yeah, the first one is.
• 0:26
  It should be within your power. So I received my iPad today.
• 0:31
  OK. Couple hours ago and I have not
• 0:33
  opened the box yet. Well.
• 0:36
  I, I question all your priorities.
• 0:38
  If you get a box, you open it. That's like, it doesn't matter
• 0:40
  what it is. Could be an iPad, it could be,
• 0:43
  you know, I think a Chapstick. You open the box when it comes
• 0:46
  in. I don't know how you just let it
• 0:47
  sit there. You're an animal, dude.
• 0:49
  Yeah, especially brand new technology so but I'm excited
• 0:53
  after the call I'm planning on cutting into it and playing for
• 0:59
  hours have. You ever had an iPad before?
• 1:00
  Is this your first iPad? This is my first iPad.
• 1:03
  Wow. 2025 Congratulations. Yeah, and it's the latest and
• 1:09
  greatness, which is even more unlike me.
• 1:13
  Well, like, oh great, the iPhone 17 came out, it's time to get
• 1:16
  the 16. Well, yeah, I'm expecting to get
• 1:20
  text from you later, like, OK, how do I do this?
• 1:22
  How do I set this up? You know, 5:00, It's a cool
• 1:27
  little device. You know, with the exception of
• 1:30
  that stupid, stupid, stupid Apple commercial of that little
• 1:33
  girl saying, what's a computer like?
• 1:35
  Very a very punchable face on that girl, unfortunately.
• 1:38
  Wow. Cut.
• 1:41
  Cut that out of the broadcast. Just so annoying.
• 1:45
  Maybe that's just me being old and decrepit, but you know this
• 1:49
  little, I don't know what she must have been like, what 1314
• 1:52
  and she kind of has this hottie. What's a computer?
• 1:55
  You you know what a computer is. Boomer.
• 2:00
  Yeah, right. OK.
• 2:02
  What else we got? Yeah, hit me.
• 2:05
  No, I'm excited for conferences, but I did want to throw in one
• 2:08
  of the thing which is kind of talked about what the later note
• 2:12
  question is. I'm really excited to jump in on
• 2:14
  that. That is like now we're getting
• 2:16
  to gym territory, something where I can actually add value.
• 2:21
  So I'm excited about that. And so for those who don't know,
• 2:25
  who never stuck around till the end, we always end every show
• 2:28
  with a later note. And it's just the idea that kind
• 2:32
  of humanizes Jeff and I and our guests.
• 2:34
  And we picked topic that's not identity related and we talked
• 2:39
  about it for a couple of minutes.
• 2:41
  I think if you're not interested in that, go ahead and drop off.
• 2:45
  Can't wait to hear it. Feel free to Fast forward toward
• 2:47
  the end. Well, I mean, don't drop off
• 2:50
  hit, you know, just turn the volume down and go somewhere
• 2:52
  else, do something. But I call that like and.
• 2:54
  Subscribe right? Yeah, that's like, you know,
• 2:57
  that's a perfect opportunity to hit the like and subscribe
• 2:59
  button. That's like meantime, like I
• 3:00
  like to have conversations about a whole bunch of different
• 3:02
  things and we usually try to pick something that's kind of
• 3:04
  fun and, you know, kind of joke around a little bit and maybe
• 3:06
  tease each other or find out something new, whatever.
• 3:09
  Maybe I could say right now I'm not going to be able to add any
• 3:12
  value to this later note. So I'm going to be relying on
• 3:14
  both you and our guests. Exactly, exactly.
• 3:17
  I know it's not your in your wheelhouse, but I was going to
• 3:20
  bring up also, you know, by the time this episode drops, it'll
• 3:23
  be probably the week of Gardner. So we're already starting to
• 3:28
  think about the 2026 calendar. You know, in the US we've got
• 3:34
  like Denniverse, in Europe, we've got EIC.
• 3:37
  Now by saying in US and in Europe doesn't mean you have to
• 3:40
  live there to only go to those conferences.
• 3:42
  I actually highly recommend folks from Europe go to
• 3:45
  Identiverse and vice versa. These folks from North America,
• 3:50
  US go to EIC usually learn so much, get so much new
• 3:55
  perspective. Yeah, We had a real good time.
• 3:58
  It was our first time going to Berlin earlier this year, 2025
• 4:01
  as we record this and we know we're so excited to do it again.
• 4:05
  So we've kind of partnered up with Cooper, Dracole and Ian,
• 4:08
  the EIC folks. So we've got discount code.
• 4:10
  So look, I get it. Going across the pond, you know,
• 4:12
  at least for the US is sometimes pretty expensive.
• 4:14
  So we've got codes on our website.
• 4:16
  So if you go to the IDAC podcast.com, just Scroll down,
• 4:19
  all of our active discounts are there, but you'll see one for
• 4:22
  EIC. So we've got 25% off ID, AC25,
• 4:26
  MKO. It's been a while since we read
• 4:28
  like a discount code for a while there.
• 4:29
  We had like 8 or 9 of them going once it was like, all right,
• 4:32
  we'll do an entire episode just reading codes.
• 4:34
  So now we just direct people, go to the website, codes are there.
• 4:37
  We're looking forward to it. You and I are going to be there,
• 4:39
  Jim. And yeah, trying to, you know,
• 4:43
  recreate the magic of 2025 there.
• 4:46
  Yeah, it was fantastic. I mean, even if like obviously
• 4:50
  it was a great conference, but it was also great being in
• 4:53
  Berlin and going on like we went on that river dinner cruise the
• 4:57
  one night and got to kind of have the canal ride.
• 5:01
  I thought it was just so cool. It's like a memories for a
• 5:05
  lifetime. It was fun.
• 5:08
  I think I sat there with Henrique and a couple other
• 5:10
  folks just kind of chatting. I, I didn't actually go outside
• 5:13
  the entire course. We were just sitting at a table
• 5:16
  just chatting about, you know, just that the other thing.
• 5:18
  So it was a fun time. It was good times.
• 5:20
  Looking forward to it. And I'm sure you'll, you'll
• 5:22
  weave in some, you know, some travel, you know, to other
• 5:24
  places. I have not yet figured out what
• 5:26
  I'm going to do yet, you know, outside of the trip to Berlin.
• 5:28
  So. Well, you mentioned how it can
• 5:30
  be expensive for people from North America to go over to
• 5:33
  Europe and vice versa. I'm sure the biggest key I can
• 5:37
  say is book way ahead of time. I mean, you know, I got my round
• 5:42
  trip to Europe last year for about the price of a last minute
• 5:47
  booking domestic flight. So it can be done.
• 5:51
  Just book, you know, six to four months in advance if you can.
• 5:55
  Yeah, or try to get your company to pay for it.
• 5:57
  That's the way to do it. Well, yeah, or both.
• 6:01
  Sure, Yeah, whatever you can get away with.
• 6:05
  All right, so let's get to our topic today.
• 6:07
  You've probably read the title already.
• 6:09
  It's identity management for agentic AI.
• 6:12
  We've got with us today Tobin S He's the Co chair of Open ID
• 6:16
  Foundation's AI Identity Management community group.
• 6:19
  So welcome to the show, Tobin. It is great to be here, thanks
• 6:23
  for having me. Thanks so much for taking the
• 6:25
  time. I feel like it's kind of like a
• 6:27
  mouthful identity management for agentic AI, you know, dot, dot,
• 6:31
  dot, and there's so much probably more behind it.
• 6:33
  Let's start a little bit though, with kind of your background.
• 6:35
  So how did you get into this space of AI and identity?
• 6:40
  Take us through your origin story.
• 6:42
  Yeah, I, I do find myself nowadays swimming in a sea of
• 6:45
  acronyms and jargon that is kind of terrifying to confront.
• 6:49
  I I didn't intend to end up here.
• 6:52
  I have always been someone who's really interested in in AI and
• 6:56
  that was the scope of what I was really interested to work on
• 6:59
  when I joined MIT for my PhD. And kind of the broad question
• 7:04
  was how do we build more robust security and safeguards on AI?
• 7:08
  And this is before ChatGPT, before it kind of aid the world.
• 7:12
  And over the course of my PHDAI got better and better chat bots
• 7:18
  became a thing. And this idea of AI agents
• 7:21
  started popping up where these crazy, wacky ChatGPT language
• 7:26
  models were going to go and do things that interact with
• 7:29
  services. And it just become became
• 7:33
  remarkably apparent that access controls and identity was going
• 7:36
  to matter more and more as part of that.
• 7:39
  And so I, I started shifting my research towards that direction.
• 7:42
  And ever since then, that has been everything that has
• 7:46
  consumed my mind for the last year or so is identity access
• 7:50
  management for agents, the future of robust consumer
• 7:53
  marketplaces for agents, as well as the kind of enterprise
• 7:56
  questions around how we deploy agents.
• 8:00
  Did your definition of AI change the first time you kind of
• 8:04
  became cognizant of things like a large language model and you
• 8:07
  know, specifically like the generative AI component?
• 8:10
  Because I feel like it was AI before it was really like
• 8:13
  machine learning and sort of pattern matching, right, and
• 8:15
  things like that. And then I saw what opened up
• 8:17
  and I was like, Oh my gosh, that's AII want that.
• 8:21
  I'm curious if you felt the same or felt differently.
• 8:24
  Even for folks who had seen language models before, small
• 8:28
  language models that would, you know, generate a couple tokens,
• 8:31
  generate a couple words, give you a poem.
• 8:33
  I do think the ChatGPT moment still changed a lot of minds.
• 8:37
  And even beyond there once tool calling anything when when
• 8:41
  language models could go and and write a Jason that could then
• 8:45
  query an API. I think that really changed the
• 8:48
  minds of a lot of people very quickly in space.
• 8:52
  Yeah. Now we've got, you know, things
• 8:53
  like vibe coding and all kinds of different things that people
• 8:55
  are doing. Tell me what about your role
• 8:57
  with Open ID? So tell me about the AI Identity
• 9:01
  Management Community group, which is not a working group.
• 9:04
  I know there's a very clear, you know, differentiation, but tell
• 9:06
  us about that role that you're serving on there.
• 9:08
  Yeah. So in, in January or February,
• 9:11
  we put out this paper on authenticated delegation and
• 9:15
  authorization and delegated authority is this topic that
• 9:20
  people have often cared about where, you know, maybe I want to
• 9:23
  delegate authority to, to an agent to, to like a human, human
• 9:26
  agent to a sub process that's constrained.
• 9:30
  But with AI agents, this kind of blows up.
• 9:32
  We're going to have these chat bots that are non deterministic
• 9:36
  that can go out and do things for us.
• 9:38
  So we hosted a series of workshop at workshops at
• 9:42
  Stanford to get people talking about this and figure out what
• 9:45
  the future is. And it was clear there was just
• 9:47
  a a need for space for a community to talk more and more
• 9:51
  about this. And so the Open ID Foundation
• 9:53
  put together a community group which I now Co chair, to host
• 9:56
  these discussions and figure out what we need to build to make a
• 10:00
  safe future with agents. So, Tobin, I wanted to ask you
• 10:04
  early. I wanted to go back out.
• 10:07
  It's happening too quickly for me to get my follow up question.
• 10:09
  But like you talked about that ChatGPT moment and I felt like
• 10:13
  it was when GPT 4 kind of came out.
• 10:17
  Like that seemed to be where everybody's like eyes started to
• 10:21
  really open and you'd hear about it and you're like, oh, I've got
• 10:23
  to go out and check this out. It's like, whoa, can't believe a
• 10:27
  computer's actually doing this. I thought like it took an
• 10:30
  instruction and it kind of like figured out how to spit out an
• 10:33
  answer. But this thing is much more so,
• 10:36
  Is that the moment for you? I think it was there's this idea
• 10:41
  in AI of instruction training. So language models were just
• 10:45
  trained on spitting out text, not doing what you want them to
• 10:50
  do. You know, anyone who was a
• 10:52
  machine learning practitioner was constantly fighting against
• 10:55
  the idea that AI does what the data tells it, not what you want
• 10:58
  it to do. That was the whole art form of
• 11:00
  being a machine learning engineer.
• 11:02
  And instruction following changed that on its head.
• 11:05
  It created AIS that could, you could give it a natural language
• 11:10
  instruction and it would go and do a thing.
• 11:12
  And kind of from that moment onwards, we've seen this slow
• 11:15
  trend towards it being better and better at doing what you
• 11:18
  want it to do, to the point where today you can tell it to
• 11:24
  do a thing and it might do it wrong, it might not use the
• 11:27
  right tools. It gets confused along the way.
• 11:30
  And that's why a lot of people kind of have this skepticism
• 11:32
  about AI agents and they haven't had their ChatGPT moment yet
• 11:35
  maybe. But to me, it seems like a
• 11:38
  really smooth trend where we're just getting better and better
• 11:41
  at having an AI do what you want it to do, no matter what that
• 11:44
  magically entails. Yeah, I'm actually working on a
• 11:48
  project that I guess I'm getting into all the details, right?
• 11:52
  It's like a chat agent or I feel like the term assistant is
• 11:58
  better, but let's say it's an agent that can have a
• 12:01
  conversation with a person to figure out what training makes
• 12:05
  sense for them. And it understands all the rules
• 12:07
  of like, hey, you got to do these things, these are some
• 12:09
  required trainings, things like that.
• 12:13
  And it kind of where I really want to take it is it can get to
• 12:16
  the point where it says, OK, Tobin, these are some courses
• 12:19
  that you should take. Would you like me to go onto
• 12:22
  your calendar and set a reminder or would you like me to go and
• 12:26
  register you for this course? And I, I think that's where the
• 12:32
  difference between a user and an agent identity really it goes.
• 12:38
  So it might let me ask the question of simply is like,
• 12:40
  what's the difference between a user and an agent?
• 12:43
  But kind of understanding that I already realized what the
• 12:45
  difference is, I guess I'm wondering from an identity
• 12:49
  perspective. Yeah.
• 12:51
  So I, I think to start off with the term agent is terrible and
• 12:55
  has made all of our lives, you know, much harder.
• 12:58
  At least it found this podcast interesting.
• 13:01
  You know, it's, it's, it does well on search engine
• 13:04
  optimization and it makes for good podcast titles.
• 13:07
  But in practice, when you're building systems, it's kind of
• 13:09
  confusing. And so AI can mean many
• 13:13
  different things. Sometimes it is a human being
• 13:17
  talking to an AI, more of an assistant context and that AI is
• 13:21
  just helping them do something. And this is not that much
• 13:25
  different from a a user interacting with the surface,
• 13:28
  subtly different but but not massively.
• 13:30
  Then there is I use an AI and ask it to go and do stuff and I
• 13:37
  gesture towards what that stuff looks like, but I don't really
• 13:40
  tell it. This is more like vibe coding.
• 13:42
  You know, my my clawed code has access to MCP servers for GitHub
• 13:46
  and to Super Bass, and it can spin up databases and use them
• 13:50
  as required. This is no longer me really
• 13:53
  being in the loop. It's more of a delegated
• 13:55
  authority use case. And then there is AI, which kind
• 13:59
  of acts like a service account. It's a theme that runs in the
• 14:03
  background. I never touch it and it goes and
• 14:05
  it just runs AII use a a workflow that drafts replies to
• 14:09
  emails on my behalf based on a bunch of custom system prompts
• 14:12
  and access to services. And this is nothing like ChatGPT
• 14:16
  right in, in an from an identity management perspective, this is
• 14:19
  just a a service account workload.
• 14:22
  And yet we kind of throw these all under the heading of agent,
• 14:26
  even if it doesn't make any sense.
• 14:29
  Yeah, I feel like I've been hearing this comment a couple of
• 14:32
  Times Now recently where like you think of Agent as a bot.
• 14:38
  We spent our career keeping the bots out and now we have to
• 14:42
  figure out how to let the bots in or the right bots in and
• 14:46
  let's do the right things. Just what do you think when you
• 14:49
  hear that? Yeah, I mean, I think the
• 14:52
  concept of a bot is a is a terrifying word to use nowadays.
• 14:57
  So one of the ways AI can interact with services is
• 15:02
  through the web. And like we've been trying to
• 15:04
  fight bots on the web for years and years and years.
• 15:07
  And now one of the things people want is like, oh, we, we want
• 15:11
  those bots. Because if, if I'm going to use
• 15:13
  Chat GPTS agent mode, which, you know, opens a browser and clicks
• 15:17
  buttons very slowly to buy stuff, if I block that bot,
• 15:23
  well, then people stop buying stuff on my website.
• 15:26
  And so then you go to whoever is building the bot blocking
• 15:29
  software and says, well, we, we don't want the bots, but we do
• 15:32
  want those bots. And this has led to this kind of
• 15:36
  blossoming of questions and solutions.
• 15:38
  Web Bot Auth, which is a standard supported by
• 15:40
  Cloudflare, a browser based to identify bots on the web.
• 15:44
  Payment questions, people doing X4O2 where you know, four O 1
• 15:48
  being the the authorization needed error code, four O 2
• 15:52
  being payment needed. And so there's all these wacky
• 15:55
  questions that you know people had written off over the last
• 16:00
  decade or two that now are coming back into question.
• 16:02
  It's a fun time to be in the Internet.
• 16:04
  I find it so fun that we are, you know, we've spent the last
• 16:07
  decade trying to prove that we are not bots.
• 16:09
  You know, pick all the pictures of a hot dog, pick all the
• 16:12
  pictures of a motorcycle. And what are we doing now?
• 16:15
  We're training our bots to do those things, you know, on our
• 16:17
  behalf. That's just where we are.
• 16:20
  If anything, I think AI is getting better at doing captures
• 16:23
  than I am. Yeah.
• 16:25
  Well, I, I'm not very good at it.
• 16:27
  A lot of times it's like pick the, pick all of the pictures
• 16:30
  with the bus. I'm like that one with traffic
• 16:34
  there. Is that a bus?
• 16:35
  I don't know. I don't know.
• 16:38
  I usually get through those, so I figure I must not be answering
• 16:41
  that poorly for now, you know? We might remove your human
• 16:46
  status soon. You know, so in my mind, there's
• 16:51
  different kinds of agents. And I like I said, I like the
• 16:54
  term assistant because I can see a scenario where people are
• 16:59
  creating their own assistance. And so wondering from like an
• 17:04
  identity standpoint when you're talking about AI agents and the
• 17:07
  ability for them to impersonate a person, right?
• 17:13
  And I think that's a big topic that's in the paper that you
• 17:15
  wrote. It's like, ultimately that's
• 17:17
  what we're driving toward, right?
• 17:19
  Is the ability to set up an identity for this assistant
• 17:25
  agent bot, you call it what you want.
• 17:28
  Who's setting that up? Is that kind of coming with the
• 17:31
  package or is the end user setting that up?
• 17:34
  Or does it make sense for an administrator to set that up?
• 17:38
  And please tell me that we don't end up just put your username,
• 17:43
  password into this agent and we're just going to log in as
• 17:46
  you. I mean, isn't that like the the
• 17:48
  worst case scenario? That that that use case scenario
• 17:53
  keeps me up at night. The so think of this back when,
• 17:57
  when we had human assistance. I like the term assistance.
• 17:59
  We've had human assistance for many years.
• 18:02
  And how would you give someone access to your services?
• 18:12
  You could just give them or your password, give them your one
• 18:14
  password admin login. Not a great idea, but people do
• 18:21
  do it. There's a version of that you
• 18:23
  can do with AI, But the question is, how much do you trust that
• 18:26
  human being? And so if you want to, you know,
• 18:30
  have some modicum of security involved, you probably want to
• 18:34
  give them selective access to some services.
• 18:38
  And you probably want to know what actions are taken by them
• 18:40
  versus taken by you. Was the e-mail sent by an
• 18:43
  assistant or was it sent by you? And the reason AI is interesting
• 18:48
  here is because this is just expanded in, you know, the the
• 18:52
  scope of what can be done is expanded.
• 18:54
  More people are going to have these assistants helping them
• 18:57
  and there's going to be, you know, exponentially greater
• 19:00
  risks as these AIS do things faster than you can oversee.
• 19:06
  And that is why I think we really need to rethink identity
• 19:09
  for this. I think to kind of answer the
• 19:11
  question you actually asked of who is setting up this identity,
• 19:15
  I think a lot of the owners in a consumer application is on the
• 19:18
  consumer chat box, people who are building these assistants.
• 19:22
  It's also on the web and standard setters to build robust
• 19:28
  notions of identity into that. And then in the enterprise use
• 19:31
  case, we, I think a lot of the really robust identity
• 19:34
  infrastructure comes to play. It's on enterprises to have
• 19:38
  sensible policies and procedures around identity access
• 19:42
  management of ages. So when we talk about this, I
• 19:46
  mean, there's essentially there's major topic that you
• 19:50
  brought on, it took on in the paper impersonation versus
• 19:53
  delegation. I haven't.
• 19:55
  It's interesting because it brought me back to what I see
• 20:00
  like in my early days in IT where an executive would have an
• 20:05
  executive assistant. The executive didn't even know
• 20:08
  how to go in and do an access review, if you even had that at
• 20:12
  that point. But they didn't know they would
• 20:14
  go in and like, approve access and.
• 20:17
  I actually work for people where their assistant would print out
• 20:21
  their emails and put a stack of paper on their desk so they
• 20:24
  could go through their emails. Yeah, such situation.
• 20:29
  But it kind of, it's kind of the human version of this process
• 20:34
  Now. I guess the question is like,
• 20:36
  what is the right way? Is that what you guys are
• 20:40
  driving toward? And is it to set up some kind of
• 20:43
  standard for like what is acceptable, what is best
• 20:46
  practice? Because I think from a
• 20:48
  practitioner standpoint, there's one of those areas where it's
• 20:50
  like, just tell us what to do. Well, Jim, I can build you an AI
• 20:56
  agent that will print out all your emails so you can go back
• 20:59
  to handwriting them. I think that maybe that's a
• 21:01
  better world. We do need to stop using all
• 21:03
  these phones all the time, so. Maybe that I'll just say please
• 21:07
  don't like. Yeah, I just got an iPad.
• 21:09
  It's still in the box but I'm going to just doing my emails
• 21:13
  there. You know, I, I think a lot of
• 21:16
  people believe that AI agents are going to solve all of our
• 21:19
  problems. And I think much in the same way
• 21:21
  you think an iPad is going to make you better at replying to
• 21:24
  emails quickly. It's not you're the problem.
• 21:28
  And it doesn't matter how good I have an AI that writes emails
• 21:31
  for you, drafts emails, and yet this limiting reagent is still
• 21:35
  me clicking set because I'm not going to let AI send the emails
• 21:39
  on my behalf. Draft, OK, but send would get
• 21:43
  hairy very quickly. So your question, what is the
• 21:47
  right solution? I really wish I could come on
• 21:51
  this podcast and say to you that like here is a magic fix that
• 21:56
  will solve all your problems that doesn't exist.
• 21:59
  It's not ever going to exist and anyone trying to tell you it
• 22:02
  will is selling you a product instead.
• 22:06
  There's a bunch of steps we can do with current web identity
• 22:10
  infrastructure, delegated authority tooling and enterprise
• 22:13
  identity infrastructure to make things ready for AI agents or or
• 22:18
  you know, be using them already and you see this already in kind
• 22:23
  of the enterprise use case. You want to have an enterprise
• 22:25
  assistant, human assistant. We have some tools for that in
• 22:28
  whatever identity provider you use.
• 22:31
  Well, most of these identity providers are also setting up
• 22:33
  tooling around agents so that you can declare that something
• 22:36
  is an AI agent, that it is delegated on behalf of a human
• 22:39
  being. You can provision it, de
• 22:41
  provision it in enterprise identity management tooling.
• 22:45
  And essentially everyone in this ecosystem is very aware that
• 22:49
  this is a problem and working towards it.
• 22:52
  And so I may not have one solution, but I'm promising you
• 22:55
  will get a lot of emails over the next year of people selling
• 22:58
  you solutions to fix them. So we've got that to look
• 23:01
  forward to it every conference. You know, it's already starting
• 23:03
  to show up as like manage your agentic AI.
• 23:06
  You talked about that delegation and you know, that's going to be
• 23:09
  a big part of it because the whole point of agentic AI is
• 23:14
  autonomy, right? It's doing things on your
• 23:16
  behalf. And so of course, it has to have
• 23:18
  some level of delegated authority.
• 23:19
  But that delegation does not absolve, you know, the
• 23:23
  governance, the consent, right, that needs to take place.
• 23:26
  Say, yes, I do authorize Jeff Bot to go do these things.
• 23:30
  How do I keep track of that? Because that human oversight
• 23:33
  that is supposed to be there, typically over like a human
• 23:37
  level ID operating at human time scale is going to be very
• 23:40
  different when you're talking about agentic, meaning
• 23:43
  millisecond time scale and hundreds, thousands maybe of
• 23:47
  these bots, maybe millions of these bots kind of all doing
• 23:49
  their own thing. Like how do you manage that that
• 23:52
  automation and make sure that's governed and that the right
• 23:55
  consent is in place to for it to for it to do those things?
• 23:59
  Yeah, I mean, I really like that comment that AI is about
• 24:03
  autonomy and I think the better AI is the more autonomy you give
• 24:07
  it. The dream of everyone building
• 24:10
  these highly effective long running AI agents is to give
• 24:14
  them one instruction or maybe an amorphous goal and they
• 24:17
  continuously run and solve problems for you.
• 24:20
  That is the exciting part. That is what all the evaluations
• 24:22
  for AI Labs is built upon, and I think they're going to solve
• 24:27
  that. I think all of the AI models
• 24:28
  will get better at long time horizon tasks.
• 24:31
  The expression that has been going around Stanford recently
• 24:33
  is that AI is the worst it's ever going to be.
• 24:38
  And so if you don't believe that an agent is capable of
• 24:41
  continuously running a task effectively using tools right
• 24:45
  now, at some point in the future, whenever that is, you
• 24:48
  can believe whatever timeline you want, it will get better.
• 24:51
  But I think the block on that, the thing that's going to stop
• 24:55
  it being able to do stuff is access is commissions, human
• 25:00
  oversight, management of the access to services for AI agents
• 25:04
  is going to be the limiter on really cool agents.
• 25:09
  So I find interesting that you mentioned that because we've
• 25:11
  been saying that for a while, like this is the worst it will
• 25:14
  ever be. I've had to change that saying
• 25:16
  though, because worst implies to some people good and bad.
• 25:21
  And So what I've said is like, this is the least capable AI
• 25:24
  will ever be is right this second.
• 25:26
  You know, as we're recording this, it will be better or it'll
• 25:29
  say it'll be more capable. 5 minutes, 10 minutes, right and
• 25:32
  so forth and so forth. So I do find interesting that,
• 25:35
  you know, hey, Jim, we're, I can, we're, I can good circles
• 25:37
  here. We're we're thinking like
• 25:39
  Stanford folks. I know it's really, it's really
• 25:43
  eye opening. You know, one thing that I kind
• 25:46
  of feel like I've found is some of the call centers that we dial
• 25:52
  into like the AI is getting really, really good.
• 25:57
  We can, I mean, some companies are under invested and you know,
• 26:01
  there should be a wall of shame for that because you get on and
• 26:04
  it's like your AI is garbage and you just say I need to speak to
• 26:08
  a human, I need to speak to a human.
• 26:10
  But there are others where they're actually good.
• 26:13
  But I feel like no matter how good they are, eventually they
• 26:17
  hit some kind of guardrail where it's like the AI agent just
• 26:23
  isn't going to solve that problem for you.
• 26:25
  It's like, hey, I got billed twice last month and I need you
• 26:29
  to refund me the money. I was like, OK, well, this could
• 26:33
  be a fraud scenario, so we're actually going to forward you to
• 26:35
  a person. And why do I bring all that up?
• 26:40
  Because I kind of feel like if you, you know, you can say, OK,
• 26:45
  here's Tobin. He's a human being and he's got
• 26:47
  these five agent identities and those agents go off and do
• 26:53
  something. You almost potentially could
• 26:57
  say, well, I didn't know what that agent was going off to do,
• 27:00
  right? Like if there's not enough
• 27:02
  logging and accountability behind those agents.
• 27:05
  So you almost have to in addition to just logging it, but
• 27:11
  you have to have the right guardrails.
• 27:12
  And maybe it's maybe it's a matter of having the right
• 27:15
  authorization so that it can't do things that you wouldn't
• 27:18
  expect it to do. But I also think there's process
• 27:21
  guardrails. And so identity management and
• 27:25
  identity security bump up to you know, we've got some of this has
• 27:29
  got to fall back to the business process.
• 27:33
  Yeah, I, I think the element that you're describing with
• 27:37
  respect to liability is a huge, huge issue here.
• 27:40
  We, we have a bunch of work on going at Stanford in
• 27:43
  collaboration with part of the Stanford Law School on what
• 27:47
  liability looks like in AI. The law has a concept of agency
• 27:52
  with not to be confused with AI agents, but the idea that an
• 27:55
  individual can take actions. And if you have the agency to
• 27:59
  take an action, you may have the liability as part of that.
• 28:02
  And it's very unclear when you tell ChatGPT or Claude to go and
• 28:08
  do a thing and then it does that thing kind of wrong, right?
• 28:12
  You go and say, buy me a new couch and it doesn't have enough
• 28:17
  money to buy a couch. So it remortgages your house in
• 28:21
  order to pay for it. Who's responsible for that loan,
• 28:25
  right? Like that's a, that's a
• 28:26
  consequential action. And so I, I deeply believe that
• 28:31
  the way organizations and business processes are
• 28:33
  structured is someone needs to be fireable for anything to
• 28:37
  work. And if you just have AI doing a
• 28:40
  bunch of things and something goes wrong, who do you yell at?
• 28:43
  You can't really yell at ChatGPT.
• 28:45
  Yeah, I mean, you can. It will.
• 28:46
  It will accept it, it will apologize.
• 28:48
  It doesn't. It doesn't feel the same.
• 28:50
  You're right, I did make a mistake.
• 28:53
  You don't want that. It, it, it doesn't fight back
• 28:55
  enough, right? Yeah.
• 28:58
  So I, I think guardrails matter there.
• 29:02
  Part of guardrails is, is you know, this alignment problem,
• 29:04
  making sure that AI stays on task.
• 29:06
  Part of it is just limiting its access so it can't remortgage
• 29:10
  your house and then part of it is understanding when you design
• 29:15
  AI system that assistance where the risks and the liabilities
• 29:19
  are in that business process automation.
• 29:22
  Well, this is sort of like the, you know, trillion dollar
• 29:25
  question is who is responsible for the actions of the agent,
• 29:28
  right? Ultimately probably a human
• 29:30
  somewhere, but they're not going to be able to keep up with their
• 29:33
  own agents, let alone all the agents within an organization.
• 29:35
  So let me try to spin this into a way that we can make this
• 29:39
  helpful, helpful for people listening.
• 29:42
  Yeah, I'm going to sit down and I'm going to vibe code, you
• 29:45
  know, an app that prints out all of Jim's emails and sends, you
• 29:49
  know, the post surface, you know, Postal Service to him.
• 29:52
  So to do that, I'm going to have to do some things around agentic
• 29:56
  AI and maybe some other things, right, to tap into.
• 29:58
  I, I want to talk about MCP in a little bit, but I'm, I'm doing
• 30:01
  this type of thing. So what do I do as a developer
• 30:06
  or an enterprise Oregon, even a consumer to say how do I get
• 30:11
  ready for this from an identity management perspective?
• 30:16
  Yes. So every tool that lets you plug
• 30:21
  in a tool such as using the model context protocol and
• 30:25
  something like clod or if you're using clod code accessing
• 30:29
  different tools, there are usually interfaces designed for
• 30:34
  humans and then there are interfaces designed for AI
• 30:37
  agents. And to give a concrete example,
• 30:40
  if you're vibe coding, you like clod code and you want to use
• 30:43
  GitHub, there are two different ways you can use GitHub to make
• 30:47
  PRS, you know, make a, you know, merge them, manage your commits,
• 30:53
  manage your, your, your branches.
• 30:55
  And one of them is the command line.
• 30:59
  You can just Claude knows how to write a, a GitHub command and
• 31:02
  the other one is an MCP server. Now this is like lesser known
• 31:07
  about that MCP server, but if you use the GitHub MCP server,
• 31:10
  it has explicit restrictions on what can and cannot be done.
• 31:14
  And it will make sure that when PRS are made, it's tied to an
• 31:20
  agent identity. GitHub is tracking when
• 31:22
  something comes through MCP, when it comes through an AI
• 31:25
  agent, as opposed to coming through your command line, which
• 31:28
  looks like you. That is one of those benefits of
• 31:32
  using the right interface for AI.
• 31:35
  And I think this analogy can apply a bunch about across a
• 31:40
  bunch of different tools. There's a bank that has an MCP
• 31:45
  server that will let you read banking information and automate
• 31:48
  your finances. Now it's read only.
• 31:51
  Everyone can remain calm. We're not vibe coding our banks
• 31:54
  yet, I hope. But having that interface and
• 31:59
  declaring an upfront, especially if you're building AI tools, I
• 32:02
  think it's super helpful, not just for you and your own
• 32:04
  management, but to offer to your consumers to, you know, other
• 32:09
  folks using the tools you build. That something is AI and work
• 32:12
  towards this better identity infrastructure.
• 32:15
  So talk to us a little bit about MCP, just if you just kind of
• 32:19
  like start with the level set of what MCP is and then talk about
• 32:25
  identity and authorization and how that plays in how that works
• 32:29
  through MCP. Brilliant.
• 32:32
  We enter the three letter acronym territory.
• 32:35
  OK, so the model context protocol MCP is a protocol that
• 32:41
  came out in November, December last year, has rapidly taken
• 32:45
  over the world. Every company I know is thinking
• 32:47
  about the MCP strategy. And I think the best, worst way
• 32:54
  to think about it is as a wrap around APIs.
• 32:58
  People are used to the concept of an API.
• 33:00
  You have some resource, you have some SAS app or Gmail, whatever
• 33:05
  that you want to expose to someone building sub link to
• 33:11
  allow them to perform actions. This is always had notions of
• 33:15
  scopes. Often they involve API keys or
• 33:19
  an overflow where users consent to give access to services to
• 33:24
  resources and MCP behaves as a kind of first approximation as a
• 33:31
  wrapper in front of that. That makes it really easy to
• 33:35
  connect to an AIA agent. So it kind of operates in
• 33:40
  natural language. It's this translational barrier
• 33:42
  between a natural language interface that's happening in
• 33:45
  Claude dot ChatGPT and the API which is in more of a machine
• 33:50
  readable format now, so. Basically, if you're so let's
• 33:55
  just make sure I understand, but if you're of course trying to
• 33:58
  develop an AI agent to do something like create that
• 34:04
  calendar reminder, the MCP gives you a simple or should I say a
• 34:13
  common way to interface with the e-mail system.
• 34:19
  Is that right? Yeah, so in this instance, you
• 34:23
  already have a Google Calendar API, but that's super painful to
• 34:28
  plug into Clot. Like if you go to the ChatGPT
• 34:31
  website right now, there's no way to give it access to that
• 34:35
  something you've custom built in terms of, you know, sending
• 34:38
  emails. But if you make it an MCP
• 34:40
  server, now you can add that to ChatGPT.
• 34:43
  And now, you know, each morning when you wake up and you want
• 34:45
  to, you know, have your like friendly morning conversation
• 34:48
  with your friend, ChatGPT, you can go and, and send some emails
• 34:52
  while you're at it. And so it's just this connective
• 34:57
  tissue that lets you write instructions in natural
• 35:02
  language. You literally, when you build an
• 35:03
  MCP server, write text that explains how to use different
• 35:08
  APIs, how to use different tools to connect to resources.
• 35:13
  So practically speaking, if I'm creating enterprise
• 35:17
  applications, am I getting my MCP server?
• 35:22
  I guess I probably have options, but what is going to be the
• 35:25
  typical enterprise you see? So are they going to have a SAS
• 35:30
  MCP server that has all the plugins that you could possibly
• 35:34
  want? Or is this something that I'm
• 35:36
  going to stand up or am I going to get it from my AI large
• 35:42
  language model provider? So I really like the internal
• 35:47
  enterprise example use case. We're going to talk about
• 35:50
  external use cases later if we want.
• 35:53
  But you as an employee have a bunch of extremely boring tasks
• 35:57
  you perform on a day-to-day basis that involve opening a
• 36:00
  website and clicking a series of buttons in the same manner every
• 36:04
  single day. We all do.
• 36:06
  And there are in many cases APIs that could do that for you, but
• 36:12
  you are not going to build button clicking automations
• 36:16
  because you're busy clicking buttons and you know you've only
• 36:18
  got so much time in the day. What MCP lets you do within a
• 36:22
  company is take those operations, which can be done
• 36:25
  programmatically and put them in a natural language interface.
• 36:29
  So wrap them in an MCP, plug that into your company
• 36:33
  employees, chat bots of choice, whether it's ChatGPT, clawed
• 36:36
  goose cursor and then the human being, it can go and say, you
• 36:43
  know, make you know, check my compliance report can pull
• 36:46
  relevant information. It can, you know, write relevant
• 36:49
  things, change state externally and do those workflows that you
• 36:52
  always had to do by hand, but now do them with a chat bot.
• 36:56
  And as you know, you find yourself doing these repeated
• 37:00
  chat bot operations. I always do the same thing.
• 37:03
  Check the versal deployment for any logs, pull it in, suggest a
• 37:07
  fix, write that code, make a new PR, and if you keep seeing this
• 37:12
  happen repeatedly, now you can build a semi autonomous agent
• 37:15
  that goes and just does this workflow repeatedly.
• 37:18
  And I like that mental model of of automating you as a human
• 37:22
  being by replacing you with an AII.
• 37:25
  Kind of think of like the Matrix and the key master.
• 37:29
  So, and this is just my feeble brain trying to wrap my head
• 37:33
  around this, and I think of each door that the key master has a,
• 37:36
  you know, key to is a portal into another world in the MCP
• 37:41
  context of MCP, maybe somewhat redundant right there at the
• 37:46
  word context, but it's how, you know, a individual, a thing, an
• 37:51
  agent, a bot or whatever it may be can traverse boundaries in a
• 37:55
  controlled method. That's the way I kind of look at
• 37:57
  it. It's like, OK, yeah, you got Neo
• 37:59
  and the agents going through, you know, a bunch of portal
• 38:02
  doors kind of in the Matrix, and they shut them behind them.
• 38:05
  And when they're shut behind, no one else can kind of flow
• 38:07
  through it or, you know, go back into it, those sorts of things.
• 38:10
  Am I crazy or you know, did the Matrix invent MCP?
• 38:14
  I, you know, I'm not sure I'm willing to go on record to say
• 38:17
  the Matrix invented MCPI. Think David from Anthropic might
• 38:21
  be a bit annoyed at that claim. But I do think that mental model
• 38:25
  makes sense. It is giving you this portal,
• 38:28
  two things that have already existed, these these resources,
• 38:31
  these APIs or always existed, but now making them accessible
• 38:36
  to both AI and through AI humans in a really interesting way.
• 38:42
  I want you to put your Nostradamus hat on a little bit
• 38:45
  here, right? And talk about MCP.
• 38:47
  Is this like the next SAML or is that not a good analogy?
• 38:51
  Like I'm just thinking you were SAML, you got this humongous
• 38:54
  adoption and like there's all these like ID PS that are SAML
• 39:00
  enabled and you can plug in and you can have single sign on to
• 39:04
  hundreds or thousands of applications overnight.
• 39:07
  And I'm wondering if MCP is going to be that.
• 39:11
  Because I think with SAML, it got to the point where I was
• 39:14
  like, if you're going to provide a web application online, you
• 39:18
  pretty much need to support SAML if you're going to sell to
• 39:22
  enterprises. Because unless you're like the
• 39:25
  only one in the market, you're going to need that.
• 39:28
  And I'm wondering if like MCP becomes that standard, do they
• 39:34
  just publish APIs and let somebody else worry about doing
• 39:37
  the MCP layer? Yeah, I, I, I, I'm not sure SAML
• 39:42
  is the protocol I'd pick, but I totally agree with that analogy
• 39:45
  of I think that's what it's going to become.
• 39:46
  Maybe it's closer to to to REST in that it's a thing that
• 39:50
  everyone decides we should start doing and then everyone does,
• 39:54
  and then we all get annoyed at the limitations of it and it
• 39:57
  causes a constant pain for developers.
• 40:02
  Speaking of that, they're coming.
• 40:02
  Behind you if it gets you. Yeah, I, I, I, I've said
• 40:05
  something wrong. I, I'm in San Francisco on
• 40:08
  Market Street, which is not the calmest place on planet Earth,
• 40:12
  but it is where all the MCPS are being built.
• 40:17
  So I, I, I really think employees and organizations are
• 40:21
  going to expect if you're selling them a SAS app, that
• 40:24
  that SAS app is accessible to people who want to just operate
• 40:27
  out of a chat bot. We work a lot with folks at
• 40:29
  Cursor who build cool apps and they love MCP at Cursor and one
• 40:34
  of many coding agents. All the coding agent folks do,
• 40:37
  and they want to vibe code their whole lives.
• 40:39
  They just want to sit and have AI do all the things.
• 40:42
  And so they really want all of their services to have MCP
• 40:45
  servers so that they can, you know, stay inside of cursor for
• 40:48
  the rest of their lives and never have to go and open a
• 40:50
  website. Well, that's a, you know, that's
• 40:53
  a monetary driver too, right? Because that MCP is a two way
• 40:56
  door. You know, if we're doing
• 40:57
  consumption based from a product perspective, that's how you
• 41:00
  adequately control and secure those things to make sure that,
• 41:04
  you know, obviously the data they're getting is is what it
• 41:07
  should be and and so forth. So to me, that makes a lot of
• 41:10
  sense. Like, yeah, of course you want
• 41:11
  to build that. It's almost like like that
• 41:13
  should be a function that is just standard at this point.
• 41:17
  Or is it too soon to call it standard?
• 41:20
  I think we are now at a point or or let me give you this in like
• 41:25
  December 20th of 2026, the standard there are working
• 41:31
  groups, essentially every company that I know has someone
• 41:37
  working on IMCP. And I like deeply believe that
• 41:41
  there will be a product push over the next year for everyone
• 41:44
  who wants to be AI forward to support something.
• 41:48
  And so we'll see if this collapses.
• 41:51
  I mean, look, AI may be one big bubble, but I think this
• 41:54
  protocol is here to stay at least as a really clean, robust
• 41:59
  way to communicate that tooling is designed for AI and optimize
• 42:05
  the agent experience, right? API is a design for a developer
• 42:09
  experience. Your website is designed for
• 42:10
  humans. We need an interface that's
• 42:13
  optimized for agents. So talk to me a little bit about
• 42:16
  the concept of recursive delegation and scope
• 42:19
  attenuation, because these are things that I picked up from the
• 42:22
  white paper. And I should say we're going to
• 42:24
  have the white paper LinkedIn, our show notes, so people can
• 42:26
  kind of go and read that. It's like 30 pages.
• 42:28
  It's, it's, it's very in depth, right?
• 42:30
  That sort of thing. But talk to me about what those
• 42:32
  things are and why should I care about them as an identity
• 42:37
  person? Those are a lot of big words.
• 42:40
  Frankly, very scary. Let's unpack them.
• 42:44
  OK, delegation. You.
• 42:46
  When you want a task dub, you delegate it to something or
• 42:49
  someone or some agent. I ask Claude, my favorite AI
• 42:53
  assistant, go read all of my emails and find what needs to be
• 43:00
  addressed. Great.
• 43:02
  That's one step of delegation. Claude's now handling
• 43:04
  everything. Now for one of those emails, I
• 43:07
  need to go retrieve information, which maybe requires me to ping
• 43:10
  someone or go do other stuff. And so Cord has this idea of sub
• 43:14
  agents where it can spin up a new AI agent.
• 43:18
  Now really all that's happening is a sub process is being
• 43:21
  spawned in a system with a clean system prompt, right?
• 43:25
  So a lot of AI is about context management.
• 43:27
  If you have a lot of nonsense in your context window, AI gets
• 43:30
  confused. And so it makes a lot of sense
• 43:32
  to spin up a sub agent with a reduced context.
• 43:36
  It's focused on exactly what it needs to do, and it can go out
• 43:39
  and do that and retrieve relevant information.
• 43:41
  So now I have delegated to an AI, which is delegated to an AI.
• 43:47
  Now let's take a enterprise use case that now this second agent
• 43:53
  needs to orchestrate something in Salesforce, needs to pull
• 43:56
  from ACRM. Well, Salesforce has Agentforce,
• 43:59
  the whole big agent strategy. And so it's now going to send a
• 44:04
  natural language request to an external system that we don't
• 44:08
  control and which, you know, agent force will go and it'll
• 44:12
  pull the CRM and find someone that might be relevant.
• 44:15
  And then it can send it back to the sub agent, sub agent, which
• 44:18
  can send it back to the sub agent, which can send it back to
• 44:20
  me. Now this is cool.
• 44:23
  It helps with context management, but this recursive
• 44:27
  delegation is multiple steps of delegation.
• 44:30
  Each one abstracts you, the original person, further and
• 44:36
  further away from what's going on and means you have less
• 44:38
  control over what's actually happened.
• 44:41
  And so I might give my main court access to all my tools,
• 44:47
  but I don't necessarily want everything that Claude creates
• 44:52
  to go do a task to have access to all my banks, all my text
• 44:55
  messages, all my emails. And so we really need robust
• 45:00
  attenuation of scope so that when AIS go and use other AIS,
• 45:05
  they don't just give all the permissions to the next AI.
• 45:10
  And that's what we mean about scope attenuation for a cursive
• 45:13
  delegation. A terrifying sentence.
• 45:16
  That is like a a $25 word. If you're playing Scrabble, you
• 45:20
  know, you probably just won right there.
• 45:22
  Or words of friends, or whatever your word choice is for.
• 45:24
  Their If anyone ever plays Scrabble and gets attenuation,
• 45:28
  please e-mail me. Immediately.
• 45:31
  There you go, open invite. I'll put your LinkedIn and our
• 45:33
  show notes. You know, you heard it here
• 45:35
  first. The the gauntlet has been thrown
• 45:37
  down for a challenge. All right, so we're like 45
• 45:40
  minutes here and I want to start to wrap things up.
• 45:43
  But what I want to do is offer the people listening some nugget
• 45:48
  that they can take away. So if I'm listening to this and
• 45:52
  I remember one thing that came out of this entire conversation
• 45:55
  around agentic AI identity management, what is it that I
• 46:01
  should be taking away from this conversation?
• 46:02
  Just remember this one thing. We don't need to reinvent the
• 46:08
  entire identity stack of the Internet to solve agents, and
• 46:12
  we're not going to. But each step of the existing
• 46:17
  identity stack, whether it's web identities logging in, whether
• 46:22
  it's SAML, whether it's Skim and identity management agents with
• 46:27
  your identity provider, each one of those steps is going to
• 46:30
  require some incremental improvements to support agents.
• 46:35
  And we're not going to change the world entirely, but everyone
• 46:38
  is going to have a responsibility to upgrade their
• 46:41
  systems to support agents. And then I think we're all going
• 46:44
  to be OK. There we go.
• 46:45
  You make it sound so simple, like this is this is it?
• 46:48
  This is just how we solve it. So you said originally you, you
• 46:50
  didn't have, you know, a way to solve it.
• 46:52
  It sounds like you just did. I think we have so many fun
• 46:56
  incremental problems to solve that whether we have AGI or not,
• 47:00
  we're all going to stay employed for a while as identity
• 47:03
  management professionals. Yeah, judge it by some of the
• 47:05
  outputs that I get out of the tools that I use.
• 47:07
  Like, OK, yeah, we're not quite there yet.
• 47:09
  But you know, this is the least capable it will ever be right
• 47:14
  this second. It's just going to get more and
• 47:15
  more capable. Now it's on us to figure out how
• 47:17
  do we make it secure? You know, how we use it
• 47:20
  appropriately, right? All that good stuff.
• 47:21
  So as long as I can, you know, use, you know, things like Sora
• 47:25
  and Gemini to like make stupid videos, you know, I'm.
• 47:27
  I'm cool. Yeah, I like you.
• 47:30
  It's neither good nor bad, but it will get more capable.
• 47:32
  Those Sora videos might become increasingly unhinged, but
• 47:35
  they'll be credible. Definition of quality.
• 47:38
  Yeah, if you have not checked out the Sora app from Open AI,
• 47:41
  it is definitely very, very weird.
• 47:45
  If you would have sent it out, don't.
• 47:47
  Straight away it is 100% AI slop and if you're into that, it's
• 47:53
  great. If you're not, you're going to
• 47:54
  be like, what the heck? Is this?
• 47:58
  So Tobin, you've been super generous with your time.
• 48:00
  I want to take us out on a lighter note.
• 48:02
  And so I have a couple options. I was going to kind of ask you
• 48:05
  about Australia and you know, something about Adelaide maybe
• 48:09
  like the people who generally don't know about, but I want to
• 48:11
  go more the coffee route. This is what Jim was alluding to
• 48:13
  earlier on of, you know, providing value.
• 48:17
  I personally do not drink coffee.
• 48:19
  I love the aroma, can't stand the flavour.
• 48:22
  And I will try coffee maybe like once every year, a couple years
• 48:26
  and it's just a quick check in. Yep, still don't like it.
• 48:30
  So my caffeine of choice is generally a cola of some sort.
• 48:33
  But I want to ask you because as I was doing some, you know,
• 48:36
  cyber stalking of you to like get some background, you know,
• 48:38
  before the interview, you mentioned somewhere that you're
• 48:41
  kind of like a coffee guy. And so I want to know in all
• 48:43
  your travels around the world, what is or where is or who is,
• 48:50
  has the best coffee that you've ever had.
• 48:53
  I have such a strong affinity for coffee.
• 48:57
  I think my PhD is was fully powered by coffee.
• 49:01
  I continue in order to get my job done, consume inordinate
• 49:04
  amounts of caffeine. I've I've a a strong belief that
• 49:08
  as well the derivative matters, you've got to keep drinking just
• 49:11
  a little bit more every day, which doesn't work at the macro
• 49:14
  scale of 1's life. But being Australian and growing
• 49:19
  up in one of the coffee havens of the world, I'm always in
• 49:23
  search of good coffee and always disappointed.
• 49:26
  San Francisco, where I live, has many great cafes.
• 49:29
  I can't, I can't knock them, but it's it's unreliable.
• 49:32
  When you travel to Europe they make coffee but it's not the
• 49:36
  same as the flat whites I care about back home.
• 49:39
  And when you fly back in to Australia, usually via Sydney,
• 49:43
  because there is no direct flight to my hometown, you land
• 49:47
  in Sydney Airport and you're immediately faced with five
• 49:51
  different cafes in the terminal, each of which has award-winning
• 49:54
  coffee. And so Australia will always
• 49:59
  have the best cafe culture in in my heart.
• 50:03
  But if you are in San Francisco, Third Wheel Coffee is my local
• 50:08
  cafe and there's a 5050 chance you might see me there at 8:00
• 50:12
  AM every single morning. So there's there's my offer to
• 50:16
  folks. All right, so go to Australia,
• 50:19
  go to Sydney for the Opera House and stay for the coffee in the
• 50:22
  airport. They're like, that's a Jim, I
• 50:24
  know you're a coffee person. You seem like you're itching,
• 50:27
  you know, to get into the conversation.
• 50:28
  I don't think that's just the caffeine or you're just excited,
• 50:30
  but what is your favorite coffee?
• 50:33
  Maybe a little bit of both, because I do drink coffee
• 50:35
  throughout the day. I'm more prepared to say what I
• 50:40
  don't like. You know, me chef.
• 50:42
  I mean, is that any surprise at all?
• 50:44
  So I. Mean, I already bashed on the
• 50:46
  poor Apple girl from the commercial, so go ahead, you
• 50:48
  know, bash coffee, that's fine too.
• 50:50
  Yeah, I don't like flavored coffee.
• 50:53
  Like if you say oh, like this is a caramel coffee or something
• 50:57
  like that, I'm like, no, that is horrible.
• 51:00
  It's ruined. I think the weird thing is like
• 51:04
  I've spent my whole life drinking different coffees,
• 51:07
  African coffee or different, different places around the
• 51:12
  world trying to find like, OK, what is my favorite roast of
• 51:17
  coffee or favorite beans? Favorite roast, The order to
• 51:21
  come back to in my old age is the good old Arabica bean, 100%
• 51:26
  Arabica coffee. And whether you make it into an
• 51:31
  espresso or you make it into a drip coffee, I just really think
• 51:36
  there's a reason that became the most popular bean in the world.
• 51:40
  That's because it's the best. It's got like the right, it's a
• 51:43
  medium flavor. And I think that's me, like I, I
• 51:48
  tried both extremes and I ended up back where I started as a
• 51:52
  teenager drinking, you know, arabica, Colombian coffee.
• 51:57
  So I'll go with Colombian coffee for 200 please.
• 52:02
  All right, coffee crossed by Tobin.
• 52:03
  Your thoughts? I actually, I, I, I think if I
• 52:07
  ain't broke, don't fix it. It's a great coffee.
• 52:09
  It's so good to the point where I lived with a machine learning
• 52:13
  engineer who had quit caffeine entirely and yet still imported
• 52:19
  really high quality arabica beans just to make decaf coffee
• 52:23
  on what must have been 5 or so $1000 of espresso machine
• 52:28
  equipment for no caffeine at all.
• 52:31
  Which is a little insane, but it's tasty.
• 52:34
  It works. OK, so Jim, you, you mentioned
• 52:37
  the bean, but like where is the best coffee?
• 52:40
  Like what's the best coffee ever had?
• 52:41
  Give me a location. So I don't know, like I don't
• 52:46
  have like that one experience. I'd say my house like I keep
• 52:50
  coming I'm. Coming to this.
• 52:52
  Yeah, I keep coming here and having coffee all the time.
• 52:55
  Actually, I don't have my really good espresso machine anymore.
• 52:58
  I've got like a Nespresso machine, which, you know, people
• 53:02
  probably out there like grumbling.
• 53:04
  I know. It's like it's a little bit of
• 53:05
  weak sauce, except convenience is very important when you have
• 53:09
  a job where you're like, you've got calls like stacked up
• 53:13
  back-to-back and it's like, oh, this call ended 3 minutes early,
• 53:17
  I have time to go make myself a coffee.
• 53:21
  So yeah, that's what I have. But I I tend toward more local
• 53:26
  places than the big chains. And some big chains are really
• 53:31
  good, but a lot of times they're very disappointing.
• 53:34
  So there's a couple local places around me here in Sturgis, SD
• 53:39
  You know, you don't think of it as the coffee capital of the
• 53:42
  world, and it's not. You know, I could tell you I've
• 53:47
  had Turkish coffee, I've had coffee in all over Europe.
• 53:51
  They've all been wonderful. Nothing stood out to me is like,
• 53:55
  oh, that's the best place to go for coffee.
• 53:59
  But. Yeah, there's a lot of them.
• 54:03
  OK, so send your angry emails to Jim on LinkedIn.
• 54:07
  Defend your coffee, people. That's the challenge right
• 54:09
  there. You've everyone's got to have a
• 54:11
  hill to die and I think it's a great, it's a great hill.
• 54:15
  That's, you know, look, we all have opinions.
• 54:16
  It's like change my mind. The opinions can change, change
• 54:19
  the facts. I'll think about it.
• 54:20
  There's nothing wrong with that. Maybe someday I'll like coffee,
• 54:24
  but for now, it's still just the aroma.
• 54:26
  And, you know, that's it for me. So all right, let's go ahead and
• 54:30
  leave it there for this week. Tobin, thank you so much for
• 54:32
  spending some time with us. I hope you'll come back,
• 54:34
  especially if things kind of develop in this agentic AII AM
• 54:39
  world and like what should be you'll be thinking about as I go
• 54:41
  along. And I hope to see you at an
• 54:43
  identity conference coming up, maybe a Gartner Identiverse or
• 54:45
  EIC or something along those lines.
• 54:48
  But thank you so much for being with us.
• 54:49
  I will have in our show notes a link to the paper that is out
• 54:53
  there as well as your LinkedIn profile so people can either
• 54:56
  defend or attack coffee preferences or whatever it may
• 54:59
  be. Please be polite.
• 55:00
  Please to yeah. And yeah, we'll go ahead and
• 55:02
  leave it there for this week. So you can find us on the web,
• 55:05
  IDC, podcast.com, like and subscribe, hit the YouTube
• 55:08
  channel. We recently crossed over 1000
• 55:10
  subscribers on our YouTube channel.
• 55:11
  So thank you for that. Again, you know, we don't do any
• 55:14
  advertising. It's all word of mouth and so we
• 55:15
  appreciate everyone supporting us.
• 55:16
  So with that, we'll go ahead and leave for this week.
• 55:19
  Thanks everyone for watching and or listening and we'll talk with
• 55:22
  you all in the next one. You've been listening to
• 55:26
  Identity at the Center. We hope you've enjoyed the show.
• 55:30
  Make sure to like, rate and review, and we'll be back soon.
• 55:35
  But in the meantime, hit the website at
• 55:37
  identity@thecenter.com. See you next time on Identity at
• 55:43
  the Center.