Latest / The Retail Podcast / Why Enterprise Retail Tech Is Broken And How to Fix It | Jason Cottrell, Orium

Transcript

Why Enterprise Retail Tech Is Broken And How to Fix It | Jason Cottrell, Orium · The Retail Podcast

• 0:00
  Something that was a competitive advantage for your business just
• 0:03
  went away overnight. At the same time, something your
• 0:05
  customers have always wanted is now possible, and so your job as
• 0:10
  a leader is to figure out both before your competitors do.
• 0:24
  Hello and welcome back to the Retail Podcast.
• 0:26
  Today's guest is a bit of a grab, if you like, a little bit
• 0:30
  of a snatch because I was lucky to see Jason or what Mac
• 0:33
  Alliance were doing shop talk Spring, but we didn't actually
• 0:37
  get to talk. And obviously, I'll ask Jason to
• 0:39
  introduce himself in a minute. But Jason, you're the CEO of
• 0:42
  Auriem and Chair of Mac Alliance.
• 0:44
  Right. That's right.
• 0:46
  Yeah. Yeah.
• 0:47
  And the elected president actually started the mock
• 0:49
  alliance. That's OK.
• 0:50
  Acting President, OK. And basically, if you don't know
• 0:52
  what the Mac Alliance says, it's an organisation that's helped
• 0:56
  define the shift away from monolithic platform, which is
• 1:00
  quite interesting because you have some platform providers as
• 1:04
  members as well. But then now the conversation
• 1:06
  has moved on again, if you like, which is why we all love tech
• 1:10
  towards AI driven agentic ecosystems where decisions
• 1:13
  aren't necessarily just supported by technology, but
• 1:17
  actually made by the enterprises.
• 1:20
  So I think Jason, that's like the high level positioning.
• 1:23
  But I'd love, I have to, you know, before we get started,
• 1:26
  find out a little bit more about you and a little bit more about
• 1:29
  the Mac Alliance. I'm sure you know, there's so
• 1:31
  much more. Welcome to the show.
• 1:33
  Why don't you tell us a little bit about you and the Alliance
• 1:35
  before we get into the question? Hi there.
• 1:37
  Yes, Alex, great to be here. And so I'm Jason Cottrell,
• 1:40
  Founder and CEO of Oreum. Oreum being kind of the largest
• 1:43
  kind of specialty provider for focus on imposable commerce,
• 1:46
  customer data and digital experience.
• 1:48
  And so we're generally working with some of the kind of fastest
• 1:50
  growing Isvs and infrastructure providers for companies doing
• 1:54
  hundreds of millions or billions of dollars in GMV.
• 1:56
  And that tends to be kind of mid to large enterprise, right?
• 1:59
  So, you know, really helping to make sure that we can use
• 2:02
  composability, you know, multiple systems to serve kind
• 2:06
  of the unique needs of these businesses across all their
• 2:08
  channels and the ways they go to market.
• 2:10
  And then as part of that, we've been involved for years now in
• 2:13
  the mock alliance. I was elected as the president
• 2:15
  of that organisation last year. And that organisation is very
• 2:18
  much focused on how, you know, companies can own their
• 2:21
  architecture and how the concepts of modern agenda, the
• 2:24
  commerce and agentic operations are going to manifest.
• 2:27
  We like to say this concept of composable, you use multiple
• 2:30
  systems to run your business. And how can you do that?
• 2:32
  Well, how can you, you said that the monolith, how can you move
• 2:35
  from, you know, break kind of free of, of the monolith?
• 2:37
  And the reality is for most modern enterprises now they're
• 2:40
  running, you know, 10s of systems to meet the needs of
• 2:43
  their business or not doing everything in one place.
• 2:45
  The same is going to be true for the agenticarer.
• 2:47
  You're not going to have one agent running in your business.
• 2:49
  You're going to have not even 10s, hundreds or thousands.
• 2:53
  And and that's, it's going to be the reality and that's going to
• 2:55
  be OK. And what we're focused on with
• 2:57
  the Mock alliance is kind of the the how, how standards and how
• 3:01
  vendors can work together to make that successful for
• 3:04
  everybody. If you could sort of help us all
• 3:07
  for a little, if we could take a look at the engineered processes
• 3:10
  around a gentic and what those key core processes look like or
• 3:15
  what they should be, what what are your thoughts?
• 3:18
  Absolutely so so I mean, if you think about where most
• 3:22
  organisations are today, you know, maybe they they they have
• 3:25
  their team, they've given everybody Co pilot or Claude or
• 3:30
  Tachi PT and and and in essence, people are using that, you know,
• 3:33
  as a better search and maybe as a bit of an accelerator.
• 3:36
  So we're most organisations are still very early days.
• 3:38
  If you look at where it's going. And if you look at where AI
• 3:42
  native start with what AI native startups do, they're building
• 3:45
  agents that are persistent and kind of scoped and secure that
• 3:49
  run all the time. And and that says organisations
• 3:52
  get more and more mature. These engineered processes that
• 3:55
  are kind of persistent and always running and kind of
• 3:57
  define to do a specific thing become very important.
• 4:00
  And I, I think what's critical there is then, you know, what
• 4:03
  you spend your time on as a business is, is kind of setting
• 4:06
  them up, getting them so they're executing successfully.
• 4:09
  But people are working on these systems instead of using them as
• 4:13
  essentially kind of a better task accelerator.
• 4:15
  So, so, so they're running even if no one's coming to the office
• 4:18
  that day, these systems are running kind of as you've scoped
• 4:21
  them. So we're seeing this emerging
• 4:23
  concepts like go to market and GTM engineering or marketing
• 4:27
  engineering, a profound launch to course last week or on the
• 4:30
  concept of marketing engineering.
• 4:32
  But I think you know consistently we're seeing this
• 4:34
  amongst our customer bases like how we know where as you get
• 4:37
  more mature, where do you start to use these these kind of
• 4:40
  standing engineer to gentic processes to take on more
• 4:44
  substantial workload on the business.
• 4:45
  And then that presents the next question, which is how do you
• 4:47
  train your team to design and build these and and and how do
• 4:52
  you rethink your architecture to support that?
• 4:54
  When you look at infrastructure and specifically agentic
• 4:56
  infrastructure, how do you think businesses have to rethink their
• 5:00
  architectures? So I think there's a few ways.
• 5:02
  So when we look at this lens of composability, one of the first
• 5:05
  things is this is early days. It's Game of Thrones right now.
• 5:08
  So you don't know which models are going to work fast.
• 5:11
  You don't, you know, last year, you know, last year or this time
• 5:14
  of year ago, you know, Claude was one of multiple options this
• 5:17
  year. They're the clear front runner,
• 5:18
  say for kind of business and coding saying next year it may
• 5:22
  be different again right. Often one year model release or
• 5:25
  feature release away from a shift in in sentiment.
• 5:28
  You look at say cursor a year ago and cursor today.
• 5:31
  Yet some of the you know, yet the announcement, you know a
• 5:34
  year from now it it it you know, they may be a back at the top.
• 5:37
  So Gabe, you think Game of Thrones, which means you've got
• 5:39
  to be careful on building everything too tightly coupled
• 5:42
  around 11 AI provider and and also recognising that many of
• 5:46
  the software vendors you have today are going to bring their
• 5:49
  own capability under the hood, right?
• 5:51
  They're building and deploying and appointing agents on your
• 5:54
  behalf that are, if you look at the stats from from, you know,
• 5:57
  from recent stage with from Google and McKinsey and also the
• 6:00
  mock alliance as well. We do an annual research report
• 6:03
  thinking about not one big agent that you just deploy at your
• 6:06
  business, but thinking about these kind of defined tasks,
• 6:09
  specific agents that you scope for appropriate security and
• 6:13
  data access that you scope for where people need to be involved
• 6:16
  in the process. And then you put those to work.
• 6:19
  That's where we're a Gentek AI is been most successful and and
• 6:23
  and can be done safely. So so the these kind of
• 6:26
  composing all these agents becomes very, very critical.
• 6:29
  I think that it what it also does is it means that you, you
• 6:32
  know, you can more rapidly change model providers, you can
• 6:35
  more rapidly change systems. You know, when you're thinking
• 6:37
  about that, that interchangeability, that setting
• 6:40
  you want on a five year basis for success.
• 6:42
  You know, we talked to too many companies right now where
• 6:45
  they've put $30 million into this broad scope AI agent that
• 6:48
  they essentially have to put in the garbage.
• 6:50
  Now. It's too tightly coupled to the
• 6:53
  model and and the AI labs or a vendor have come along and
• 6:57
  replaced it overnight. And so you really got to, you
• 7:00
  know, think very critically about where you put your
• 7:02
  investment, but also keeping that flexibility to shift models
• 7:05
  to take bits out of out of, you know, what you're doing in one
• 7:08
  system and put it in another. That's where where as a mock
• 7:10
  alliance, we think about we're encouraging that concept of
• 7:14
  composability because it gives that adaptability going forward.
• 7:17
  The last thing that that we are really trying to encourage, I
• 7:20
  think it's, it's probably 1 of the biggest things that's being
• 7:22
  underestimated in the architecture for all of this.
• 7:25
  There's new architecture layers that need to come out to allow
• 7:28
  hundreds or thousands of agency when everybody in your company
• 7:31
  can increasingly be not say that coder, but a builder, right?
• 7:34
  And, and as you get more advanced in these engineered
• 7:37
  flows, the way I like to look is you don't want policies enforced
• 7:41
  in the agents themselves. So you need, you need to rethink
• 7:44
  other parts of infrastructure as the safe and secure way that
• 7:48
  agents and people can access and take action.
• 7:51
  You know, you don't if you think of how you'd hire an employee,
• 7:54
  right, you wouldn't Great. You know, Alex, day one, here's
• 7:56
  your handbook. You're training.
• 7:58
  I'm going to give you super user access to all of our all of our
• 8:01
  systems. Please don't break anything,
• 8:04
  right? It's not what you do you, you,
• 8:06
  you slope the access of employees to what they can
• 8:09
  safely do and what they need to do to do their job.
• 8:12
  And then you also like you train them and supervise them.
• 8:15
  And they only get more access as they as they've proven
• 8:18
  capability, if anything, just to protect the employee too, right?
• 8:21
  Like they don't want to make errors.
• 8:23
  So. So when we see people developing
• 8:25
  agents often right now they're thinking about, well, you know,
• 8:28
  I'll programme in the agent, please don't delete all my
• 8:30
  orders. Please don't e-mail all of my
• 8:32
  customers. And we think that that yes, you
• 8:35
  can do that, but it needs to be enforced separately in policy
• 8:38
  too. So we're really discussing, you
• 8:40
  know, this concept of systems of action.
• 8:42
  So, yeah, where do agents route in to change orders, to issue
• 8:47
  refunds, to message customers? Those systems need to have
• 8:50
  policies enforcement too, as they do for human users.
• 8:54
  And so thinking about the the critical intersection points and
• 8:57
  are those vendors upgrading in that regard or is there an
• 9:01
  intermediary layer that you put in place in between to do that
• 9:04
  policy enforcement? Then the last part is
• 9:07
  governance. So then you know, there's
• 9:09
  vendors coming on the market like a Galileo, which was
• 9:11
  recently bought by Cisco, but there's there's others that are
• 9:13
  essentially doing kind of company wide in line enforcement
• 9:17
  of policies, but also monitoring and aggregating.
• 9:20
  So you can see how all the agents in your business are
• 9:22
  running. Are they, are they running
• 9:24
  effectively and are they running in a compliant way?
• 9:26
  So these are systems that didn't exist two years ago, but but to
• 9:30
  run agents at scale, you'll need them in your business.
• 9:33
  There's so many God I wish this conversation was like for an
• 9:36
  hour because there's so many different points I want to talk
• 9:38
  about for example QA and how the human in the loop is still part
• 9:43
  of. Even if you have loads of agents
• 9:44
  doing it, you always will need chat TPT to also whoever else
• 9:48
  the evil or to be that guard well, to be the QA, to be the
• 9:52
  human in the loop. QA is, is definitely a
• 9:54
  fascinating 1 and, and I think you know, both both QA and also
• 9:59
  development operations infrastructure, right?
• 10:01
  Like you're, you're shipping as you're shipping more code and,
• 10:04
  and over time as agents, you know, doing more and more of
• 10:07
  that, you know, it, I like to think of development operations,
• 10:10
  even going back 15 years ago and I remember it, they're behind
• 10:14
  people saying, but are we just going to automate all the
• 10:15
  developers? And then let's look forward to
• 10:17
  today. We still don't have enough
• 10:19
  developers, right, because we write so much more code.
• 10:22
  And so, and, and what we do write is often more
• 10:25
  sophisticated. And I think the same thing's
• 10:26
  going to happen again. You know, we, we use, you know,
• 10:29
  some great agentic coding or sorry, QA infrastructure.
• 10:33
  Our QA is dramatically more efficient.
• 10:35
  And you know, for better or for worse, there's never enough time
• 10:38
  for QA, right. There's no company in the world
• 10:41
  that has as many QA's as it should to be able to check
• 10:43
  absolutely everything. So you get a wide.
• 10:45
  So first you get a wider scope. You can test more robustly every
• 10:49
  time and, and so, so you know, that fills some of the
• 10:53
  efficiency gain, but you know, you almost need to actually
• 10:56
  start having agentic do the first line QA on Agentic coding
• 11:00
  is to keep pace. But also we find there's a
• 11:03
  quality improvement. It's not just speed it, you
• 11:06
  know, even in the last year, the quality of these tools has
• 11:09
  improved that they do better review.
• 11:11
  And then, then, then you also need to think about where people
• 11:14
  are in the loop and how do you upgrade also your deployment
• 11:17
  infrastructure in DevOps so that it can keep pace with all this
• 11:20
  code movement. So it kind of triggers these
• 11:22
  upgrades. But you know, across the board,
• 11:26
  I think as you rethink that system, you start to get some of
• 11:29
  those efficiency gains. You can start to get much more
• 11:31
  personalised experiences. You can start to ship far more
• 11:34
  features for customers. You can start to that kind of
• 11:37
  compete in new segments and with new abilities and that that's
• 11:40
  where you want to push to as an organisation.
• 11:43
  This all lends itself really nicely into the next point about
• 11:47
  Cowork, which is the book that you recommended that I read, and
• 11:51
  you have it right there. Oh no.
• 11:55
  I keep it on my desk because it's one of my favourite books
• 11:58
  right now. Ethan I.
• 12:00
  Mean It's a cracking book and I wish I read it two years ago
• 12:03
  because that is about hybrid work and now humans agents
• 12:07
  basically produce better results together.
• 12:10
  What how do you translate that? What does that mean to you?
• 12:13
  And like and the Mac Alliance if you like.
• 12:15
  Yeah, I think that one of the questions that I often get is
• 12:19
  how do I deploy a Gentek in my business safely?
• 12:22
  And I think the answer is with human in the loop, right.
• 12:26
  So don't just jump to agents running even if you're doing
• 12:29
  these engineered processes, you know, it's not.
• 12:32
  You think about where a human is checking or that, and then you
• 12:35
  build agents that do some of those cheques in the humans
• 12:38
  check them, but you know, you still keep an individual in the
• 12:41
  loop. And and I think that allows you
• 12:43
  to get to not only that the right end outcome, but it allows
• 12:46
  the level of kind of safety and governance in the mix too.
• 12:49
  And what I think is exciting is, you know, right now we're also
• 12:53
  in a window where, you know, it it it's making some things much
• 12:56
  more efficient or kind of actually finally giving us the
• 12:59
  time to do things the way we were always supposed to do them.
• 13:01
  We just never had had time or budget to do it.
• 13:03
  But then, you know, it's giving people new capabilities new and
• 13:07
  I think we spend a lot of time right now maybe focused on the
• 13:10
  1st efficiency bet. What I really what I my
• 13:12
  favourite quote from Ethan Moloch is, you know, something
• 13:15
  that was a competitive advantage for your business just went away
• 13:18
  overnight, right? You're, you're there's new
• 13:20
  competitors who can now replicate that and they're
• 13:23
  they're coming at you at the same time, something your
• 13:26
  customers have always wanted is now possible.
• 13:29
  And and so your job as a leader is to figure out both before
• 13:33
  your competitors do. So I think the interesting thing
• 13:35
  about Co work is thinking about, you know, where can you deploy
• 13:38
  the specialty knowledge of your team or your data assets or your
• 13:41
  market position for new advantage that AI is now made
• 13:44
  commercially feasible, you know, and so I, I think I think that's
• 13:48
  exciting both at an individual level, new, new things that we
• 13:51
  could always, you know, if you're, if we look at like what,
• 13:53
  what do our teams love doing, right?
• 13:55
  They didn't come to work to do manual tests.
• 13:57
  If you're AQA, they didn't come to work to kind of shift a
• 14:00
  wireframe. If you're a designer, liners
• 14:03
  love building products. Product designers love building
• 14:06
  products that people love yeah, we can adapt their tooling to do
• 14:09
  that They've got new ways to do that whenever possible for QA's
• 14:12
  kind of love quality and finding errors and, and, and shipping a
• 14:17
  higher quality system. But you know how they do it can
• 14:20
  ship. I think our our VP technology
• 14:22
  developers love puzzles, right. The puzzles are changing, but
• 14:25
  the puzzles are still there. There's new puzzles.
• 14:27
  So I, I think as we rethink how people do their work, new
• 14:31
  abilities they've just been granted in other domains, and
• 14:33
  then as a company, you know, extract that out to a company
• 14:36
  level and strategic advantage. I, I think it, it becomes a very
• 14:39
  exciting time. And I'm hoping as we all kind of
• 14:41
  get our head around these systems on the efficiency side
• 14:44
  that, you know, really trying to encourage as an allies, trying
• 14:47
  to think about the, these new ways of working, these new
• 14:50
  advantages as a big place that leaders can put time in 2026 and
• 14:54
  27 for the sake of their business.
• 14:56
  But also it's part of kind of bringing the team on the journey
• 14:58
  with you. People need to see what they're
• 15:00
  going towards. You know, look, we don't have
• 15:02
  many telephone switchboard operators anymore, but we have
• 15:05
  network engineer and network engineer years and always on
• 15:08
  global communications and, and I think it's worked out OK.
• 15:11
  And I think, I think that that's, you know, what part of
• 15:13
  it we can paint as a, as a leader in 2020 says.
• 15:16
  What do you think? A final question, what as
• 15:18
  leaders who are basically being exposed to so many, so many new
• 15:23
  agentic possibilities, how do you think or have you seen
• 15:27
  examples of the best ways of how leaders are sharing that with
• 15:30
  their team? So there's certainly an element
• 15:33
  of kind of demonstrating by example, right, And some of
• 15:36
  that's a topic because we, you know, we're busy day-to-day, but
• 15:40
  it is finding kind of days to carve out to actually, you know,
• 15:43
  you don't have to be like, I don't necessarily have to be
• 15:45
  like programming things to production, but but showing
• 15:48
  especially like rethinking that you're thinking about your
• 15:50
  market position, whatever your strength is that you're thinking
• 15:53
  about your market position, that you're, you know, hey, I went on
• 15:56
  Friday and I rebuilt this process.
• 15:58
  So an agent is doing it even if it's an agent in in Claude.
• 16:01
  But but I, you know, we're running this 1st and I built a
• 16:05
  heuristic the agent runs on and I'm tuning that and I'm just
• 16:08
  showing people an agent first process.
• 16:10
  So I think building our own capabilities, demonstrating that
• 16:12
  in practise to the team and and then I think providing them
• 16:17
  space, but also being, you know, one of the things we like to
• 16:19
  look at, there's a four stage maturity from Zapier, A maturity
• 16:23
  model that we really like on kind of how people build skills
• 16:26
  towards kind of more advanced sygentic use.
• 16:28
  And one of the things that we try to normalise is nobody skips
• 16:32
  to the most advanced state. Everybody, everybody goes
• 16:35
  through each step. So we can.
• 16:37
  So if you're, if you're halfway through, that's OK, Everybody's
• 16:40
  there at some point. So let's, you know, let's have
• 16:43
  the conversation around like, how do people move along faster?
• 16:46
  But I think normalising that this is a journey for nearly
• 16:49
  everybody right now. And so the focus is just how
• 16:52
  actively are you advancing, not where you are.
• 16:55
  I think that's one of the other key things that we can set.
• 16:57
  So otherwise it becomes this scary, intimidating thing as
• 17:00
  opposed to just like kind of being where you are.
• 17:03
  And let let's try to support and encourage you to reach step as
• 17:05
  quick as you can. No, again, Jason, I've loved
• 17:08
  this conversation. It's gone past so quickly and it
• 17:11
  feels like I have to invite you back at least, you know, monthly
• 17:14
  to have a conversation about AI. And you know, because I just
• 17:19
  think it's such a topical. For example, there are thousands
• 17:22
  of merchandisers around the world being told by their
• 17:25
  business to go build some sort of merchandising dashboard with
• 17:30
  some sort of agentic capability. And then you've got a whole
• 17:33
  bunch of you're saying, hey, you don't need to do that.
• 17:36
  Let us build it for you. And that whole process of
• 17:38
  building it step by step and then getting it to be business,
• 17:42
  as you were saying, you know, being able to have the guard
• 17:44
  whales that it's a business app that the business can use.
• 17:47
  I think that's the like in my mind, those merchandisers of the
• 17:51
  people sort of screaming for help right now in terms of how
• 17:54
  do they navigate this new world. And I think some of the stuff
• 17:58
  you provided today hopefully goes takes a step in that
• 18:00
  direction you. Know like it's a it's a great
• 18:03
  topic. It's a timely topic and one
• 18:05
  that's evolving so quickly as I was really glad that you're
• 18:07
  taking the time to cover it. Brilliant.
• 18:09
  Thank you, Jason. Thanks, Alex.