- 17 hours ago
AI is no longer just a tool. It is becoming a workforce. Yet despite widespread adoption, very few companies report meaningful P&L impact. Most remain stuck augmenting existing workflows rather than rethinking execution itself. The real breakthrough comes from reinventing how work is done. When cognitive and physical workflows are redesigned around hybrid human-AI systems, the result is not an optimized version of the current enterprise. It is a fundamentally different model: the symbiotic enterprise, with new roles, new organizational forms, and new economics. This model unlocks growth well beyond productivity. But it also erodes many traditional sources of competitive advantage, creates new dependencies, and raises a critical question: as AI capabilities become available to all, where will durable differentiation actually come from? In this keynote, Kim Baroudy, Stéphane Bout, Senior Partners at McKinsey & Company, will explore how the symbiotic enterprise reshapes operating models and competitive dynamics, what it takes to navigate the transition without destabilization, and where lasting advantage will emerge in the AI era.
Category
🤖
TechTranscript
00:00So we want to start with something that makes most companies at least
00:04uncomfortable and we realize that we are consultants and we typically see large
00:09companies but we also do see scale-ups but most of the large companies invest
00:14a lot in AI these days they invest probably hundreds of millions of
00:19dollars 80% of them actually do but they don't get anything out of it so we'll
00:24come back to the some of the numbers and statistics around this but reality is it
00:29doesn't work for them and we would argue that it's not a technology problem it's
00:35an execution problem and we'll come back to the description of what we will call
00:39this symbiotic enterprise which is basically how the cognitive and the
00:45physical AI world meets how humans agents and robots work together not to replace
00:52each other not to compete but in a new operating model
01:01these are the key messages we want you to take away from today so in all honesty AI is
01:08no longer a tool it's a workforce but very few companies have found the way to fix
01:15that they are applying AI even agents into old workflows old processes and
01:24everything that comes with that they're not reinventing how work orders and
01:30workflows are produced that also means that if they do this they will succeed and
01:38eventually they will and if not they will be taking over by AI natives and in
01:42this new world as you can see here we are arguing that the old modes or
01:48differentiation and competitiveness of companies will change expertise is no
01:54longer a mode scale is no longer a mode and of course coordinating an ecosystem is
02:00no longer a mode because agents will do this on your behalf so the winners will not
02:05be the ones that actually buy the most AI they will be whoever rewires the
02:11fastest this is our attempt to describe the last four years I'm sure you have
02:20your own version we all back in 2022 got excited when chat tpt arrived we all got
02:27excited about its ability to produce fluent sometimes useful language and but the
02:36problem is it was stateless it didn't really have anything to do with real life it
02:41didn't have any memory it didn't invent or interact or engage with tools inside
02:47organizations and we as humans had to engage and check everything so honestly it
02:53was a nice tool and the beginning of something exciting then a few years ago
02:59the agentic era came agentic is amazing we all know agents I think you all play
03:06around you all use them in your professional and private life so I hope and the
03:11challenge with that was that they were very junior so they didn't understand the
03:15local context they didn't understand the rules of the game of your
03:17organization or your personal life so it was a little bit too commoditized and
03:23generic so it behaved like what we would call a new hire or a complete junior
03:28analyst and then two things happened last year the first thing is that the LLMs or
03:37the models actually became much better significantly better so instead of doing
03:42I would say random stuff and hallucinating they actually became top of the
03:47par so on all human skills you can imagine they actually score pretty well today so
03:52that's the first thing that happened the second thing that happened was of
03:56course the invention of skills or plugins which is my favorite this is
04:01actually the difference between a smart generalists as we would say an agent would
04:06be to something really useful where you replicate or clone the best employees in
04:12your organization and if you are able to do that instead of applying an agent and
04:18then having to fix everything afterwards it will work and that basically clone the
04:24best employees when you have an organization so if the big organizations
04:28don't do this I promise you the smaller AI natives will do that that didn't keep
04:37inside the what we would call the digital world so this the PC world on screens the
04:42stuff that we do it actually moved into the world into robots and you I'm sure
04:49you know that for 30 years we had robots they were extremely useful but in a very
04:54narrow use case so you have to define a sandbox where everything is fixed you
05:00can't have too many variables of volatility around the process and in reality every
05:06time there's something that goes just slightly wrong it has to call a human for
05:10help that was actually not a mechanical problem was a cognitive problem and in
05:16the last six months I would say we've seen three new battlegrounds appearing that
05:23is changing this physical world the first world the first one is the world models or
05:28the vision language action and I'm sure you all heard about it we would say this is
05:33basically an attempt to predict and anticipate what happens around you so if
05:39you were a robot today you actually in the good old days you did something and
05:43you didn't know how it impacted and influenced the world today the world
05:48models will allow you to do that as a robot in a system that is designed for it
05:54secondly we call it embedded intelligence you can call it many things but we're
05:58basically putting brains into the machines so instead of having to loop back to
06:03the cloud and ask for how what do I do it will actually do it what we would call
06:07near real time and then thirdly which actually is an underestimated impact it's
06:14the digital twins we all talk about it but for robots this really matters because
06:18they have always been constrained by the learning experience of quote-unquote
06:23reality today with this new approach you will be constrained by software instead so
06:30basically you can put a robot into the gym as we would call it for a couple of
06:34years for whatever touches the real world and it will come out as a
06:38significantly more productive and effective system so when you put all this
06:43together it's actually pretty impressive what you can do and by the way you see
06:47some of this on the right hand side you I'm sure you've all seen pictures or even
06:51visited the Amazon warehousing the Amazon warehousing is not robots instilled or
06:59implanted in a normal warehouse it doesn't look like a warehouse 10 or 15 years ago
07:04it is actually redesigned to work together between the humans the agents
07:10and the robots and that's in a nutshell what we're gonna describe it for the
07:15future so now we have the cognitive revolution we have the physical AI
07:21revolution so I guess in theory this should all make sense so we should have a
07:27lot of impact we should have a lot of productivity uplift in the world and for
07:30companies altogether this is a very complicated consulting chart what you see
07:35on the left hand side is that that's all the tasks in Europe done without the
07:41use of physical stuff so that's give or take two-thirds of all activities done in
07:47Europe doesn't nearly need a physical intervention capability on the right hand
07:51side one-third does and then what you see here if you're good in PowerPoint you
07:57will see at the bottom of the chart basically two-thirds of the non-physical
08:01and one-third of the physical can be automated so in the perfect world with
08:06what I just described on the cognitive and the physical we can automate away 60% of
08:12activities but that's in a theoretical world it obviously takes time we also
08:17with the electricity it took 30 years before it diffused into the production
08:22plans and systems altogether industrial robots by the way the same 30 years I'm
08:27not saying this will take 30 years but the direction and end state is very clear
08:31and it's by the way still moving so this is the number I started with 80% of
08:38companies today invest in AI by the way 94% will say they invest more tomorrow than
08:43they did yesterday so everybody's doing this the problem is it doesn't work for
08:48it doesn't work in the P&O so only 6% of companies today see an impact from AI by
08:56the way that number was 4% last year so it's not really moving and and it is not
09:02because they're doing something necessarily wrong but we think that
09:05they are caught in a trap on the left hand side of this page the trap is that we
09:11have talked for years about putting humans in the loop that was a mistake
09:16because humans in the loop is a constraint it basically creates bottlenecks it
09:23creates a system where everything needs to go to the human for acknowledgement
09:27approval execution ordering and then it goes back so in other words the
09:33companies on the left hand side of this page which is somewhere between 60 and
09:3680% depending on where you are they basically apply copilots or agents to an
09:42existing system and flow and operating model that doesn't give you any impact
09:46it will give you 5-10% productivity that is all great it doesn't lead to a
09:51meaningful enterprise great upgrade we would argue that you actually need to
09:55work in the symbiotic enterprise that's the fun we'll talk about in a second this
10:01is an example everybody talks about software development and yet software
10:06development productivity doesn't go up so productivity actually is only up 10% plus
10:12minus 5 if we look at the world today there's produced 5 or 10x more code so
10:18that's not the question the question is what do you get out of it and the problem
10:22is that you basically give tools to coders and programmers that helps them
10:27writing so they will write faster but it still goes back to the human in the loop so
10:32again you create a bottleneck so what we have found and we're seeing some
10:37companies do this on the right hand side of this page is that you can actually
10:41change the operating model you can change the engine room of how you develop
10:44software and instead of having a two-week sprint still being a two-week sprint
10:48maybe minus one or two days you can change it to a one-day cycle and you will
10:54actually see significant liftoff and this goes for software development it also goes
10:59you can apply the same logic for call center we've seen telcos and banks
11:03applied to call centers that means that they will apply a new system of how you
11:08work and that works very elegantly to produce productivity as just as just
11:19illustrated by Kim individual augmentation use cases deliver only
11:25incremental productivity gains on the contrary the reinvention of an entire
11:31functional domain such as IT marketing customer customer operation unlocks the
11:37full potential of AI so what happens when workflow reinvention is scale across the
11:45whole company this is going to give rise to a new enterprise model the symbiotic
11:52enterprise in which humans agents and robots are going each to contribute
11:58according to their specific strength this is not a technology deployment the
12:05war operating model is going to change all are going to switch from producing work to
12:12setting priorities orientation and supervising processes become fluid
12:20adaptive organizational organization move from sea load multi-layer structured to
12:27dynamic network of outcome oriented teams and economics shift from labor intensive to
12:34technology technology technology the value of the symbiotic enterprise goes far beyond productivity
12:47history has shown that productivity gains are never durable competitive advantage typically they are
12:56progressively replicated by all competitors and at the end ultimately a significant share of the value
13:03created is captured by customers themselves so what does it mean you know that means that the symbiotic
13:10our enterprise is not a productivity story it's a growth and competitive advantage story it enables
13:16organization to innovate faster to continuously adapt to changing marking market condition to create a new
13:24new sources of revenue and also to scale through software rather than labor it is a powerful model but it's
13:37also change that is going to challenge erodes the traditional sources of competitive advantage notably expertise is going to
13:48to become commoditize the capabilities held by expert can be replicated by agent scale advantage diminish you know small
13:59organization can operate with a reach and productivity of large companies similarly coordination barriers for agents can
14:11orchestrate complex workflows across multiple teams and and companies and similarly also market friction
14:20decline search comparison transaction costs disappear through automation and agent-to-agent interaction that means that for
14:28incumbents the challenge is not only to smartly adopt AI it's also to find a way to defend against the
14:36erosion of
14:37traditional sources of competitive advantage while building new forms of differentiation but this is not the only
14:45challenge in addition a new form of strategic dependency on AI infrastructure provider is emerging as AI becomes
14:56embedded in a company core business process and the critical activities you know organizations are exposed to pricing
15:04power to vendor looking to operational vulnerabilities what recalls a cognitive tax an increasing share of enterprise value could be
15:14transferred to infrastructure provider as AI becomes a commodity as AI models commoditize
15:24the key question becomes how can we differentiate we see free new sources of competitive advantage first the ability to
15:32build
15:32proprietary intelligence by combining company specific data company specific expertise encoded into agentic skills by customizing the models and
15:44self and by setting up learning loops that will ensure continuous enrichment of this proprietary intelligence layer
15:53secondly the ability to control strategic points within customer workflows and within ecosystem and finally the ability to design
16:05orchestrate continuously optimize large scale system combining humans agents and then robots
16:14the destination is the destination is clear but the journey itself is far less obvious you know with two failure
16:21modes
16:22to avoid on one end incrementalism move to sleep too quick sorry move to slowly and you risk you know
16:32optimizing your pre AI operating
16:34model while AI native competitors are redefining the economics of your industry move to aggressively and you risk
16:45overwhelming your organization or facing technology crashes we do think that four condition must be met to successfully
16:54navigate this transformation dilemma first establish a clear vision of the future business model grounded in a future sources of
17:04revenue
17:04and differentiation and not only on simply on productivity improvement this vision this vision will guide your investment
17:13decision with guide organization redesign as well as workforce transformation
17:18secondly develop a balance transformation road map conducting in parallel the progressive reinvention of the company
17:27of the company domain per domain while deploying some individual augmentation use cases to ensure organizational
17:33learning the pace of the transformation itself should balance early value realisation technology
17:39technology stabilization and workforce transformation scalable foundation in particular
17:45a modular vendor agnostic AI foundation that will allow to scale without being locked in with a specific vendor
17:54and finally an expanded leadership coalition this is a strategic transformation it has to be led by the CEO
18:02but the CHRO a chief transformation officer and the CEO must also be involved
18:11if you want to know more about the symbiotic enterprise you can download the white paper on our website
18:19you have a QR code here that will allow you to access to this white paper
18:23thanks a lot for your attention
18:24thanks a lot for your attention
Comments