Agentic AI and the End of Organisational Functions

Organisations have forever oscillated between centralised and federated models. Despite this decades-long dance, I wonder if the balance is going to be tipped once and for all. It is my view that Agentic AI may just spell the end of the centralised function as we currently know it.

The logic around decentralisation has always been clear. We ideally want the locus of activity to be as close to the customer or frontline staff as possible. However, federated models have come with a cost. I call it the Decentralisation Tax. This tax reflects the reality that devolving activity closer to the front lines makes it more difficult to coordinate, pool expertise and drive consistency. The Decentralisation Tax can be driven by cost duplication, but could equally involve inconsistent standards or slower learning across an organisation… and for these reasons, centralised functions have to some degree made sense.

The Impact of Agentic AI on Today’s Functions

AI, and in particular agentic AI (where expertise is available on tap and actions can increasingly be automated based on that intelligence) is eroding the foundations of the Decentralisation Tax. A full-time pricing analyst or a customer insights manager may not have been required in each of three BUs and hence the roles are centralised, but if the tasks performed by a particular role are candidates for agentic AI, then why wouldn’t you have a pricing agent or a customer insights agent specific to each business area? The nuances for that area could be better accounted for, the ability to monitor BU needs and expert-driven outcomes would sit closer to the ultimate owner of the P&L, and most importantly, the cost of resource duplication – a full-time role for something that may not have been justified based on workload – is no longer relevant. Pricing and customer insights are just examples, and perhaps not even the best ones. I could run very similar lines of thinking around call centres, marketing, procurement, traditional IT and even sales.

As the cost of many tasks decreases, so too does the logic around centralisation. However, this is not only a financial consideration. Central functions have also existed to pool expertise and drive consistency. The entire concept of a Centre of Excellence is predicated on an inability to coordinate knowledge if devolved across BUs, products or geographies. Not only does this lead to lower domain prowess, but it also allows potentially costly inconsistencies (e.g. with brand or data classification). As fallible humans who take time to write emails, who may not always be the most skilled communicators, and who are influenced by parochialism for our organisational silo, decentralisation leads to sub-optimal outcomes that can be simply overcome by increasing our proximity to each other.  However, where that knowledge resides in non-human agents, the cost of coordination approaches zero. The ability to coordinate, communicate and share knowledge suddenly requires little to no marginal effort, and the logic of centralisation becomes increasingly moot.

The Functions of the Future

All of this said, I don’t believe that central functions will completely disappear. In fact, I have a view that three specific functions will survive and create a more ubiquitous functional structure even across different types of organisations. To be clear, this is very much crystal ball stuff. I won’t pretend that this is going to happen overnight. Nor do I profess to be a futurist who can see around corners. With those caveats in mind, and acknowledging that I have yet to see a consultant conclude that your current operating model serves you well and you should just keep going, let me introduce you to the concept of Offense vs Defence Functions… admittedly a somewhat unusual lens to apply.

Offense Functions

Two surviving functions in my future view of the organisation are Finance and Capabilities. Both require allocation across an entire organisation which becomes impossible if devolved into BUs.

Finance is an obvious one. The decision around the use of capital is ultimately the big strategic lever that sits with any CEO. It requires a whole-of-firm outlook and therefore needs to be independent of any one particular business unit agenda. That said, certain parts of finance may no longer make sense from a central perspective. If agents can do expense management, reconciliations, accounts payable and receivable, invoicing and many of the more process / operational elements of a finance function, then my logic would say that centralisation no longer makes sense in a world of minimised decentralisation tax. That would leave Finance as a more strategic function – focused more on top-down strategic allocation decisions than operational tasks.

The second surviving function is People, although not as we know it today. In fact, I am not using that as a label as it evokes a traditional HR lens, but rather, I am thinking about Capabilities, or even the concept of Intelligence as a Function. Intelligence could reside in humans, but in today’s world, it could equally reside in non-humans. Many pundits talk about AI as having zero cost when this is not quite the truth. I said earlier that costs can approach zero, but there is still a material cost around Agent-related tokens at scale, as well as the cost of setting up and overseeing non-human decision-making. I am sure that this will decrease over time, but ultimately, the choice between human and non-human intelligence is an overarching optimisation decision, and then an allocation exercise. Like Finance, allocating capabilities is a strategic consideration that must first be done at the whole-of-organisation level before more specific disaggregation occurs within the BU.

Both Finance and Capabilities are resources available across an entire organisation as the fuel to get stuff done, which is why I classify them as being related to offense. If I lean into the sporting analogy, we deploy both types of resources to help us kick goals.

Defence Functions

If Finance and Capabilities help kick goals, then who is ensuring we don’t get scored against? Enter the one defence related function, which in my contemplation of the future will become an amalgamation of Risk, Legal, Regulatory and Cyber. These are the types of activities that in the absence of a whole-of-organisation approach, could lead to vulnerabilities that centralised coordination can help mitigate.

This defensive function answers the question of ‘who is accountable when things go wrong’? Agents can detect, draft, monitor and remediate, but ultimately, regulators, courts, customers and boards will continue to require a human point of accountability for consequence. If agentic AI increases the attack surface area (more autonomous actions without human oversight), the need for humans with ultimate decision authority only becomes more concentrated. It is not surprising that virtually every AI regulatory framework that is emerging across the globe explicitly assigns accountability for AI to a named human role within the organisation. The direction of travel is clear – as AI scales, human accountability for AI intensifies.

The potential implications

If my navel-gazing is directionally correct, then the implications for the organisation of the future are profound. No more centralised strategy team. No more centralised sales, marketing, customer service, procurement, transformation or digital teams. The PMO dies. Technology as a function gets absorbed other than with a few exceptions where it becomes a defensive priority such as Cyber.

On the topic of technology, it would be reasonable to question the ongoing logic of consolidation where you have whole-of-organisation platforms. My view is that even this will get decentralised. If we end up with teams of agents who coordinate with each other, I can see a world where every BU has its own platform-specific agent who interacts with the platform-specific Agentic chief of staff. This is neither an allocation decision, nor an external defensive imperative. The centralisation of platform capability is simply a matter of cost and expertise, and hence no longer makes sense.

*********

Agentic AI is changing how we think about job roles and tasks, but I have yet to hear how it will change the fabric of organisational structure, in particular the role of functions. I acknowledge that my hypothesis is extreme, and I may well look foolish 10 years from now. However, organisation design has always relied on one constant, irrespective of the era, geography, industry or size of company, and that is a reliance on humans as their ultimate instruments for action. As this foundational premise gets challenged, organisations will evolve. If I am right, I don’t imagine all functional leaders will willingly accept a new reality and embrace the demise of their domain. This will be gradual. But maybe, just maybe, we are at the start of the end of the centralised to federated merry-go-round.

PS – this article was lovingly handcrafted using a pen and paper. I did use a spell checker though.