The Blended Workforce: Why Agentic AI Is Not Just Another Tool

By Arjen van Berkum

Rethinking the Paradigm

There is an almost standard pattern in how organisations respond to technological discontinuity. A new technology emerges, early adopters integrate it into existing workflows, leadership issues a press release declaring commitment to innovation, and then the organisation waits to see what happens next. This pattern has repeated itself with the internet, with cloud computing, with mobile platforms. It is repeating itself now with artificial intelligence, and for most organisations it will produce the same outcome it always has: late adoption, competitive disadvantage, and a scramble to catch up.

The difference this time is that the gap between those who understand the shift and those who do not will be catastrophic. Agentic AI is not another tool to be absorbed into an existing workflow or process. It is a fundamental reorientation of how work is structured, how information is gathered and processed, how decisions are made, and ultimately how leadership functions. To treat it as anything less is to misread the nature of the transformation entirely and will lead to abysmal failliure.

The Architecture of the Blended Workforce

The concept of the blended workforce is already out there, but its implications are rarely examined with the attention they demand. A blended workforce is not simply a team that uses AI-powered software, although I see that a lot…. It is a structural reorganisation of labour in which human cognitive capacity and machine processing capacity are allocated according to comparative advantage of each. Humans will increasingly own the domains of oversight, ethical judgement, relationship management, and strategic direction. These tasks are not easy to automate, the real question is however should they be... The difficulty is precisely what drives human value. The responsibility, the ambiguity, the relational complexity and the moral weight of consequential decisions are features of human work, not inefficiencies to be optimized away. But it does require reskilling.

What shifts is everything adjacent to those domains. The gathering of intelligence, the processing of data, the execution of micro-tasks, the monitoring of systems, the aggregation of signals from disparate sources and the generation of actionable summaries all move into the domain of the machine. And critically, not into a single monolithic AI system, but into swarms of micro-bots, each performing a discrete, well-defined task, orchestrated together to produce composite outputs of significant intelligence and value. This is the architectural reality that most organisations have not yet internalised.

Micro-Tasks, Orchestration and the End of the API

The implications for systems architects are profound and underappreciated. Traditionally, enterprise technology design has been dominated by the logic of the platform: identify functional requirements, select a platform, integrate it with adjacent systems through defined APIs, and manage the resulting complexity. This logic is being superseded. Agentic AI systems do not route information through predefined endpoints. They navigate, interrogate, and synthesise information across systems dynamically, without explicit integration logic for every possible connection. The micro-bot does not need a pre-built integration. It needs a task, a set of rules, access to relevant environments, and the capacity to reason about how to fulfil its objective.

For architects, this represents a conceptual shift. The question is no longer simply what the landscape looks like and how the systems connect. The question becomes: what are the micro-tasks that constitute our total operational activity, and which of those are candidates for agentic automation? Answering this requires decomposing complex processes into their atomic constituent actions, assessing which require human judgement and which do not, and designing orchestration logic that governs how agents collaborate, sequence activities, and escalate when necessary. The end of the API as the primary integration mechanism is not science fiction. It is, in emerging form, already here. It is also the end of the architects that are technology lovers, however it is the rise of the process owner.

The Platform of Platforms Era and Contract Management

We are living in the age of the platform of platforms. The most sophisticated enterprise adopters have already moved past the question of which individual system to use. They are building meta-architectures, layers of orchestration that sit above existing systems and coordinate activity across them through intelligent agents rather than rigid integration code. The speed differential between organisations operating with this architecture and those still locked in the traditional model is already visible and will become decisive.

Contract management illustrates the stakes clearly. The post-award management of contracts, the discipline concerned with ensuring that what was agreed is actually realised, has historically suffered from fragmentation and information deficit. It is now shifting towards “did we realize what we wanted to realize”. From buying or selling to meeting objectives. A typical senior contracts manager operates in an environment where information about performance, supplier behaviour, risk, change events and compliance is scattered across CLM platforms, ERP systems, supplier portals, communication tools, financial systems, and spreadsheets. Synthesising this into a coherent picture requires significant manual effort, is invariably out of date, and is rarely available at the speed required for proactive decision-making. A swarm of micro-bots continuously processing information across that ecosystem produces instead a living, real-time intelligence picture. It surfaces anomalies, flags risk concentrations, identifies emerging disputes before they escalate, and monitors performance against commitments without waiting for a quarterly review. The conversation about who is to blame, a retrospective exercise in attributing fault after value has already been lost, becomes unnecessary because the conditions that give rise to it are detected and addressed earlier (finally we can go really pro-active). And critically, the backend systems do not need to change. The legacy ERP, the incumbent CLM, the procurement platform acquired five years ago can remain. Agentic AI acts as an intelligent intermediary, funnelling information from existing systems rather than replacing them.

The Employee Experience

There is a dimension of this transformation that is easily underestimated: the experience of the individual employee. Consider the administrative friction embedded in the daily working life of a contracts professional. Document retrieval, status updates, approval workflows, data entry across multiple systems, version control, stakeholder notifications: the cumulative burden is extraordinary. The PDF that must be printed, completed by hand, scanned, and emailed back is not merely an inconvenience. It is a symbol of how far the current architecture of work is from what it could be. Agentic AI replaces this friction with an interaction model in which the employee communicates intent and the machine executes, leaving human cognitive energy for the judgements, relationships, and strategic thinking that genuinely require it. Employees gain clarity: not a vague sense of what they probably need to do, but a precise, contextualised understanding of where their attention adds the most value. This will require significant reskilling of the current workforce. 

The Risks That Demand Honest Accounting

It would be not very cool to present this vision without confronting the risks, and those risks are very substantial. The most fundamental is process quality. Agentic systems execute against the logic of the processes they are built on. Poorly designed, incomplete, or internally contradictory processes are not corrected by AI; they are amplified and propagated rapidly at scale. Organisations that deploy agents without first achieving process clarity will not merely fail to realise benefits. They will generate new categories of operational risk. And lets be honest, who has its processes really in order?

Data integrity is of fundamental importance. Corrupt, outdated, incomplete or biased data produces corrupted, outdated, incomplete and biased outputs. A senior leader relying on an AI-generated risk dashboard built on poor data is making decisions based on a sophisticated confabulation. The third risk concerns cost sustainability. The economics of AI appear compelling today, but the trajectory is not guaranteed. Emerging tokenisation of inference capacity, combined with the already significant environmental costs of large-scale AI computation in terms of both energy and water consumption, introduces real uncertainty about long-term cost structures and ethical costs. Organisations committing to agentic architectures must model scenarios in which AI compute costs are substantially higher than they are today, and build business cases that are robust under those assumptions.

The Skills That Actually Matter

There is a widespread assumption that the critical future skill in an AI-intensive environment is the ability to prompt effectively. This is insufficient. Prompting is a technique. What is needed is understanding. The professionals who will thrive in blended workforces are those who understand the journeys (and thus processes) that key stakeholders undertake: the customer journey, the supplier journey, the employee journey. They understand where those journeys encounter friction, where they are vulnerable to disruption, where system shocks originate, and where dependencies create fragility. They understand how to design audit mechanisms, checks and balances, and governance structures for systems in which much of the execution is automated and how to embed human oversight as a genuine control point rather than a formality.

Regulatory pressure will grow. Legislation governing AI in consequential decision-making is a matter of when, not whether. It will require organisations to reproduce the reasoning behind AI-assisted decisions, demonstrate that human oversight was substantive, and retain the evidence required for audit. The data storage implications alone are substantial and largely unplanned for. The future professional in contract management or procurement is not a prompt engineer. They are a systems thinker who understands human and machine behaviour equally well, can design for resilience, and can distinguish between a process that looks efficient and one that is genuinely robust.

Augmented Intelligence and the Imperative to Act

The framing that best captures what is at stake is the distinction between Artificial Intelligence and Augmented Intelligence. Artificial Intelligence, as a concept, positions the machine as the primary agent and the human as an observer. Augmented Intelligence inverts this: the human remains the primary agent, the bearer of purpose, judgement and accountability, and the machine extends the scope and quality of what that human can achieve. This distinction is not just semantic. It shapes how agents are designed, how people are prepared, and how success is measured. An organisation building for Augmented Intelligence measures not how many tasks have been automated but how much better its people are able to do what only people can do.

The final question is whether senior leadership will engage with this agenda with the urgency it requires or whether they will wait. Waiting is a choice with consequences. The organisations already building agentic architectures, already decomposing their processes into micro-tasks, already thinking in platform-of-platforms terms, are acquiring capabilities that compound over time. A press release announcing that the organisation will do AI is not a strategy. A proof-of-concept that never scales is not a transformation. What is required is the mindset of the startup applied to the resources of the enterprise: dream big, start small, act fast, and learn from both the failures and the successes. The blended workforce is not a future state. It is an something that is starting to spring to life now, and the window in which early action creates durable advantage is here....

Arjen van Berkum is the contract management wizard and entrepreneur, speaker, teacher and conceptually creates work for CATS CM, specializing in post-award contract management best practices, organisational design, and the integration of intelligent systems into complex commercial environments.