Beyond Headcount: How AI Is Changing The Way Businesses Plan Their Workforces

Historically, workforce planning has largely meant working out how many people a business needs, where they need them and when they need them. But that model is becoming harder to maintain as AI moves from experimentation into day-to-day operations.

That was one of the central questions discussed at the latest La Fosse AI Council roundtable, hosted by La Fosse Chief Marketing Officer Lucy Kemp. Bringing together senior technology, data, AI, procurement and business leaders, the discussion looked at what workforce planning should mean when an organisation’s capacity increasingly comes from a mixture of people, software and AI agents.

The conversation quickly moved beyond the familiar question of whether AI will replace jobs. Instead, the group considered a more complicated issue: if AI becomes part of the workforce, how should businesses actually plan, budget and take responsibility for it?

From Headcount To Capacity

Barry Sacks, Chief Technology Officer at La Fosse, offered a straightforward reframing of the problem: “You’re planning for capacity.”
Rather than thinking exclusively about the number of employees a business has, organisations may need to think about how much work they can actually deliver and what combination of human and AI capability is required to do it. After all, AI capability is changing much faster than traditional workforce planning cycles, so it’s only fair that our conceptualisation of the issue ought to change too.

An annual headcount plan can quickly become outdated if the tools available to a team change substantially within a few months. As Sacks discussed, workforce planning needs to follow how work is actually being performed. If agentic technology changes the amount of capacity available to a team, then the assumptions underpinning an annual plan may no longer hold.

Srimanth Rudraraju, Engineering & AI Transformation Leader, pushed the idea further, describing a model in which AI effectively becomes “another team member”, a way of looking at the issue that several people have started to shift towards in order to better make sense of things in a way that feels at least somewhat familiar. Now, that doesn’t mean that AI should simply be treated as a free replacement for a human employee. Of course, AI has its own costs, including token and runtime usage, and those costs need to be incorporated into how organisations think about capacity.

In other words, the question is no longer just how many people a business employs based on units. It’s how much capacity it has available, what that capacity costs and how it can be deployed.

Who Actually Owns An AI Agent?

Once AI starts to look less like software and more like a contributor to business processes, the questions of ownership and accountability loom large.

The council discussed whether AI agents should have owners in the same way that human employees have managers, particularly as some organisations begin giving agents defined responsibilities, KPIs and roles within workflows.

For Tia Cheang, Non-Executive Director at Investindustrial, there is an important difference between owning the technology as opposed to owning the outcome.

An AI agent working within finance, for example, should ultimately sit within the business function responsible for that work. The CFO may, therefore, be accountable for the business decision or outcome, while the technology function remains responsible for the platform on which the agent operates.

Hiten Parmar, Chief Procurement Officer at Beazley, was in a similar frame of mind, asserting that organisations may need to think beyond technical competence when designing their future workforces. As AI begins taking on more execution, characteristics like curiosity, critical thinking and human expertise that have been integral to how we’ve dealt with things in the past could become more important than ever, not less.

The implication here seems to be that AI adoption isn’t just about saying yes to new technology; at least, it shouldn’t be. Rather, if and when done properly, it ought to completely change how organisations define roles, responsibilities and capabilities across the business.

AI POCs Vs. The Real Business

One of the strongest themes to emerge from the discussion was the gap between experimenting with AI and actually deploying it at scale. Rudraraju pointed out that AI proofs of concept can behave almost like hackathons in that teams have favourable conditions, fewer constraints and can focus on demonstrating what the technology might do.

Production, however, is a very different story. Once an AI system enters the real organisation, it has to interact with procurement, engineering, operations, governance, data and existing processes. It’s possible that the technology may work perfectly well, while the surrounding organisation simply may not be prepared to support it.

That’s why Rudraraju argued against thinking of AI as simply another vertical within an organisation. In fact, AI cuts across existing processes (at least it should), meaning that its successful implementation can require changes far beyond the technology team.

Daniel Ferguson, Associate Partner at JMAN Group, made a similar point. Rather than simply bolting individual AI tools onto existing processes, businesses need to think about the broader “agentic system of an organisation”.

The challenge, therefore, goes far beyond finding another AI tool that’s helpful or may improve specific aspects of your business. Rather, it’s about working out how the organisation itself needs to change across the board.

The Data Problem Hasn’t Gone Away

If AI agents are going to make decisions, retrieve information and perform work across an organisation, the quality and accessibility of the underlying data becomes critical. But the council highlighted an all-too-familiar problem: that is, organisations often struggle to establish who actually owns that data.

Loretta Franks, Global Chief Data, Analytics & AI Officer, asserted that it’s easy for organisations to say someone owns a process or decision. But if the data feeding an AI system is inaccurate or poorly structured, the data function becomes central to whether the system can actually work.

The old “garbage in, garbage out” problem hasn’t disappeared with AI, and in some cases, AI simply makes it much more visible.

For example, an organisation might use AI to generate an email or analysis from a collection of internal documents. But if those underlying documents contain poor or inconsistent information, then no matter how good the AI model is, it can end up making the problem obvious very quickly.

Daniel Ferguson also argued that strong data foundations are becoming more important in an agentic environment as time goes on. Organisations with well-structured data platforms are better positioned to add semantic and AI layers on top of them.

However, the group also acknowledged that the idea of needing to fix data first, before dealing with everything else, also has the potential to become an excuse for delaying AI adoption indefinitely.

Thus, the challenge is finding the balance between addressing foundational problems and making practical progress.

AI Isn’t Always The Answer

Perhaps one of the more interesting threads running through the discussion was a warning against treating AI itself as the objective.

Rudraraju questioned the tendency towards “AI for AI’s sake”, particularly when pressure from boards or leadership teams creates a desire to demonstrate that an organisation is doing something with AI.

Instead, businesses should start with the outcome they’re trying to achieve. Sometimes, AI may absolutely be the right answer, but other times, traditional software, conventional machine learning or another technology may solve the problem more effectively and at lower cost.

Just because AI is part of the solution, that doesn’t mean the solution is inherently better. And the inverse is just as true.

Tia Cheang also questioned the assumption that AI automatically means cheaper operations. After all, the real calculation needs to include implementation, maintenance, token consumption, governance and AI operations costs.

As vendors are competing aggressively on price, the question of cost is becoming more important than ever. A heavily discounted AI product might look attractive initially, but businesses still need to understand what the economics look like when usage scales.

Is AI About Efficiency Or Growth?

There was also a broader question about what organisations should actually be trying to achieve with AI.

The obvious answer is often efficiency: automate tasks, reduce costs and potentially reduce headcount. But, Cheang noted that many of the organisations she works with are increasingly looking beyond efficiency and more towards growth.

Efficiency savings can be valuable, but they may be one-off. Growth opportunities can potentially compound, whether that means entering new markets, developing products more quickly or increasing the size of a sales funnel. Growth may even end up leading to efficiency, in the long run.

And that shifts the workforce conversation once again. If AI is viewed primarily as a mechanism for reducing headcount, the discussion becomes centred on how many people can be removed from a process. But, if it’s viewed as a source of additional capacity, the question becomes what the organisation could do with that capacity, which is something entirely different.

The Workforce Of The Future May Not Be Counted In People

The La Fosse AI Council discussion ultimately pointed towards a workforce model that has become more complex than traditional headcount planning. Businesses may now need to account for human expertise, AI agents, automation, token consumption, data infrastructure and the organisational processes connecting them.

That doesn’t necessarily mean that humans become less important. In fact, in many respects, the discussion suggested quite the opposite in some areas. As AI takes on more execution, human qualities like judgement, curiosity, critical thinking and domain expertise could become more important and more valuable than ever before.

The bigger change may end up being that businesses can no longer plan their workforce by looking only at the people on the payroll, and at the outset, that may certainly seem to complicate things significantly.

But as Sacks put it, the fundamental question is capacity. And as AI capability continues to change on a quarterly (or even faster) basis, businesses may need to rethink not only how many people they employ, but rather what their entire organisation is capable of doing.

Find the Beyond Headcount Blueprint here.