Building Talent Acquisition From Scratch: What the AI Era Changes (and What It Doesn't)

I had a conversation recently with a senior HR leader inside one of the world's largest creative holding groups. He had approval. He had the mandate. He had been fighting the same battle for three years.

He wanted to build an internal talent acquisition function from scratch.

And the closer we got to talking about how, the clearer it became that the real problem was not the technology. It never is.

The Vendor Problem Nobody Wants to Name

Large organisations do not build internal recruitment functions because external vendors are convenient. That is the polite version.

The less polite version: external vendors fill a gap that no one wants to own internally, and after a few years, they become load-bearing walls. Difficult to remove. Expensive to maintain. And frustrating, because the data they hold is not always yours.

This particular leader had been running recruitment through an outsourced model across multiple agencies. Different platforms per entity. Fragmented candidate data. Systems that never spoke to each other. And a provider relationship that had become a dependency rather than a partnership.

The decision to build something internal had finally been approved. The harder question was now in front of him: how do you build a recruitment operating system that actually works, in a large multi-brand group, in 2026, when everything is changing?

The Three-Tier Model Worth Building

The instinct in most organisations is to either go fully internal or fully external. Both are wrong.

The structure that makes sense at scale, particularly inside a creative holding group with diverse talent needs across markets, is a deliberate three-tier model.

The first tier is an internal team of midweight recruiters. Not senior strategists. Not junior coordinators. People who can run volume, manage pipelines, handle the operational grunt work of account management-level hiring day to day. This is the core machine. Without it, nothing else holds.

The second tier is a small group of strategic preferred partners. Not a panel. A genuine, curated shortlist of specialists with deep knowledge of the industry, specific markets, or specific talent communities. People you trust. People who know your organisation well enough to represent it accurately in the market. This tier is not for efficiency. It is for quality, reach, and the kind of search work that requires relationship depth.

The third tier is executive search, applied selectively. Senior appointments, confidential mandates, regional leadership roles. Not every hire needs a search firm. But the ones that do, need one that operates at the right level with the right network.

The model only works if the boundaries between tiers are clear. When they are not, you get duplication, budget leakage, and frustrated suppliers who do not know what they are actually being asked to do.

What AI Actually Changes

This is where the conversation shifted, and it shifted fast.

The traditional argument for not building an internal TA function was always about capability: you cannot compete with an external firm's research capacity, candidate database, or market mapping infrastructure. That argument is gone.

Tools like Clay have fundamentally changed what a small internal team can do. An exec search that previously required weeks of desk research can now be initiated in hours. AI systems can take a detailed job brief, identify candidate pools across multiple channels, and return enriched profiles including background, contact details, and relevant career context. Not a shortlist. A researched, structured, candidate intelligence report.

The output I have seen from a well-constructed AI search surprised me. Not because the technology was perfect, but because the gap between what required a five-person research team and what one person with the right tools can now produce has narrowed dramatically.

You do not need a researcher anymore. That role, as it existed ten years ago, is gone.

What you do need is someone who can prompt well, think structurally, and build a process around the output. The technical capability is accessible. The discipline to use it properly is still rare.

The Platform Question Nobody Has Fully Answered

The technology conversation inside large holding groups is almost always the same. There is a global HRIS rollout coming, usually Workday. There is a freelance platform that someone championed and no one has fully implemented. There are agency-specific ATS tools that were inherited and never rationalised. And there is a CRM for candidates that lives in spreadsheets or in the vendor's system, not yours.

Building an internal TA function requires cutting through this. Here is what is actually worth separating.

Workday as an HRIS is fine for what it is. As an ATS, it is built for large corporate infrastructure, not for recruitment agility. A separate, purpose-built ATS running alongside it is a legitimate and sensible choice. The options worth evaluating, particularly for a creative services environment, are Ashby, Greenhouse, and Luxo. Each has trade-offs. What matters is picking one and committing to it, not waiting for global IT to solve the problem in three years.

For freelance and production hiring, a dedicated platform is not a nice-to-have. The volume and speed of freelance requirements in a creative group is simply incompatible with a generalist ATS workflow. Having a freelance platform that feeds into rather than replaces the core HRIS is the right architecture.

The candidate data question is the one that bites organisations hardest. If your recruitment vendor owns your candidate records, you are not running a recruitment function. You are renting one. Building internal means owning the data, which means the first step, before anything else, is defining what data you have, where it lives, and how you are getting it back.

The Operating System Problem Is Bigger Than the Tech Problem

Here is what I see every time a large organisation tries to build something new: they underinvest in process and overinvest in the announcement.

They approve the headcount, buy the tools, and then six months later wonder why it is not working. The answer is almost always the same. There was no operating system. No clear roles. No KPIs tied to actual outcomes. No swim lanes between internal team, preferred partners, and search firms. Everyone is doing a version of the same thing, stepping on each other, and the hiring managers in the business have no idea who to call.

The question I always ask is simple: if your three internal recruiters, two preferred partners, and one search firm all received the same mandate today, what would each of them do with it? If the answer is unclear, the operating system does not exist.

Getting the operating system right before scaling the team is not conservative thinking. It is the only way scaling works.

That means job descriptions. Real ones, not aspirational ones. It means KPIs that sit above everything as the north star, so that every tool, every hire, and every preferred partner relationship is measured against outcomes that actually matter to the business. And it means a clear decision-making framework: which roles go internal, which go to a preferred partner, and which go to a retained search firm, and why.

Transition Is the Hardest Part

Moving from a vendor-dependent model to an internal one is not a clean switch. That is the part most leaders underestimate.

You cannot cut the external vendor until you have something to replace it with. You cannot build the internal function at full capacity before the vendor is cut. You are managing both simultaneously, which is expensive and uncomfortable, and the instinct to accelerate one or the other is usually wrong.

The discipline is in the sequencing. Wind down incrementally as internal capacity builds. Keep the vendor in place for volume or for markets where the internal team is not yet operational. Do not compromise the hiring outcomes for the business while the transition is underway.

This is one of the reasons most internal TA build-outs stall. The organisation gets approval for the strategy but not for the transition period. It looks expensive on paper to be running both. It is more expensive when the internal function launches half-built and the business loses confidence in it before it has had a chance to work.

What Has Not Changed

The tools are better. The data is more accessible. The research that once required a team can now be done by one person with the right subscriptions and the right discipline.

None of that changes what actually determines whether a talent acquisition function works: the quality of the relationships it builds, and the speed at which the business trusts it.

Trust is built through consistency. Through knowing that when a hiring manager brings a brief, it gets treated properly. Through candidates having an experience of the organisation that reflects well, even if they do not get the role. Through the internal team being knowledgeable enough about the business that they can have an intelligent conversation with both sides of a hire.

Technology accelerates a function that is already working. It cannot substitute for one that has not been designed properly.

Build the operating system first. Get the team structure clear. Define the boundaries between tiers. Own the data. Then use AI to do what used to require a research operation.

In that order.