AI and Orchestration Start with the Work, Not Technology

At the most recent CWS Summit Europe, I had the opportunity to moderate a panel, “Getting Work Done: Orchestrating with AI.” It was one of those conversations that felt particularly timely — not because AI is new, but because the pressure to apply it is accelerating faster than many organisations are prepared for. The discussion also reinforced an important point: orchestration as a process can utilise AI, but it requires human input to be truly effective.

Through the conversation with our panelists, Brian Delle Donne, co-founder of Talent Tech Labs; Ian Blake, executive director of AI and technology transformation at Allegis Global Solutions; and Jonathan Shilton, external workforce global program office lead at Sanofi, a central theme stood out: AI and orchestration is not really a technology conversation; it is a work conversation. The real question is not how many agents an organisation can deploy, or how quickly it can adopt the latest tools, but whether it can make better decisions about how work gets done in the first place. That distinction is highly significant.

The Real Problem Is Misallocation, Not Lack of Automation

Workforce research conducted in partnership with Deployed, based on more than $1 billion in spend data, found that organisations make the wrong workforce category decision around 40% of the time.

That means businesses are routinely choosing the wrong route to get work done. They may bring in consultancies for too long, continue renewing services when the need has changed, keep roles on shore that could be off-shored or fail to automate work that should no longer be manual. This is why orchestration matters.

Orchestration is the process by which a manager’s request is translated into the right business need, in the right place. In a nutshell, AI can help identify where a hiring manager’s request may differ from the business need, but experienced people are still needed to work with the business, determine the better work option and navigate the organisation to execute it.

In other words, if you can get the allocation right at the start – i.e., identify the right type of resource, the right channel and the right operating model before execution begins – the productivity opportunity is significant. This is not marginal improvement. It is a material business issue with broad operational and financial consequences.

AI Should Improve Decisions, Not Reinforce Flawed Ones

That point connected directly with something Ian Blake said during the panel that I think will resonate with many leaders: sometimes AI is simply creating a faster mistake. 

Too often, organisations are layering AI onto workflows after a decision has already been made. In other words, the workflow may be faster, but the underlying judgement has not improved. If the initial request is flawed, the downstream process can still run efficiently — and still produce the wrong result.

That is why I believe the highest-value use of AI is not at the end of a process, but at the intake point. The moment when a hiring manager, procurement leader or business stakeholder first describes a need. That is where orchestration with AI belongs.

It should inform the decision, challenge assumptions, provide relevant context and help determine the right path before work begins. If it does not do that, the organisation risks automating the wrong answer, just more efficiently.

Start With the Work, Then Determine the Solution

Talent Tech Labs’ Brian Delle Donne expertly framed the broader market context. We are seeing an explosion of AI tools, agent platforms and vendors claiming to “agentify” every workflow imaginable. There is clear momentum, but also a lot of noise.

What I found especially valuable in Brian’s contribution was his reminder that this is now an ecosystem challenge. Infrastructure providers, bot factories, agent studios and traditional automation players are all entering the conversation at once. Without a clear orchestration model, the result is complexity rather than coherence.

That is why the idea of orchestration is so important. It is the logic that determines how tools, systems and decisions work together. It gives structure to an otherwise fragmented landscape. But even then, technology cannot lead the strategy.

The panel kept returning to the same core principle: start with the work. What needs to be done? What outcome is required? What tasks are actually involved? And only then—once the work has been properly understood—should organisations determine the best channel, worker type or solution.

Why “Moment of Need” Is Such a Useful Reframe

Jonathan Shilton shared one of the most practical ideas of the session when he described the “moment of need.” Rather than framing every demand as a recruitment issue or a resourcing issue, he suggested that managers really have a simpler problem: they need work to get done. That is the moment orchestration with AI should respond to.

I think that framing is incredibly helpful because it shifts the discussion away from process silos and back toward business needs. It encourages organisations to stop thinking in predefined categories and instead focus on what completing the work actually requires.

It also aligns with the larger point that orchestration should help businesses make better decisions about work — not simply utilise familiar channels by default.

Process Design Still Comes Before AI

If there was one warning that came through consistently across the panel, it was this: do not automate broken processes.

Many organisations are under pressure from senior leadership to move quickly on AI. That pressure is understandable, but speed without clarity is dangerous. If the underlying process is inconsistent, outdated or disconnected from how work actually happens, adding AI will not solve the problem. It will just scale it up.

Jonathan spoke candidly about the gap between documented processes and real-world behaviour. That honesty mattered, because it is a challenge many organisations recognise. What sits in a process map is often not the same as what happens in practice. And if AI is trained or deployed against the wrong version of reality, trust breaks down quickly.

This is where foundational work becomes critical: cleaning data, aligning stakeholders, defining governance and understanding how work actually gets done before introducing automation into the flow.

Why Governance and Trust Cannot Be Added Later

Another important theme from the session was trust. Ian made the point clearly that adoption is, in many ways, a trust problem. People need confidence that AI is being used responsibly, that governance is built in, and that decisions remain transparent and explainable — particularly where compliance, risk and budget are involved. I think that observation is well founded.

For AI and orchestration to succeed, organisations need more than technical capability. They need governance by design. They need human oversight where it matters most. And they need to be thoughtful about where AI should and should not be applied. That becomes even more important in talent and hiring environments, where the stakes are higher and the ethical implications are harder to ignore.

Re-envision the Work, Don’t Just Replicate It

One of the strongest takeaways from the discussion was that organisations seeing meaningful returns from AI are not simply digitising existing workflows. They are rethinking how work should be designed. That is a more difficult challenge, but it is also a more valuable one.

AI orchestration is not about inserting another layer of technology into an already crowded stack. It is about improving how organisations understand demand, classify work and route it to the best possible solution. Done well, it complements existing systems and increases the value of technology already in place. Done poorly, it adds more complexity without addressing the root issue.

That is why I believe the conversation around AI orchestration needs to stay grounded in business outcomes. If we focus only on tools, we will miss the bigger opportunity.

The First Decision Matters Most

If there is one idea I would leave leaders with after this session, it is this: the first decision is often the most important one. Before a requisition is opened, before a supplier is engaged, before a workflow begins, there is a decision about what kind of work needs to be done and how it should be solved.

That is where many organisations get it wrong. And that is exactly where AI and orchestration has the greatest potential to help. Not by replacing judgement, but by improving it. Not by speeding up every process, but by making sure the right process starts in the first place.

To me, that is what it truly means to get work done the right way, the first time.