The rapid diffusion of artificial intelligence is generating a massive wave of research measuring and forecasting its impacts on global labour markets. But the track record of past economic approaches gives reason for humility.
We have seen these predictions before with job offshorability and the introduction of industrial robots. Often, the direst predictions fail to materialise. Yet, the latest data from Anthropic suggests something fundamentally different is happening this time.
This is not a distant science fiction scenario. It is a present reality creeping into our project delivery structures. In our view, the project delivery sector is sitting at an awkward intersection. We generate enormous volumes of unstructured data, from requests for information (RFIs) and meeting minutes to variation orders and cost reports. Historically, this data has been too expensive and too sensitive to process comprehensively with AI.
But as it reveals, the gap between what AI can do and what it is doing is closing fast.
The Gap Between Theory and Reality
Anthropic recently introduced a new measure of AI displacement risk called 'observed exposure'
This metric combines theoretical large language model (LLM) capability with real-world usage data. The findings are stark and highly relevant to anyone managing complex knowledge work. AI is far from reaching its theoretical capability; actual coverage remains a fraction of what is feasible
For instance, their data shows that 97% of the tasks observed across their previous four Economic Index reports fall into categories rated as theoretically feasible by researchers
Yet, the actual application is lagging. Why? Some tasks are slow to diffuse due to legal constraints, specific software requirements, or human verification steps. In project delivery, we see this constantly. We know an LLM could theoretically draft a preliminary risk register, but internal compliance rules and the lack of integrated software tools prevent it from happening at scale.
But this lag is temporary. As Amodei notes, the timeline for disruption is shrinking.
The memo sparked debate about the 'Ghost GDP trap'—a scenario where machines boost output but spend nothing, potentially destabilising the broader economy
This macroeconomic theory has direct microeconomic implications for how we staff and run large infrastructure projects.
The Impact on Project Delivery Professionals
What does this mean for those of us managing multi-million-pound construction projects? It means the administrative heavy lifting that forms the backbone of entry-level project management roles is highly exposed.
According to Anthropic's study, occupations with higher observed exposure are projected by the Bureau of Labor Statistics to grow less through 2034
The most exposed occupations include:
Computer Programmers (75% coverage)
Customer Service Representatives (67% coverage)
Data Entry Keyers (67% coverage)
In the context of project delivery, think about the roles focused on data entry, basic scheduling, and initial document review. These are the tasks ripe for automation. We feel this isn't necessarily a negative development.
By automating the mundane, we free up human capital for complex problem-solving, stakeholder management, and strategic foresight. These are areas where human nuance remains irreplaceable. The role of the junior project manager will shift from being a data gatherer to being an AI orchestrator.
The Demographics of Disruption
Interestingly, the demographic profile of those most exposed to AI disruption challenges our traditional assumptions about technological displacement. Workers in the most exposed professions are more likely to be older, female, more educated, and higher-paid
This is a profound shift. Previous waves of automation primarily impacted blue-collar or lower-wage roles. AI is coming for the knowledge workers. However, Anthropic's research notes that there has been no systematic increase in unemployment for highly exposed workers since late 2022, though there is suggestive evidence that hiring of younger workers has slowed in exposed occupations
This aligns with what we observe in the industry. Firms aren't necessarily firing their junior staff en masse; rather, they are rethinking their hiring strategies and the skill sets required for new entrants. The focus is shifting from raw data processing capabilities to prompt engineering, AI orchestration, and critical thinking. If a machine can process the data, the human value lies in asking the machine the right questions.
The sensitivity of commercial data, subcontractor pricing, and preliminary estimates has kept many firms cautious about adopting cloud-based AI. But as models become more secure and local inference becomes viable, these blockers are weakening. Project intelligence is stopping being a reporting function and is becoming a real-time operating layer.
"The pace of progress in AI is much faster than for previous technological revolutions. It is hard for people to adapt to this pace of change, both to the changes in how a given job works and in the need to switch to new jobs."
— Dario Amodei, CEO of Anthropic
These quotes should serve as both a warning and an opportunity. The technology is capable of much more than we are currently asking of it. The firms and individuals who figure out how to bridge that gap first will hold a significant competitive advantage. We are currently in a window of opportunity where the tools are available, but widespread adoption hasn't yet occurred.
As we integrate these systems, we must fundamentally rethink how we value human capital on our projects. If the baseline tasks are automated, how do we train the next generation of senior leaders? Historically, project managers learned the ropes by doing the grunt work. If the grunt work disappears, we must create new pathways for experiential learning.
This might involve simulation-based training, where junior staff manage virtual projects against AI adversaries. It might involve shadowing senior leaders much earlier in their careers. Whatever the solution, the old model of progression is broken. We must actively design the new one.
Furthermore, we must address the psychological impact of this transition. The fear of redundancy is real and justified. Leaders must communicate transparently about how AI will be deployed, not as a replacement, but as an augmentation tool. We need to foster a culture of continuous learning where adapting to new AI tools is rewarded and celebrated.
Avoid falling behind in the AI shift. Subscribe to the Project Flux newsletter and podcast for weekly insights on navigating the future of project delivery.
Takeaways
From our lens, the following three areas demand immediate attention:
Conduct workflow audits: Identify tasks that are highly repetitive and data-intensive. These are your prime candidates for AI integration. Look closely at your RFI processes, your monthly reporting cycles, and your document control systems.
Upskill the workforce: If 50% of entry-level tasks are automated, the entry-level role itself must evolve. Junior staff need to be trained not just in traditional project management methodologies but also in how to manage and interrogate AI systems. They need to understand the limitations and biases of the models they are using.
Monitor economic implications: The 'Ghost GDP trap' is a valid concern. As we drive efficiencies through AI, we must also consider how value is distributed within our organisations and the wider economy. We must ensure that the productivity gains from AI translate into better project outcomes, higher wages for skilled workers, and a more resilient industry.
Links and Stuff
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
/
