Tuesday, August 5th, 2025, 12:30am
| update:
Monday, August 4th, 2025, 3:40pm
Accepting uncertainty, but clearly communicated from AI about upcoming Upheavals
AI no longer quietly rewrites workflows. The alarm comes from inside the boardroom. Executives at Ford, Amazon and JP Morgan have publicly warned of deep unemployment as AI forms an operating model. Ford's Jim Farley says half of all white-collar roles are under threat. McKinsey says, “This is existential.” Andy Jassy of Amazon calls AI the next wave of confusion. JP Morgan is already accelerating large-scale automation in closed rooms. These are not predictions from think tanks, but strategic signals from people closest to the machine, and it's fair to say they have a thumbs up on scale.
While recent US labor data supports this shift, it certainly has tariffs and other Trump monomic insanity, but certainly other factors. A US government analysis released in July showed that 3% of American workers have already been replaced by AI tools, and this trend is the fastest accelerated in professional services and management. The World Economic Forum expects to eliminate up to 83 million jobs worldwide by 2030. The forecasts are different, but the momentum is clear. AI is shrinking tasks and teams just as streamlining workflows. For most of us, self-realization and UBI are not turning the corner.
So, what's next?
Executives must be honest about what is known and what is unknown. McKinsey and other analysts predict that full AI adoption across the industry could grow in the 2040s. However, most predictions underestimate the effects of compound interest in self-improvement tools. Not only are AI models faster and cheaper, they also learn how to automate new roles without human prompts. Generator AI's ability to code, summarise legal briefs, interpret medical data, and manage customer relationships is growing weekly with the troublesome “urgent ability.” When a single process is automated, it affects subsequent processes through adjacent roles. The results are exponential, not linear destruction.
We can't say for certain how many AI engineers and quick supervisors the world needs in three years. My bets are less than people would like. Planning a historic timeline is useless. The industrial revolution of the past was physical. AI is cognitive and scalable despite the large infrastructure needs. Shifts can occur within the browser window, and are so. The analogy is not electricity or printing press. AI is a billion light bulb moments where everything is happening at once, with some earthquakes, some endless knock-on effects. We have never experienced the perfect storm of infrastructure, distribution, choppy social waters, and geopolitical uncertainty.
AI is a billion light bulb moments where all capacity is happening at once, some earthquakes, some infinite with knock-on effects of gagantuan
Executives should allow generative models to introduce things that previous technologies cannot: unpredictability. AI tools can hallucinate, deceive, or manipulate outside the intended range. And that's now. What is permitted to happen in the future seems less likely to be policed if current policies are adopted. Engineers now spend as much time overseeing and modifying as they generate output. Increased productivity is real, but so is the risk of new forms. Redundancy is not only about efficiency, but also about trust, responsibility and system management. Today, many CEOs boast that once bloated companies are streamlining. These lovely people are missing out on the point. We have many poor people who can't afford their own things.
Employees are watching what's coming. Silence from leadership is a risk in itself. Even without a full plan, managers have to start talking. Start talking because it's an honest communication that allows uncertainty to build trustworthiness. With staff delaying adaptation and waiting for clarity before amplifying anxiety, workers do not expect a perfect answer, but they expect respect and context. Today's advisors (including myself) are flooded with calls to not only explain technology and possibilities, but also shape a consistent internal narrative so that no one is surprised and make some people jump.
Reconstructing the next era
Sam Altman himself warned in June that AI would pose unemployment and national security threats, and that he was not alone. Future-building executives are more vocal about social impact than many governments. Companies that dismiss these warnings as hype or worst-case stances will miss out on strategic opportunities. Companies still have time to shape how AI enhances its employees. Those who treat this shift as an issue of HR are based on those who reframe it as a strategic change. More than ever, my favorite phrase seems appropriate. Confusion is an invitation.
The reorganization process is just part of the puzzle. Three major new models are about to drop that draw the roadmap again. Openai's GPT-5 (end of August), Grok 5 from Elon (probably just dropped 4 times since December), Deepseek's R2 Reasoning Model (any day). These systems not only accelerate automation, they also trigger strategic resets. Inference, planning and real-world integration is the next frontier, and businesses waiting for certainty before responding are already behind. You have to move now.
Executives need to build adaptation strategies that assume the basis of shifts. We plan to not only increase but also remove employment. Model impact scenarios across operations, customer service, design and compliance. Not all roles disappear, but all roles change. Secondary influences can be brutal. A decrease in employment means a decrease in consumer spending. Dislocated workers bring reputation, political and economic backlash. Governments need new plans for income security, training and digital infrastructure. The current policy is likely not ready, and is not ready for a dramatic change.
Next 6-12 months
Over the next six to 12 months, companies should focus on three actions. Create inter-professional teams responsible for auditing all repeatable processes for the possibilities of automation, building internal AI flow ency across tools to include critical monitoring, and exploring known unknowns. Strategic foresight needs to be operational muscle. Leaders of law, HR, data, products and C-suite need to work together now, not to have layoffs launched.
A good start is creating a procurement kill list. Audit all current high-tech vendors and platforms for AI forecasting and long-term viability. Flag agreements that do not provide a roadmap for intelligent automation, which envisions manual workflows, or resist integration. Instead of making legacy processes a bottleneck, start replacing these vendors now.
Inserting gaps in understanding is not an option. These systems are fast, opaque and increasingly becoming agents. I got a model that lets me decide what to do next if a human was a gatekeeper in the past. We have now permission to do so, and it takes about the time until it is written by humans or machines. The problem is not just work, but the redefinition of agency and accountability within the organization. The smartest move executives can get now is to build a learning agenda, engage with the outside voices they see across the industry, and be unfairly honest about what might come next.
Confusing and exploring it is a wise move before it bumps into you, taking the invitation and treating it as a gift. All businesses can choose whether to cross prepared thresholds or get caught up in the understood of tools they don't understand yet.
Paul Armstrong Founder of TBD Group and author of disruptive technology
