In February, a flyer from Pause AI at an anti-AI march in London echoed the age-old question of every startup pitch deck: ‘Step 1: Build a digital supermind. Step 2: ? Step 3: Profit.’

While Elon Musk and others cheerfully skip over step two, the reality is murkier. Studies suggest that while AI might transform workplaces, it’s not as straightforward as deploying code. For every claim of a sunny upland of economic transformation, there are sobering studies showing AI often falls short in real-world tasks.

Even top-tier models from OpenAI and Google DeepMind failed to complete 480 common workplace tasks when tested by researchers at Mercor. What’s missing is clear agreement on how AI will integrate with existing workflows without causing more problems than it solves.

The tech industry is betting big on AI, but the road ahead remains uncertain. More transparency and evidence-based evaluations are needed to bridge this gap. Until then, the world of work might just be waiting for a real solution—like someone who can actually do step two.