Concepts
AI Economics Work Automation

So-So Technology

Origin : Daron Acemoglu & Pascual Restrepo, NBER (2019)

A 'so-so' technology automates human work without creating enough new roles to compensate. Acemoglu & Restrepo's (2019) concept to distinguish technologies that enrich the economy from those that simply redistribute value toward capital.

Klarna replaced 700 employees with AI. IBM eliminated 8,000 HR positions. The productivity gains are real — but where do they go? Not to workers. Not to new jobs. To margins. This is what Acemoglu and Restrepo call a “so-so technology”: a technology that replaces without creating.

Origin

Daron Acemoglu and Pascual Restrepo, economists at MIT, introduced the concept in Automation and New Tasks (NBER, 2019).

“Automation technologies that are only marginally better than the workers they displace and do not generate enough new tasks to compensate for their labor-displacing effects.” — Acemoglu & Restrepo, 2019

The classic debate on automation assumes that technologies create as many jobs as they destroy — the “enabling” hypothesis. Acemoglu and Restrepo show this is not automatic. If the technology is “so-so,” it displaces without compensating.

Enabling vs So-So

TypeDefinitionExamples
Enabling technologyCreates new tasks only humans can doInternet → data science, UX, cybersecurity
So-so technologyReplaces tasks without creating enough new rolesSupport automation → no equivalent new positions

Key criterion: does the technology open new spaces for human activity, or does it merely close the old ones?

Application to Current AI

Some AI is enabling: it creates new roles (prompt engineering, agent orchestration, AI red teaming, specialized fine-tuning).

Much of it is so-so: it replaces support tasks, data entry, document research, standardized writing — without creating equivalent new roles in number or accessibility.

The structural risk: if the current automation wave is mostly so-so, productivity gains accumulate on the capital side without redistribution toward labor. Inequality widens even as GDP grows.

The Paradox of System Builders

Individuals who build AI tools in their organizations may believe they’re creating value (enabling) — and this is often true in the short term. But collectively, they may be contributing to a so-so infrastructure: productivity rises, positions aren’t recreated, headcount shrinks through attrition.

This is not malice. It’s local rationality whose collective result is so-so.

Indicators of a So-So Technology

Sources

Concepts