Imagine a car whose engine gets a new turbo every few months, while the steering, the brakes and the driver’s sense of where to go stay exactly as they were in 1990. That is roughly where many societies stand with AI and robotics. Talcott Parsons gave us a useful lens for this more than seventy years ago: his AGIL scheme. Read through it, today’s unease is less about the technology itself and more about a system whose parts no longer move at the same speed.
AGIL in two minutes
Parsons argued that every social system has to solve four problems to survive. Each problem is handled mainly by one subsystem of society, and the four are bound together by exchanges: money, power, influence and commitments flow between them.
| Function | The problem it solves | Main subsystem | Medium |
|---|---|---|---|
| Adaptation | Getting resources from the environment | Economy | Money |
| Goal attainment | Deciding what to do and mobilising for it | Politics | Power |
| Integration | Keeping the parts coordinated and loyal to each other | Law and community (the “societal community”) | Influence |
| Latency (pattern maintenance) | Passing on values, motivation and identity | Family, education, religion, culture | Value commitments |
The key point is balance. Parsons did not think the subsystems had to be equal, but they had to stay in step. When one changes much faster than the others, the exchanges between them break down, and the system experiences strain.
AI and robots: an adaptation shock
AI and robotics land squarely in the A-function. They are tools for extracting more output from fewer inputs: more code per developer, more parcels per warehouse, more diagnoses per radiologist. Firms adopt them because competition forces them to, and capital flows toward them at enormous speed.
In Parsons’ terms, the economy is innovating its way to higher adaptive capacity. That is not the problem. The problem is that the A-subsystem now changes on a cycle of months, while the other three still run on cycles of years or decades. The economy is sending signals the rest of society cannot yet read, let alone answer.
Why G, I and L are not keeping up
Goal attainment (politics). Political systems decide through elections, consultations and legislative cycles. These are deliberately slow, because legitimacy takes time. But a law drafted for one generation of models is often outdated before it applies. Politics ends up reacting to yesterday’s technology rather than setting goals for tomorrow’s. Power, the medium of the G-system, loses value when decisions keep arriving too late.
Integration (law and community). Our ways of holding society together still assume that a job is the main ticket to belonging: income, status, social security and daily contact all come through work. If automation hollows out whole occupations, the integrative glue weakens. Labour law, liability rules and social insurance were built for human workers and human decisions. Who is responsible when an autonomous system causes harm? Who pays into the pension system when a robot does the work?
Latency (education and values). This is the slowest layer, and that is by design. Schools, families and cultural institutions pass on what a “good life” and a “meaningful contribution” look like. Many of those patterns are tied to paid work and to skills that machines now do well. Education systems still train for tasks that may not exist in ten years, and our values have not yet answered a basic question: what is a person for, if not mainly for producing?
The result is a familiar kind of strain. People feel the change in their jobs (A) but see no credible plan (G), weaker safety nets (I) and no shared story about what comes next (L). In such gaps, distrust and polarisation grow.
What a more resilient system would need
The answer is not to slow down adaptation by force. It is to make the other three functions faster and more flexible:
- G: adaptive regulation that is reviewed on a fixed schedule, so politics can learn alongside the technology.
- I: belonging and security that are less tied to a single job, through portable benefits, new funding bases for social insurance and clear liability for automated decisions.
- L: education that teaches judgement, collaboration and learning itself rather than fixed tasks, and a public conversation about worth and contribution beyond paid work.
Parsons is often criticised for being too focused on stability and order, and fairly so. But his scheme makes one thing very clear: a society does not fail because its economy becomes too capable. It fails when its capacity to steer, to hold together and to make sense of change falls too far behind. AI and robots are testing exactly that gap, and closing it is a social task, not a technical one.