Social impact
If the problem is institutional, so is the impact. We are not selling a widget that “makes society happy”. We install the layer that lets organizations that already serve the public deploy agents without giving up identity, provenance, judgement, and a human gate.
Who gains when the institution can answer
| When the institution… | Society gets… |
|---|---|
| Can show who produced a deliverable and from which corpus | Fewer opaque decisions and fewer black boxes in processes that affect rights |
| Keeps human oversight with the power to interrupt | Less uncorrectable harm and less automation bias with no recourse |
| Isolates corpus and identity per client or department | Less cross-leakage of sensitive data across purposes |
| Evaluates before promoting a skill or a behaviour change | Less systemic error propagation at scale |
These are not marketing promises: they are conditions the design requires so a deployment is diligence, not theatre. What is already measured is in What is live.
Beyond a single buyer
The exoharness is built to federate across institutions: published skills, shareable facts with provenance, handoffs with an audited envelope — the same trust vocabulary models like X-Road use between administrations. When two bodies can exchange agent work without losing the thread of authority, the benefit stops being internal only: inter-institutional cooperation without centralizing all data in one operator.
That matters where the bottleneck is institutional capacity, not lack of models: justice, regulation, administrative health, complex citizen-facing services.
What we do not claim here
This dossier does not include citizen adoption figures, verified budget savings, or stories from a specific deployment. Those proofs belong to a design partner or a measured pilot, not a public draft.
Social impact is contingent: it depends on the deployer using governance instead of hiding it; on human gates staying open; and on improvements being evaluated before they spread. That is why Limits names social risks with the same seriousness as Governance names the gates.
One line to close
Giving institutions a workforce they can hold accountable widens regulated public and private service without normalizing algorithmic irresponsibility.