Skip to content
InnovAIte

Research & development

Can governed AI increase human capability while preserving human agency?

That is the core research question, and it is genuinely open. We are building products in order to test it under real conditions rather than argue about it in the abstract — and we would rather be seen as rigorous than as finished.

The honest answer is that nobody knows yet.

There is no shortage of confident claims about what AI will do for people. There is a considerable shortage of evidence about whether sustained use of an AI support system leaves someone more capable or less capable over a period of years.

That question cannot be settled by a benchmark. It requires systems that real people use for real decisions, over enough time for the effect to show, with an honest account of what was measured. So InnovAIte is structured as a company that builds products and treats the resulting evidence as the point — not as a side effect.

This page exists to say clearly what remains unproven. A research company that only publishes its successes is not doing research.

A mix of current investigations and longer-term questions, not thirteen parallel research programmes.

Questions guiding the work.

AI-assisted living architecture

What the structural components of a governed personal support infrastructure actually are.

Human capability augmentation

Whether targeted support produces durable capability gains rather than temporary output gains.

Cognitive operating systems

What it takes for a conversational layer to behave like an OS rather than an interface.

Doctrine-driven AI behaviour

Encoding operating principles that hold under pressure, ambiguity and user disagreement.

Long-term human–AI interaction

How a relationship with a system changes over months and years, not sessions.

Outcome-learning loops

Learning from what actually happened rather than what was predicted or clicked.

Capability measurement

Measuring capability without collapsing into engagement or self-report.

Personal AI privacy and security

Minimum viable access for systems that hold genuinely sensitive context.

Human approval and auditability

Approval that is meaningful rather than a habituated confirmation click.

Dependency prevention

Detecting the point where support becomes substitution, and designing against it.

Core interoperability

Sharing context between systems without merging them into one undifferentiated store.

Multi-system orchestration

Coordinating specialised systems into one coherent response under governance.

Trust and justified reliance

When reliance on a system is warranted — and how a system should behave when it is not.

Five levels of evidence.

We distinguish between what is established, what is prototyped, what is being built, what is hypothesis and what is ambition. Anything on this site can be placed on this scale, and we will say which.
  1. Established capability

    Demonstrated repeatedly under real conditions. We currently make no claims at this level for public products.

  2. Functioning prototype

    Working software, exercised in practice, not commercially released. BuilderCore operates here as internal infrastructure.

  3. Active development

    Being built now, with a defined scope. Finance Core sits at this level.

  4. Working hypothesis

    A reasoned position we expect to revise. Most of the capability-measurement work is here.

  5. Long-term ambition

    A direction we are building towards without claiming a route to it yet. Capability stacks sit here.

Governance, privacy and trust

The principles every system is built under.

These are architectural commitments. They constrain what we are able to build, which is the point of having them.

  • Meaningful human approval for important actions
  • Data minimisation and purpose limitation
  • Clear, revocable permissions
  • Auditable actions and decision records
  • Visible uncertainty and stated confidence
  • No hidden execution
  • Privacy by design
  • Doctrine subject to evidence and revision
  • Capability growth over dependency
  • User agency preserved throughout the system

InnovAIte optimises for justified trust: competence, consistency, transparency and respect for human agency.

Status language used across this site.

Live
Available for public use.
Private Beta
Being tested with invited users.
Prototype
Functional but not commercially released.
In Development
Actively being built.
Research
A concept, framework or system under investigation.
Planned
Approved direction but not yet in active development.
Internal Infrastructure
Used by InnovAIte and not intended for public release.

Working on adjacent questions?

We are interested in research collaborations, academic partnerships and conversations with people working on human–AI interaction, governance and capability measurement.