Assessment framework initiative
Contact

Advisory

Every education actor already has solutions in place. That is where we start.

Wiquid helps institutions integrate AI into what they already run: platforms, item banks, curricula, practices and teams. The aim is controlled change, not a new system to replace the old one.

The WASPL platform is a prototype that shows what is possible. It is our playground! You can play too!

Discuss your contextRead the research

Start from what exists

AI extends the systems that already work. It does not replace them.

Platforms
Delivery and scoring systems that work today. Their exact rules stay; AI is added around them.
Item banks
Their metadata decides how useful any AI agent can be.
Specifications
Purpose, population, curriculum, psychometric targets, delivery rules: the documents that frame each assessment.
Practices
Marking, review and validation routines, and the human rules that govern them.
Teams
Item writers, markers, psychometricians and operators. AI has to support their expertise.

Optimise what you already have

All of this is available today, with the models we already have.

  • Use AI where it is already reliable

    Content quality control, short-answer scoring, feedback drafts.

  • Keep deterministic scoring where it works

    Never replace a rule that is exact with a probability.

  • Right-size the model for each task

    The smallest model that does the job, not the most capable one.

  • Trade speed for quality in batch work

    Overnight pipelines, several passes, discounted batch processing.

  • Expose your platform’s functions as tools

    Today it saves integration work. Tomorrow it lets an AI agent drive the platform.

  • Audit continuously

    A human-marked sample, every session, on every population.

Anticipate the change

Preparing is mostly organisational work, and it can start now.

  • Write specifications that machines can read

    Purpose, population, curriculum, psychometric targets, delivery rules.

  • Invest in item-bank metadata

    The richer the metadata, the more useful every agent becomes.

  • Prepare a sovereign infrastructure

    Local or co-constructed, sized for the tasks that matter.

  • Define the human rules now

    Who validates, who arbitrates, how a score can be contested.

One signal to watch: the day an AI agent drives your platform better than a human operator, the rest follows quickly.

Six priorities to settle before delegating anything

They structure the work: first a diagnosis of what exists, then a roadmap. No task goes to AI before they are settled.

  1. Priority 1ReliabilityMake the result dependable, and prove it.
  2. Priority 2FeasibilityMix deterministic and probabilistic systems.
  3. Priority 3ControllabilityKeep humans in command.
  4. Priority 4InfrastructureDecide where the models actually run.
  5. Priority 5Agentic designSpecialised agents, one task at a time.
  6. Priority 6OrchestrationCoordinate the agents into one process.

From diagnosis to roadmap

Diagnosis
A review of existing systems, content and practices, against the six priorities.
Roadmap
A sequenced plan for integrating AI: what to optimise now, what to prepare for.
Model choice
Selecting and benchmarking AI models for your tasks: open source, sovereignty, cost.
Workshops
Training for teams, on the model of the IAEA 2026 workshop on AI-assisted correction and feedback.
Governance
Contributing to the governance of formative assessment resource banks.

For the institutions that run assessment

  • Education authorities
  • Assessment organisations
  • Platform providers
  • Schools

References you can check

Our references are public: conference talks and the research published on this site.