Forecasting & evaluation
Design questions, collect forecasts, measure calibration and compare human and machine judgement against resolved outcomes.
Human judgement + AI
Better decisions from messy information.
PK2SL designs and tests practical systems for forecasting, evaluation, matching and operational decision-making.
What we help with
- Forecasting and decision support
- Human judgement and AI evaluation
- Matching, routing and prioritisation
- Small, measurable AI pilots
What we do
Simple engagements around real decisions.
Applied AI systems
Turn fragmented workflows into bounded, testable systems with clear inputs, decision rights and evidence of what happened.
Matching & routing
Improve how people, opportunities and actions are prioritised when a thin profile or simple score loses important context.
How we work
Start small. Measure what changes.
1
Define the decision
Agree the problem, constraints and what a useful result would look like.
2
Build a bounded pilot
Use the minimum data, workflow and automation needed to test the idea safely.
3
Measure the outcome
Compare the result with the starting assumption and decide what is worth keeping.
Current work
Pukot and Sulog
Pukot
Tools for capturing human judgement, context and evidence in forms that can be tested and scored over time.
Sulog
Tools for turning that signal into useful routes, matches and next actions, then learning from the outcome.
Both are experimental and under active development.
About
PK2SL is a small UK company building practical human-AI decision systems.
We work across forecasting, evaluation, research, matching and operations. The emphasis is on measurable outcomes rather than broad claims about AI capability.
PK2SL is the public brand of pk2sl.un limited, company number 17446976.
Have a decision problem?
Send a short written brief.
Describe the decision, the available evidence and what would change if the problem were solved.