Learning Analytics Agent
An agent that watches for risk, not a dashboard that waits to be read.
This one decides what to do about it while it still matters: reading learner activity, scoring risk in real time, and assigning an AI tutor session automatically.
Client
Role
Sole builder, brief to handover
Scope
xAPI schema · agent · dashboard
Stack
xAPI · Rules engine · AI tutor · Database · Dashboard
LEARNERS TRACKED
Learner activity, read as it happens.
RISK TIERS
Medium risk gets a nudge; high risk gets a nudge plus an AI tutor session.
AGENT STAGES
Signals into a decision.
XAPI COMPLIANT
Built on standard xAPI, so it fits the platform a team already runs.
dots | tiers | chain:Read,Score,Decide,Act | grid
How it’s built
Point at a step
Signals into a decision.
Read
pulse
1
xAPI activity
Learner activity, read as it happens.
Score
nodes
2
Four signals
Stalled progress, a failed quiz, low tutor confidence, skipped practice.
Score
diamond
3
Rules engine
Two or more signals means high risk.
Act
bell
4
Medium risk
Medium risk gets a nudge.
Act
persona
5
High risk
A nudge plus an auto-assigned session with the AI tutor.
Act
db
6
Audit trail
Every decision lands in the database with its reasoning attached.
01
/ 01 · Every reply, live
The rep
Practises the pushback their next call is likely to bring.
Process
Discover · Scope · Build · Prove
01
Discover
Finding out too late
Every platform can tell you who dropped off, after the fact. By then the learner has gone quiet and the manager hears about it last.
02
Scope
Act while it still matters
Read learner activity as it happens, score risk, and intervene before a human has to ask. Built on standard xAPI so it fits the platform a team already runs, with no rebuild.
03
Build
Signals into a decision
Four signals feed a rules engine: stalled progress, a failed quiz, low tutor confidence, skipped practice. Two or more means high risk. Medium risk gets a nudge; high risk gets a nudge plus an auto-assigned session with the AI tutor.
04
Prove
An audit trail, not just an output
Every decision lands in the database with its reasoning attached. You can see why the agent did what it did, which is what makes an automated intervention defensible.
More work
All work
