Day 4 · AI-Ready Data Preparation

Thursday 28 May 2026 · 13:30 – 17:00 · Main venue

A joint session on what it takes to make organisational data ready for agents. The host team’s substrate meets the office-library pipeline.

Why this session matters

The five-stage memory pipeline (Pullers → Converters → Analyzers → Story-builder → Person-watcher) assumes the raw files are reachable. The data-prep layer — metadata, search, lineage, retention — is what makes that possible. Without it, the agents have nothing to read. This session is where the two halves meet.

Format

Block	Activity
13:30 – 14:15	Host team presents — data-prep work, metadata catalogue, search infrastructure
14:15 – 15:00	The five-stage office-library pipeline walked end-to-end
15:00 – 15:15	Break
15:15 – 16:15	Joint design exercise — one in-house dataset, mapped through every stage
16:15 – 17:00	Synthesis — what a joint deployment would look like

The joint design exercise

One dataset, walked through six steps:

1. Where it lives today — schema, storage, access pattern
2. Stage 1 Puller — what an agent connector would look like
3. Stage 2 Converter — what readable .md sidecar makes sense
4. Stage 3 Analyzer lenses — which three to five perspectives fit (Operator · Auditor · Forecaster · Customer, or domain-specific equivalents)
5. Stage 4 Story shape — what a daily story for this dataset would look like
6. Stage 5 Inject — which agents wake up with this in their cold-start

What we produce

• A one-page integration brief showing how the pipeline plugs into the host substrate
• A candidate first joint office — small enough to scope in eight weeks
• Material that lands in the Day 5 final presentation

Materials

• Office library + memory — The five-stage pipeline architecture this session walks.
• Day 5 ahead — Where the output of this session lands.

Next: Day 4 overview →