rb2 hackathon / FieldPack

The first two weeks of an FDE pod
should produce a buildable AI slice.

FieldPack is the tool that turns a sales-stage hunch and messy field evidence into a validated first intervention.

01 / 10
Why this still fits the hackathon

The hackathon should not demo a 3-month engagement.
It should demo the front door.

Too broad
“Can we run the whole FDE model?”
That becomes methodology, staffing, governance and delivery process. Too much for a hackathon.
Sharp enough
“Can we find the first safe AI intervention?”
That is concrete, demoable and directly useful before a pod starts building.
02 / 10
Where FieldPack sits

We already have a sales hunch.
The field work decides if it survives contact with reality.

Before
Early sales signal
We have a suspected opportunity: a workflow, pain point, data source, or automation angle.
Week 1–2
FieldPack phase
The FDE verifies, clarifies, constrains and selects the first safe slice.
Week 3–12
Forward deployed pod
Builder + operator work with the client to build, test, adapt, embed and decide what scales.
03 / 10
FDE job in the first two weeks

Not a discovery workshop.
A controlled way to turn uncertainty into a build path.

1
Verify
Is this really the pain, or just the sales story?
2
Clarify
How does the work actually move across people, systems and decisions?
3
Constrain
What must stay human, approved, auditable or out of scope?
4
Select
What is the smallest AI slice that would prove value?
04 / 10
Starting material

Two client people are enough for v1.
If they bring evidence from the real work.

Person 1
Knows the normal workflow.
Person 2
Knows exceptions, rework and judgement calls.
meeting recordings
transcripts
documents / SOPs
chats
emails / tickets
charts / dashboards
screenshots
rough notes
Important
They do not need to write a clean brief.
05 / 10
What the tool does

FieldPack should not ask many questions.
It should ask the questions that unblock a first build.

Input
Sales hunch + field evidence
What we think is possible, plus messy proof.
Extract
Work Map
People, systems, objects, handovers, decisions.
Test
Confidence gaps
What is unclear, risky or contradicted.
Clarify
Targeted questions
Only where the slice depends on the answer.
Select
AI Intervention Slice
Small, safe, demoable, valuable.
06 / 10
The FieldPack artifact

The output is not a requirements document.
It is a decision artifact for the pod.

Validated Work Map
How the work really happens.
Shared reality for rb2 and client.
AI Intervention Slice
What AI should draft, classify, summarize, enrich or suggest.
A first build target, not a platform vision.
Guardrails
What AI must not do; where approval and audit are needed.
Keeps the client comfortable and the scope honest.
First 5 days
Builder actions, operator checks, demo moment.
Turns the first two weeks into movement.
07 / 10
After the first two weeks

The pod still follows a process.
FieldPack makes the process start from validated reality.

Build
Prototype the slice
Turn the selected intervention into a working demo in the real workflow context.
Operate
Test with field users
Check usefulness, edge cases, failure modes and adoption friction.
Decide
Scale, adapt or stop
Use evidence from the slice to shape the rest of the 3-month engagement.
08 / 10
Hackathon build

Use Mastra for the demo.
Position Pi and Pulse as the future operating layer.

Input
Simple web UI
Paste transcripts, notes, documents, charts.
Sales hunch
The suspected opportunity we want to validate.
Workflow
Mastra FieldPack workflow
Focused agents for extraction, gap detection, risk and writing.
workMapAgent
Maps actors, systems, objects, decisions.
interventionAgent + riskAgent
Selects the slice and guardrails.
Output
FieldPack JSON
Structured enough to render, inspect and reuse.
Markdown / HTML / Mermaid
Readable artifact for client and pod.
Later: Pi / Pulse
Profiles, audit, permissions, tools, channels.
09 / 10
The clearer storyline

FieldPack is the first two weeks of FDE,
compressed into a hackathon demo.

Sales hunch + messy field evidence
→ validated Work Map → safe AI Intervention Slice → first pod build.
Hackathon scope
Demo the front door.
FDE scope
Use it to start a 3-month pod.
Avoid
Turning it into a planning system.
10 / 10