rb2 idea sprint / focused idea

FieldPack turns a sales-stage AI hunch
into one safe buildable slice.

Not an FDE methodology. Not discovery automation. A decision artifact for the first two weeks of a forward-deployed pod.

← / → to navigate
01 / 08
Problem

AI opportunities are easy to name.
They are hard to make safe enough to build.

What happens now
A promising sales story enters delivery with missing context.
The workflow sounds clear until real exceptions, approvals, systems and ownership appear.
What we need
A narrow, evidence-backed slice the pod can build first.
Small enough for a demo. Safe enough for client trust. Useful enough to continue.
02 / 08
User

The first user is the rb2 FDE pod.
The client benefits because the pod starts from reality.

1
Sales / lead
Brings the hunch, client pressure and commercial reason to care.
2
FDE operator
Collects field evidence, spots constraints and checks adoption reality.
3
Builder
Uses the FieldPack to build the first working slice without guessing the workflow.
03 / 08
Input → output

The input is deliberately messy.
The output is deliberately small.

Input
Sales hunch
What we think AI could improve.
Evidence
Transcripts, notes, docs, tickets, screenshots
Messy material from the real work.
Map
Work Map
Actors, systems, objects, decisions, exceptions.
Filter
Confidence gaps + guardrails
What must be clarified before building.
Output
AI Intervention Slice
The first safe thing to prototype.
04 / 08
Demo in 2 minutes

Paste messy context.
Watch the tool choose a first intervention instead of writing a big plan.

1 / Brief
“Client support escalates order issues manually.”
Add call notes, a ticket export and a rough process screenshot.
2 / FieldPack
The tool maps the work and names the uncertainty.
It asks only questions that change the first build decision.
3 / Slice
“Draft escalation summary with human approval.”
A small intervention with clear boundaries, evidence and next steps.
05 / 08
Artifact

The FieldPack is not a requirements doc.
It is a build decision the pod can inspect.

Work Map
How the work actually moves.
Actors, systems, handovers, decisions, exceptions.
Slice
The first AI intervention worth prototyping.
Draft, classify, summarize, enrich, route, suggest.
Guardrails
Where humans, approvals and audit remain required.
Keeps trust and scope under control.
First 5 days
What the builder and operator do next.
Concrete actions, demo moment, stop/continue signal.
06 / 08
Reusable after Friday

The reusable part is the intake-to-slice workflow.
Not a generic planning platform.

For sales
Qualify AI hunches faster
Turn weak signals into evidence-backed options before over-promising.
For delivery
Start with less ambiguity
Give the pod a clear first build and visible risks from day one.
For clients
See value without a big programme
A narrow AI intervention is easier to understand, approve and test.
07 / 08
Decision

Pick this if we want a Friday demo
that also becomes a reusable rb2 service motion.

Build the front door:
sales hunch + field evidence → Work Map → AI Intervention Slice → first pod build.
Keep it narrow
One workflow, one slice.
Use Mastra
Workflow agents for map, gaps, risk and slice.
Avoid
Full FDE methodology theatre.
08 / 08