Danielle Beram

Case study · Sandbox VR · Marketing

AI enablement: capability that stays in the department

Build the builder.

Problem
Manual marketing workflows consumed capacity, and every automation request routed back through me.
Intervention
A two-week applied cohort where two Marketing Super Users mapped, built, tested, calibrated, and shipped their own AI-enabled workflows.
Result
Roughly 16 hours a month returned without openings and roughly 32 with one to two, plus two local builders who can build the next one.

Overview

What this was

AI could clearly speed up a marketing workflow. The open question was whether the people closest to the work could identify, build, validate, operate, and improve those workflows themselves.

  • Designed the cohort sequence and the work-based assessment
  • Facilitated live instruction on workflow decomposition, context design, and Claude Projects
  • Coached the builds between sessions and ran calibration
  • Handed ownership, maintenance, and the executive presentation to the Super Users

The system

Diagnosis, build, and how it ran

The constraint

A marketing team was spending recurring hours on assembly work: the same collection, formatting, and drafting steps every cycle, and more of them in months with an opening. Automating one of those workflows myself would have solved one workflow and created a permanent dependency on me for the next one.

Success meant two people could explain, operate, maintain, and extend the workflow themselves.

The sequence

The capability journey

Select a stage to see what it had to produce.

The work was the assessment

The final session tested four capabilities, each against something the participant had actually built:

  • Diagnose: what should be automated, and what should stay human?
  • Build: can they create and refine the workflow?
  • Validate: can they find failure cases and tell plausible output from usable output?
  • Own: can they run, maintain, teach, and extend it independently?
  • 16 hrs/mo

    Capacity returned, months without openings

  • 32 hrs/mo

    Capacity returned, months with 1–2 openings

  • 2

    Local builders maintaining and extending the work

  • 2 weeks

    Pre-work through handoff

I report the two operating conditions separately as hours of capacity returned, using the participant-tracked time for each condition.

What was actually transferred

The scale loop

  1. Spot
  2. Map
  3. Build
  4. Ship
  5. Teach
  6. repeats inside the department

The loop runs inside the department. The people who own the work spot, map, and build the next candidate workflow, and the Super Users teach the next person.

A successful pilot ends with less dependence on the person who launched it.

The durable result was a department with the capability to build the next workflow itself.

Measurement

What changed

  • 16 hrs/mo

    Capacity returned in months without new-store openings

  • 32 hrs/mo

    Capacity returned in months with one to two openings

  • 2

    Local builders able to maintain and extend the workflows

  • 2 weeks

    Pre-work through handoff