Prioritize the service dependency
Start with guest needs, coverage, production and the next handoff before adding a new process.
CASE 05 / Operations leadership
Leading managers across three businesses
Manager ownership, staffing context and a bakery pilot with a deliberate automation gate.
DELIVERED
Developed practical manager routines, training and checklists, and introduced a short real-task AI exercise tied to work managers already own.
EVIDENCE LIMIT
The reviewed evidence supports the routines and the exercise invitation. It does not establish independent AI adoption, completed exercises or time savings.
Would a busy manager actually use this?
A manager can be helping a guest, answering a staffing question and checking an event setup within a few minutes. I don’t want to hand them another process that only works when the restaurant is quiet. The routine has to earn its place in the shift.
What I owned.
I lead staffing, hiring, training, purchasing, service, events and P&L across our restaurant/bar, coffee shop and bakery. The bakery supplies our shop and other Spill the Beans locations. I develop routines managers can use, combine sales history with their shift context, and define manager-reviewed pilots before changing an authoritative process.
Where I used judgment.
Give the task an owner
Attach the routine to an existing manager responsibility with an action, follow-up and verification.
Make adoption observable
Introduce AI through a bounded real task in quieter service. Ask for usefulness and friction; do not count an invitation as independent use.
Who had to be in the loop.
I coordinate managers, business partners, kitchen, bar, floor and coffee teams. Bakery production and supply to other Spill the Beans locations add upstream planning responsibilities; the operating scope does not imply ownership of those other shops.
A forecast has to earn its way into an order.
I developed a manager-reviewed ordering pilot around counts, usable carryover and bakery supply. The source-readiness review showed why a useful forecast should stay in shadow until those inputs reconcile.
- The priority
- Make a useful production/supply decision without treating sales, orders and physical consumption as interchangeable.
- The alternatives
- Expose suggestions from incomplete history, or retain manual ordering while checking the source gaps. The pilot kept live suggestions locked.
- The responsibilities
- I defined the operating requirements and review gates. Store-manager counts, bakery commitments and receipts are distinct dependencies; no new staff assignment is implied by this example.
- The practical control
- Keep expected and actual stock separate. Record missing information as unknown. Require prospective evidence before calling a result an ordering improvement.
- What this proves
- A developed pilot and explicit readiness decisions. It does not prove automated multi-store ordering, lower waste, reduced stockouts or savings.
Leadership judgment: keep a promising tool small enough to evaluate while managers retain the authoritative order. Expanding access is not the same as establishing trust.
What was actually delivered.
- Operating procedures, checklists, staff training and manager coordination across three businesses.
- Staffing planning informed by POS history and managers’ observations around sports and citywide events.
- A bakery-ordering pilot and source-readiness review that keep uncertain forecasts out of live ordering; a bounded manager AI exercise introduced.
What remains to demonstrate.
Observe an independent manager completing a task, record friction, and validate bakery inputs prospectively before exposing live suggestions. No savings or broad AI adoption is claimed.
I can be excited about a tool and still need to find out whether it helps someone else. I want to see a manager use it on their own, understand the answer and know what to do next.