Shaping the work instead of estimating it
Fixed time, variable scope, and bets instead of backlogs. Why we scope engagements around appetite — not guesses — especially when the work involves AI.
The default way to plan a project is to estimate it: define the scope, guess how long it'll take, then watch the date slip as reality intrudes. For work with real uncertainty — which AI work always has — that's a recipe for scope creep and broken commitments. There's a better model, drawn from Basecamp's Shape Up methodology: fix the time, flex the scope, and bet on shaped work.
Fixed appetite, not fixed scope
Instead of asking "how long will this take," you ask "how much is this worth" — and set an appetite: a time budget you're willing to spend. Six weeks, say. Then you shape the work to fit the appetite, rather than estimating an open-ended scope and hoping. The appetite is fixed; what you build flexes to fit it.
Shaped work — rough but bounded
Before committing, work is shaped: defined enough to be understood and bounded, but not so detailed that it's already designed. Shaping sets the boundaries and names the risky unknowns, so you're betting on something real without pretending to have foreseen every detail. Example: "a board-packet workflow across Intacct, Cvent, and NPSP, in six weeks" is shaped; "47 Jira tickets" is a false-precision estimate.
Bets, not backlogs
You don't maintain an ever-growing backlog of guilt. You make bets: a decision to spend one appetite on one shaped piece of work, with a circuit breaker — if it doesn't land in the budget, it doesn't automatically get an extension. That pressure forces scope decisions early, when they're cheap.
Why this fits AI work — and our engagements
AI work is uncertain by nature: you don't fully know what's possible until you're in the data. Fixed-scope/variable-time planning turns that uncertainty into slipped dates and blown budgets. Fixed-time/variable-scope turns it into focus — you ship the most valuable version that fits the appetite. It maps directly onto our milestone-gate engagement model: each gate is a bet, the appetite is fixed, and scope flexes to deliver real value at every step rather than a big-bang reveal at the end.
Estimating asks "how long will the scope take?" Shaping asks "what's the best thing we can ship in the time it's worth?" For uncertain work, the second question is the only honest one.
That's why we scope around appetite and prove value at gates — so you're never betting the whole engagement on a guess made before anyone touched the data.