Nasser Ghanemzadeh

AI Feature Sprint

Workflow Scorecard

Two minutes. Tells you whether your AI idea is a ten-day workflow, and what to fix if it isn't.

It runs in your browser. Your answers stay on this page unless you book the call.

01 · Scope

Write the workflow as one sentence

Fill the five blanks. If a blank is hard to fill, that's the first thing the call will fix.

The one person who runs this: a role, not "the team."

What arrives or happens to start it.

What it does, in one verb.

The artifact, in one noun.

A number that moves: minutes, errors, throughput.

02 · Score

Score it on eight dimensions

One to five each. The verdict is only as honest as the scores.

1: nice to have. 5: the board asks about it.

1: a few times a year. 5: daily.

1: mild. 5: people avoid the task.

1: scattered or absent. 5: in one system with an API or export.

1: needs new infrastructure. 5: fits your existing APIs.

1: a wrong output is irreversible, legal, or financial. 5: a human reviews before anything leaves.

1: no owner. 5: the owner asked for it.

1: one-off. 5: the same shape as three other things you need.

03 · Readiness

Check what the Sprint needs from your side

Four things have to be true by Day 1.

Read-only codebase access can be granted within a week
Sample data exists and can be shared
One decision-maker will answer within a business day during the Sprint
A data processing agreement is in place with your model provider (Anthropic, OpenAI, or other)

Reference

How to tell whether an AI feature can ship in ten days

An AI feature can ship in ten days when it is one narrow workflow: one user, one input, one output. Its data already sits in one system you can export, it runs on APIs you already have, and a person reviews every output before anything leaves. Miss one and it takes longer.

The one-sentence scope

Every Sprint is written around one sentence with five blanks. If you can't fill all five, the workflow isn't scoped yet.

For [user], when [trigger or input], the workflow will [process] and produce [output], so that [value].

Example. "For the founder, when monthly metrics and three to five sentences of context are in, the workflow will draft and produce an investor update covering metrics, risks, and asks, so that sending it takes 8 minutes instead of 90."

The eight dimensions

Each is scored from 1 to 5, for a total out of 40.

Business value

Whether the workflow moves a number leadership already tracks. A 1 is a nice-to-have nobody would miss; a 5 is something the board asks about. It most often fails when the value is stated as "saves time" with no number attached.

Frequency

How often the workflow runs. A 1 runs a few times a year; a 5 runs daily. It fails when the idea is a quarterly report: too few runs to tune the prompts or to pay back the build.

Pain intensity

How much the people who do the task today dislike it. A 1 is mildly annoying; a 5 is a task people avoid or hand to the newest hire. It fails when the pain belongs to the founder rather than to the person who does the work.

Data availability

Whether the inputs exist and can be reached. A 1 means the data is scattered or absent; a 5 means it sits in one system with an API or export. It most often fails on data that exists but lives in five places with no owner.

Technical feasibility

Whether the workflow fits the systems you already run. A 1 needs new infrastructure; a 5 calls APIs you already use. It fails when the plan starts with building a platform instead of the workflow.

Risk manageability

What happens when the output is wrong. A 1 means a mistake is irreversible, legal, or financial; a 5 means a person reviews every output before anything leaves. It fails when version one is designed to act on its own instead of drafting for review.

Adoption likelihood

Whether someone will use it every week. A 1 has no owner; a 5 was asked for by the person who will use it. It fails when the workflow is built for "the team" and nobody in particular.

Reusability

Whether the same build serves more than one need. A 1 is a one-off; a 5 has the same shape, input to draft to review, as three other things you need. It fails when the workflow is so custom that the next one starts from zero.

Green, yellow, red

Green means Sprint-ready. The total is 30 or more, data availability and technical feasibility are 4 or more, risk manageability is 3 or more, at least three of the four readiness checks are a yes, and every blank in the sentence is filled.

Red means not a Sprint yet. Business value, data availability, or technical feasibility is 2 or lower, risk manageability is 1, or the total is under 22. Any one of these is enough. The fix is usually a different workflow, a month of sample data in one place, or turning the workflow into a draft a person reviews.

Yellow is everything in between: fixable. The scorecard names the two lowest dimensions and any readiness gap, with what closes each one. An empty blank in the sentence also holds a result at yellow.