Create an AI Product Business Proposal

Are all required files submitted?

The submission includes a Proposal file (a pdf) and images of the user interfaces of the proposed product.

Is the proposal complete?

Every section of the proposal file has been completed.

Does the proposal describe a business problem?

The proposal includes a short description of the problem the product seeks to address. The problem is described in business terms; for example, the problem might be the need to increase customer satisfaction or drive repeat customership. What is the business benefit of your proposed solution?

Does the proposal make a business case for the product?

The proposal contains an argument in favor of the product that derives from impact on revenue, market share, and/or other drivers of business success.

Does the proposal describe a narrow task that AI/ML could help solve?

The proposal should describe what the AI/ML model will actually do.

Do the success metrics relate to the business goal(s)?

The success metrics measure how well the product achieves the business goal(s).

Does the proposal describe the size of the data? Does the proposal describe potential biases in the data?

The proposal should include an estimate of the size of the data. The proposal should describe the categories/types of data that will be under/overrepresented in the dataset(s).

Does the proposal describe how the data will be acquired and discuss issues that may arise in acquisition?

The proposal should discuss the following considerations: buying vs. collecting data; privacy/personally identifying information (PII)/sensitivity issues; cost; ongoing data vs. one-off data dump (which would need to be refreshed).

Does the proposal justify the choice of data labels?

The proposal should explain the proposed labeling scheme, and why the chosen labeling scheme was chosen. What are the strengths and weaknesses of such a labeling scheme?

Does the proposal describe how the model will be built?

The proposal should include a description of how the model will be built, and should discuss considerations such as the likelihood an external platform will satisfy the specific use case, the need for certain controls on the model, data sensitivity/security, and willingness to give an external platform access to the data.

Does the proposal describe the planned use of ML metrics to measure the performance of the model?

The proposal should describe how ML metrics such as accuracy, precision, recall, F1 score, etc., will be used to assess the performance of the model.

Does the proposal include sketches of main user interfaces of the product?

The proposal should include a few sketches of main user interfaces of the product.

Does the proposal describe the “types” of users who will use the product?

The proposal should describe a few prototypical users and their use cases.

Does the proposal lay out a general plan for building and launching the product?

The proposal should have a general pre-launch and post-launch plan.

Does the proposal describe strategies to improve the product in the long term?

The proposal should describe a plan for testing, versioning, and active learning (learning from new data).

Does the proposal include a plan for monitoring bias?

The proposal should mention a plan for mitigating unwanted biases.