Being an AI professional, you often have to advise your client in situations such as
- defining scope of a problem,
- design of a data science or machine learning solution,
- deployment of machine learning model,
- AI strategy for a business unit
- selecting vendor solution
As an AI advisor, you would need to leverage your expertise, experience and ability to collect and analyze relevant information. For a given problem you would build an argument, present it clearly, and then defend the argument.
A client can seek your advice on two types of AI problems;
Advising on Simple AI Problem
On simple AI problem, you could provide your advice quickly based on information from the top of your head. For instance, how often should model X (which is ready for deployment) be re-trained after deployment in production. You may answer this question right away because you have been closely working on this type of models and know the seasonality patterns in training data.
Advising on Complex AI Problem
For this type of problem, you would need to spend some time in order to collect and analyze the information, build your argument and present it properly and defend it by answering any follow up questions.
For instance, your client is thinking to move machine learning development and deployment to cloud and would like to know which vendor (AWS, Google, MS) to go with. To advise on this, you would need to collect information about client's existing on-premises infrastructure and future AI vision, as well as detailed information about offerings of different cloud vendors. Then you would be able to build the argument for a suitable option and present it to the client.
To build your argument, you can adopt one of following two approaches;
Deductive argument building approach is analytical in nature. The argument must be based on a set of solid and factual premises. The construction of deductive argument follows funnel style flow of premises.
The risk in this approach is that striking down of one premise can bring the entire argument down.
This approach is persuasive in nature and doesn't have any logical order of premises e.g. give 5 reasons why we should move AI solution development to AWS cloud.
Whichever approach you follow, the argument must be based on
a) valid premises and
b) the premises must be aligned with client's context and decision making process.
For instance, you may advise your client to migrate to AWS cloud, because of
- High availability
- Strong on-demand ML capabilities
- Competitive pricing
- Easily available AWS talent
Above premises are valid but the client may belong to a highly regulated industry and not allowed to move data to cloud. In this situation, your argument may not work. Therefore, you must think of questions and situations which may arise during presentation of your argument.