Will your views on the project survive this meeting?

You can take your corporate PowerPoint out and bore your audience, or you can captivate it.

I have gone through these meetings hundreds of times with a 90% sales success rate and have refined a method that can be summed up in my 2-minute pitch, which always goes something like the following.

I will display a chart and the following graph and say:

"Let's go straight to the point. We're trying to analyze the details of products based on their images and find the most profitable ones. Here is a one-page chart showing the list of the most promising products. Here is TensorBoard; it comes with TensorFlow package. The computation power of TensorFlow's data flow guarantees disruptive innovative profits for this project.
At the bottom of the screen, you can see that we are going to input store products (point to the input_store_products box). What we expect is the profitable classification of products: the best ones that represent our brand AND the ones that also generate the most profit. So basically, there are two main types of features to extract: forms, shapes, and colors of the most trendy products; and profit.
To achieve that, we feed (that's literally the language we use in code) the system with the results we expect, a dataset of good choices (point to the input_expect_classification box). Then we run the system; it optimizes until it has been trained through gradient descent (point your arm and finger downwards), which is simply a cost function to lower (CEO's understand lowering costs in 1/1,000,000th of a second) the errors. Once the weights (just as we weigh decisions) are optimized (I point to the weights), that's it. We will provide you each day with reports to drive the marketing department."

A dataflow graph with input_store_products

Then I stop immediately, slowly look at everybody, and say:

"To sum it up, we select the most profitable products in real time with this system and provide the marketing department with real-time AI reports to boost their marketing campaigns. By analyzing the features of the products, they will have more precise marketing data to drive their sales."

If you read this aloud slowly, it should take you be under 2 minutes, which is your first-level survival time.

The rules I've applied successfully are as follows: 

• Explain from off the top (global concepts).
• Use as few words as possible, but choose the right ones. Don't avoid technical concepts, but break them down into small key parts. This means that, once again, TensorBoard architecture will develop this ability.
• Speak slowly. Speaking slowly, stopping at each sentence, emphasizing each keyword, and pointing to the graph at the same time is the only way to make sure you are not losing anybody.
• After 2 minutes, sum your introduction up and ask: "Am I clear? Do you want me to continue. Questions?" If they want you to continue, then you can drill down, stopping at every key point to make sure that everybody is still with you.

Bear in mind that many CEOs are business visionaries. They see things we don't. So, if they come up with an idea, be very constructive and innovative.

Was the sales pitch convincing?

At this point, if you have a good project, the CEO will most probably say one of the following: 

• "Gee, that was clear and I understand what you are doing. Thanks. Bye." That's not too bad. 
• "That looks too simple to be true. Can you give some more details?" This could be an opportunity. The CEO wants me to explain more. So I drill down another 2 minutes around the way the weights are optimized in a clear way just like the first 2 minutes. At that point, in 20% of my meetings, I ended up having lunch or dinner with the CEO, who always has 100 other ideas for uses of such a program. Bear in mind that many CEOs are business visionaries. They see things we don't. So, if a new idea comes up, I generally say, "Great. I'll look into this and get back to you tomorrow." Then I spend the night coming up with something that works and is explained clearly.
• "Good. I'd like you to talk to X and see where this is going." Not bad either. Hopefully, X will listen to the CEO.

Statistically, being short and concise with real technical tangible objects to show worked 90% of the time for me. The 10% failures are lessons to learn from to keep the statistic up the next time around.

The percentage of failure of our presentations is our cost function. We have to bring that number down by learning from these failures by backpropagation as well. We need to go back and change our parameters to adapt to the next run.