When business leaders come to picking the right AI project for their company, this often boils down to having the right ingredients and knowing how to combine them. Almost every functional AI system today works through some manner of rules that are encoded in the system and effectively hold everything together.
Vital to success, too, is how their teams can effectively communicate as data scientists, and repeatedly asking themselves a series of questions throughout the development process.
We can break AI down into ingredients because there’s a whole menu of things that it can do. When you have an idea of what they are, it gives you an idea of what its powers are.
The first ingredient is “yes” and “no” questions. For example, if I send you an email, are you going to open it or not? These give you a probability of whether something is going to happen. We ask this question at every stage of the AI project.
The second ingredient is numeric prediction. For example, how many days is it going to take you to pay your bill? Or how long is it going to take me to fix this person’s refrigerator?
Thirdly, we have classifications. For instance, when you take a picture of your team, ask, “are there people in this picture?” “How many people are in this picture?” There are text classifications, too, which you see when you interact with a chatbot.
The fourth ingredient is conversions – taking information and translating it from one format to another. This could be voice transcription, or translation.
Where to Start
Starting your AI project journey, the fundamental questions you need to ask is, “What data do we have?” And, “What concrete problems can I solve with it?”
Take, for instance, something that every salesperson tracks as a natural part of their job: categorising a lead by giving it a score of how likely it is to close. Data sets like these are a key source of truth from which to develop an AI-based project.
People want to do all kinds of things with AI capabilities, but if you don’t have the data, then you have a problem.
How to Get from Pilot to Rollout
To get the project from the pilot to rollout stage, the first question you need to answer is what is the problem you are trying to solve. Am I trying to prioritise people’s time? Am I trying to automate something new? Then, you can confirm that you have the data for this project, or that you can get it.
The next question you need to ask is: Is this a reasonable goal? If you are seeking to automate 100% of your customer service queries, put simply, you are setting yourself up for failure. However, if 25% of your customer service queries are requests to reset a password, and you want to automate that and take it off your agents’ to-do list, that is a reasonable goal.
Another key question is: Can a human do it? Most of the time AI can’t do anything that humans can’t do.
There are two reasons why AI projects tend to – and should – have uncomfortably long pilot periods.
First, you need to determine whether it actually works the way it should. From healthcare diagnosis to movie recommendations, the context and importance of AI-powered recommendation will vary but ultimately you need to share the explanation so your users will trust it.
Second, you need to measure the value of the AI solution versus baseline — human interaction. Think about automating customer service queries. If your chatbot can’t answer your customers’ questions, you will end up with frustrated customers who hate your chatbot and end up talking to a human anyway.
Why We Need to Stay Grounded
We know that understanding and using data sets to inform machine learning systems can solve problems more effectively. Yet intuition also has an important role to play in this process.
Take for example building a custom prediction for questions like, “Will my customer pay their bill late or not?”
Often data scientists in the AI field have a tendency to think about algorithms, or maybe slightly higher level abstractions. What we really need to do is get into our customers’ heads and express the solution to the problem in terms that they will relate to. This means that it’s not just about making a recommendation; it’s specifically recommending the part that goes into a project. Also, it’s not just making a prediction, but specifically answering the question, are you going to pay your bill or not?
From here you have to decide, if I make that prediction, I give you a probability of the guy paying late, what are we going to do about it?
Ultimately teams need to stay grounded when considering what problems they should try to solve with AI and what they have on hand that can help them do it. It’s going back to the question of: Can a human do it? If they can, maybe AI is a great way to take that task off a human’s plate to free them up for more strategic tasks.
By Robin Fisher, Senior Area Vice President, Salesforce Emerging Markets