In our last post, we discussed why having a data strategy is essential for the modern CMO. CMOs can no longer rely on CTOs to get their team the data they need to make critical business decisions and empower their teams.
Often when building a data strategy, the question of how and when to leverage AI (artificial intelligence) comes up.
Before even considering AI for your business, it is essential to take a step back and evaluate the problem you are trying to solve.
“Tech cannot solve people and process problems. It can only help scale a solution.” - Francis Brero, Co-Founder and CPO at MadKudu.
Often there is a misconception about the use of AI. It is a hyped-up industry buzzword that makes most people feel it will solve all of their challenges. Contrary to popular belief, technology, and especially AI, alone won’t solve your problems.
Start with the problem first.
First and foremost, it is critical to understand the challenge you are trying to overcome. For example, don’t use AI to create new content until you have a content strategy in place and understand what’s working well and what isn’t.
AI, ML, etc., are there to help scale the decision process, but you need to know how and why you’re going to use the AI to be effective. As Jeff Ignacio, Sales Operations Lead, West at AWS, said, “Any good technologist is more of a business problem solver, and tooling is just a tool.”
Utilize data available to you now.
Once you understand the problem at hand, see what can be done without AI first. Eugene Yan discusses this concept in the post “The First Rule of Machine Learning: Start Without Machine Learning.” He advises starting with your data and drawing conclusions. It is time to utilize machine learning only after you have a non-ML baseline that performs well and can’t scale without breaking.
For example, if you have 50 or fewer leads a month interacting with your website chat start with simple engagement rules based on the most frequent questions asked. Adding AI to your website chat isn’t necessary until you have a high volume of chats and cannot rely on simple engagement rules anymore.
Take it from the experts – start with the problem, see what you can build with your data (and knowledge of your business, marketing is part art and science, after all) before diving into the world of AI.
Buddy vs. Police
Once you’ve evaluated the problem and reached a breaking point in what you can achieve with your current data and infrastructure, it may be time to consider AI. A great way to think about AI and what is right for your business is in the framework of personality types – buddy and police.
Police AI: Police AI is a system designed to make better decisions and enforce them. This type of AI prevents some choices and enforces others.
Buddy AI: Buddy AI is meant to provide suggestions and let end-users make the final decision. This type of AI balances technology with human input.
The type of AI selected ultimately impacts end-user adoption.
AI & Design
Many companies struggle with the adoption of artificial intelligence because there aren’t any clearly defined design standards.
Design in AI is critical because how you present AI to the end-user can drive adoption or destroy adoption and fast.
Simply put, the power is in your hands.
To illustrate, let’s look at predictive lead scoring.
Marketing often views predictive lead scoring as police AI, defining how the leads should be routed (or not) to sales. However, sellers are often reluctant (and rightfully so) to police AI because they feel it ignores nuances inherent to what they know about the business and their prospects.
Sales teams typically prefer buddy AI that guides them in the right direction with helpful context but doesn’t contradict what they know about their leads and prospects. Presenting buddy AI typically helps build trust and adoption. Predictive lead scoring is just one example but important to consider when designing a system that works for the end-user.
To sum it up:
- Start with the problem
- Utilize your data and tools
- Evaluate AI once you understand the problem and need to scale your current solution (considering design and AI types)