As the AI tidal wave surges and we see more and more reports firms like McKinsey predicting record GDP growth due to AI, companies across the globe are feeling the pressure to begin adopting AI. This pressure is coming from venture capital, private equity, shareholders, or other investors. These investors are looking for opportunities to change marginal cost activities to one-time investments, and achieve superior profitability and disrupt incumbent players. The issue is, these firms don’t know how, where, and, most importantly, they don’t take the time to establish a suitable why. They simply “need AI”. This is representative of a common issue we see among current and potential clients.
The pressure firms face to adopt AI is real and the resulting mindset tunnel vision of adopting AI in any form is understandable, but it’s crucial to the long-term success of any holistic AI endeavor that the guiding logic is sound. Doing AI for the sake of doing AI is not a strategy, it’s a knee-jerk reaction.
When projects are begun in this way they most often result in a flop. Why? They are not tied to business objectives. Often money and resources are spent on a project that sounds good but lacks substance. As the energy is funneled to the work and the results are slow in coming, enthusiasm for AI projects wanes. This leads to a challenging cultural outcome: lack of support. Your backers may not wish to pour more money down the drain. Your executive champion may not want to put their neck out again. Your end-users may not want to hassle with spotty service. Getting a second project off the ground will be a feat in those conditions. Thus, starting off your AI journey with the right first project can make the difference between a one-off and a holistic endeavor that will reshape your company. The question is…
Where Do I Start?
Here are some basic questions you should ask yourself as you begin to explore AI:
- What pain am I feeling in my organization? Or, what opportunity am I missing out on?
- If I solved that pain, would I be able to produce more? Would I reduce costs?
- If I seized that opportunity, would I be able to generate new revenue?
- Do I have data that reflects the pain or opportunity? Am I capturing data regarding how the human currently solves the problem? Can I find data to support it?
- Is this an AI problem? How do I know if AI is the right tool for the job?*
- Assuming this is an AI problem, is it feasible?
- Do I need a custom solution, or is there a product I can use?
Answering these questions qualifies three important prerequisites:
- The project has business value
- The project is an AI project
- There is sufficient data & technological capability to support the project
With those three things in hand, you can feel much more confident in pursuing a solution, be it a bespoke model or an off-the-shelf product.
We put an asterisk next to the “Is this an AI problem?” question for two reasons:
- Often that answer turns out to be something like “No” or “not necessarily”. AI is often used in the same breath as automation but never forget that automation can be accomplished through simple rules-based approaches too.
- The second half of the question – is AI the right tool – is harder to answer and requires further exploration.
Is AI the Right Tool for the Job?
This can be a difficult question to answer. As discussed earlier, many times AI fits the bill. Other times, AI could work but might not be the most efficient tool to solve the pain. Understanding which situation is which takes exposure and practice. There are, however, some questions you can ask yourself to guide this exploration.
Let’s assume that you’re considering an automation project. You have an expensive team of experts who spend their days reviewing data and making decisions based on what they see. You think to yourself that it’d be lovely to automate their tasks – you could save potentially hundreds of thousands of dollars and you could also enact their decisions much more quickly. On its face, this sounds like a great AI project, right? Maybe, maybe not.
What types of decisions do those workers make? When evaluating data, are they comparing it to a simple set of rules or are they making judgment calls? How many different potential decisions could they make based on what they see? How many different data points could they choose to evaluate before making a decision?
If the decisions are relatively simple and based on a straightforward set of rules, AI is not the right tool for the job. There is likely a software product available geared towards your problem. If not, it is likely not a terrible lift for a software engineer to develop a rules-based engine. If, however, the aforementioned workers have to make tough judgment calls based on “gut” or have a panoply of possible data points and decisions, you may have an AI problem after all.
What kind of data do they look at? And just how much data is there?
A prime example would be automating the task of QA on assembly lines. If you produce cars in a factory, you may have humans that walk the floor and visually inspect elements for defects. Is this a deep thinking, “human intuition” problem? No, but it does require a human’s ability to see and understand, and there is a lot of visual data to process. Humans get tired; Machines do not get tired. This is a ripe opportunity for introducing AI to the mix to augment humans by reducing their cognitive load and presenting them with the most challenging questions.
At the end of the day, there are precious few simple heuristics for determining if your project is an AI project. The closest I’ve come is the question “Is this problem something that requires a human to evaluate inputs, consider its options, and make a decision?” If the answer is yes, AI could probably apply. If the answer is no or maybe, start asking more questions.
If you need help sorting through these questions or want guidance on how to begin an AI project, reach out to us at email@example.com. Hope to chat with you soon!