Over the last decade, I’ve watched countless “AI companies” fade away because they could never explain any real ROI. And I’ve seen twice as many data analytics projects crash and burn because they chose the wrong AI/machine learning tools. But with so much hype and so many options out there, how can you select the right AI vendor?
Choose wrong, and you could throw millions away on a poorly-implemented AI that leads to disaster. Never forget: The MD Anderson cancer center spent more than $62 million on IBM’s Watson for Oncology for a product that “didn’t deliver“.
Based on my observations and Lexalytics’, an InMoment company, experiences using AI and machine learning for NLP, I’ve identified 4 key questions to help you choose the right AI vendor:
- What’s the ROI going to be?
- How experienced is this company?
- How much can they tell you about their own systems?
- What’s the real cost to me?
In this blog, I’ll discuss each question in turn and explain how they should guide your search for an AI vendor.
1. What’s the ROI Going to Be?
AI’s true potential is in targeted solutions to specific challenges. That flashy “AI company” you’re chatting with might look good, but how will they deliver real business value? Even before you begin looking for a technology solution, make sure you’ve identified the data problem or question you’re trying to solve. For example, maybe you’re analyzing video gaming forum posts, and traditional sentiment analysis can’t figure out when “that sucks!” is positive or negative.
Once you’ve identified a specific problem, insist that your potential AI partner demonstrate the real quantified value they will deliver. If they can’t talk ROI, go find someone who can.
2. How Experienced Is this Company?
Everyone wants AI and machine learning, but nobody wants to work with an unproven startup. That’s because the key to a successful AI system, one that delivers real benefits, lies in the implementation.
Remember: while degrees and certifications are useful, they’re no substitute for real-world experience.
If they can’t talk ROI, go find someone else who can.
Make sure the company you’re partnering with has a proven track record of building and implementing AI and machine learning systems that deliver real business benefits.
3. How Much Can They Tell You About Their Own AI?
A company should speak confidently and openly about the techniques they use to build machine learning models and develop AI systems. So, ask your potential software partner to explain how their AI and machine learning systems actually work.
Some questions you can ask:
- How do you choose which machine learning model to use for each task?
- Why is your approach better than others?
- How do you handle hyper-parameter optimization?
- What’s your methodology for maintaining gold-standard test sets?
Do they speak with certainty, or are they unwilling to go into details? Their answers will be revealing.
4. What’s the Real Cost to Me?
A complex AI system that claims to solve huge problems may look good now, but you’ll be far behind your competition after it takes years to implement. On the other hand, a “one-size-fits-all” machine learning tool will cost you far more in the long run, even if it can get up-and-running in a few days.
At best, that “one-size” tool will return mediocre results. At worst, it’ll lead you to make bad decisions based on misleading data.
Take the time to do your research and find an AI vendor that combines sophisticated software with a practiced methodology for implementing them. You’ll get a better solution in less time, while saving thousands of dollars in unexpected fees, costs, and delays.
Choosing the Right AI Vendor
A targeted, well-trained AI system can deliver huge benefits to your company and your own clients.
By using these 4 questions to guide your search for an AI vendor, you can realize AI’s huge benefits while avoiding the common traps that doom so many AI companies and projects.
Successful AI Applications
9 Ways AI is Reshaping Retail Banking – TechBullion
How AI can Optimize your Sales Funnel – MarTech Advisor