by Dharmesh Syal
Artificial intelligence (AI) has made breathtaking strides in the past few years embedding itself into the consumer subconscious. Computers are using advanced data science, machine learning (ML) and neural-network programs to learn more and more about individuals’ traits and experiences, predict outcomes and prescribe actions.
As impressive as this progress may be, current AI technologies have yet to master the abstract thinking, instincts, and judgement that drive human behavior. The technological advances that will make AI truly valuable for business (behavior intelligence and deep learning) are just over the horizon. When computers attain these human-like capabilities, AI will finally begin to create value for business and enable strategic growth by building meaningful customer relationships based on trust.
Given the rapid pace of technological change, many companies may prefer to wait and see how the AI revolution unfolds before making significant investments—but this could be a grave mistake. Very soon, machines that think, talk, and persuade like humans will become a reality and those who embrace the technology early and learn how to use it effectively will gain a powerful advantage.
But given the current state of AI, how do businesses functionally do this? Not in the next few years—but in the next six to nine months? At BCG Digital Ventures, we employ minimum viable product (MVP) methodology to quickly launch our AI concepts. The ideal innovation process for any product should live at the intersection of desirability, feasibility and viability and developing your MVP will help get there.
Desirability – do people want it?
Feasibility and viability don’t matter if you don’t first create desirability. This is the value proposition–understanding who your customer is and what they really want. Will the AI solution you’re developing fill a genuine need? Are you building something that won’t just fit into your customers’ lives, but that they truly cannot live without?
AI ventures deliver exponential value in the long run, so initial use cases should be around validating concepts. Focus on frictions that are measurable explicitly in customer transactions. Adoption is key, so any engagement created has to fulfill a specific need and create positive change.
Feasibility – can it be done?
Yes, there are many problems that can be solved by AI and opportunities to cash in, but it’s important to determine what bite-size portion of the problem you can tackle. Ask yourself: Is the solution to the problem you’re using AI to solve really within reach? Can you actually make it happen? And if so, how long will it take?
MVPs tend to need human interaction and engagement to build an initial human and machine intelligence system. Data for machine learning is often inconsistent and has gaps and heuristics dominate initial algorithms. Built-in customer interactions create a lot of baselining data for algorithms and also generate learning data for next phases.
MVPs should create a feedback loop and foundation for learning systems. Some pockets of ML using clustering and regression techniques which may be implemented to test your hypothesis.
Viability – does it have a strong business case?
Last but not least, how do you determine if your MVP is going to be a viable business concept? Validating the accuracy of data or predictions will only happen over time, so don’t waste your energy here. What you should focus on instead is how to build enough so you can ‘directionally’ validate the business concept you are driving. This should show outcomes that can lead to revenue and growth goals.
Business partnerships and ecosystem foundation are key for MVPs and may involve equity relationships. This can often make or break a startup. An early alpha release creates opportunities for several beta tests. Synthetic data can aid alpha, but we should be wary of the risk of creating bias. It takes a village, so it’s important to take a federated approach to data and algorithm expertise.
The same technology that started with credit card fraud detection has proliferated into your Netflix account, shopping habits and mode of transportation–and this is just the beginning. If you aren’t considering how AI can create value for your business and improve your customers’ experience, you should be. By using MVP methodology you will be able to quickly launch your AI concepts into the market and not be left watching from the sidelines.
Dharmesh Syal has worked for two decades with Technology and Telecom Leaders launching new products and business models using technology innovation. He currently serves as a Partner and CTO at BCG Digital Ventures.