A few weeks back, I had the opportunity to visit a tax department for one of the world’s leading global economies. My colleague Alexander Fred-Ojala and I were there to discuss how artificial intelligence and data technologies might fit into their strategic plans.
It was a great discussion, and a lot was learned on both sides. We started with two fundamental topics:
- Understanding the Silicon Valley perspective on data-driven business models and
- The tools and techniques used in AI, machine learning, and data borrowed from our very applied Data-X course (see data-x.blog).
This was a great start.
Then we asked them about the opportunities for using these emerging technologies in a tax service. It was insightful to understand some real-life examples that any tax service around the world might be concerned about. They highlighted one case where a bad actor might log in and impersonate a proper taxpayer, then file a false return to have the return directed to them. Then later, the innocent taxpayer is on the hook to resolve and possibly payback the false return, while the bad actor has taken the funds. It’s basically a fraud scenario.
Now here is the important insight: What type of problem is
- an authentication problem
- a system problem
- a data science / AI problem
- a data security problem
- a website problem
- an operations problem
- a policy issue
- or something else?
The answer is really all of the above. It would be a gross overstatement to say that this problem is solved simply by using data signals to separate authentic users from bad actors.
This is really the case with most problems and projects. While most projects have a problem/solution nature, they rarely look like AI / ML / Data problems at the start. And when data, AI, and ML are part of the solution, then this technology is certainly not the only or even the main component in the solution. This is a call for stepping back and looking at problems and solutions at the systems level.
In real life, the entrepreneurial and innovative process starts with a story or narrative of the problem, and this still holds true in this case — and is in fact very necessary. In Data-X, we teach students to write a “story” in the first 4 weeks, and then turn it into a low-tech demo without technology. Then they work on agile sprints for the next 8 weeks. Students (as well as executives) come to us to learn how to use AI, ML, and Data, but they soon realize that the system, solution, and process is much more integrated. If they learn how to first view the real problem, they learn how and where to start, and they learn when and where AI or Data can be part of that solution.
Without learning how to set up the problem in a broad-thinking
manner, the solution will be either too focused on one technology or too
focused one narrow aspect of a problem. Why does this matter? The reason is that without an integrated approach, the end result will be ineffective, overly complex, and/or overly expensive to implement.
For years, great technical architects have understood how to solve complex problems by breaking them down at a fundamental systems-level while keeping the design as simple and effective as possible. Successful entrepreneurs/innovators have always known how to target the most
valuable of problems by using story and by using inductive and agile learning processes.
|System Architecture Development||Entrepreneur/Innovator|
|Start with the user’s story -> Get to Effective Implementation 1.0||Start with user’s story -> get to Business Model|
|Break it down: first principals/relationships||Use story to gather stakeholders/resources|
|Agile execution||Inductive learning|
|Minimal Viable System: Keep it Simple||Minimal Viable Product for market fit|
|Use broad thinking to reduce failures/flaws||Use broad thinking to reduce business risk|
This table illustrates the parallel of effective System Development
vs Entrepreneurship/Innovation Process
You can learn more about our approach for Innovation Engineering by combining the two approaches (of architects and entrepreneurs/innovators). We do this already within Data-X and our industry projects, but it is an approach that is extensible beyond this domain.
I feel it would be valuable if we could further distill this approach to help companies and organizations around the world to create innovations that are effective without incurring the massive cost overruns often associated with new technology projects. Of course, we would love to hear your feedback to make it work even better.
Could be written as increase robustness, increase reliability, reduce cost,
increase performance (all are true while holding other variables constant)