Blogs

A HiPPO's Guide: From Data to Artificial Intelligence

By Zara Palevani posted 09-05-2017 07:11 AM

  

zara-palevani-data-artificial-intelligence.jpg



HiPPO, chef, captain, or whatever cool name your team uses to address you, you are sitting at the top and leading the ship. Where should you take it in the next 12 months? I may not be able to answer that question precisely for you but in this post I would like to present you with a high level road map from data to AI. After all that is what every one is talking about these days, AI, eh? :)

HiPPOs of 2017 are the coolest of their breed in my humble opinion :) Not only they invest in educating themselves and their board to learn about the data ecosystem but also they are very well aware of the value of data as an asset to their organizations. This was not the case in 2008 when I started my career in the measure community. We have come a long way and we are going through an exciting time.

Sooner or later you need to update your business plan this year and if you are leading a large enterprise you will need to consider how to build capacity for machine learning and AI in your business.

In the illustration in the top of this blog entry, I am providing a road map that can be used as a guide for any organization. I'll try to keep this post short and focused on explaining what each layer means. Graduating from each level will qualify your business to move up to the next layer.

Business Objective: You can find a reference to it in every marketers deck these days. It's an attempt to use terminology that resonates the best with the CEO and CFO. I personally prefer Business Intelligence (BI) instead. BI doesn't fall under the technical domain. The CEO and the CFO can ask business questions, analyze information and make strategic decisions. If you are interested in some formal education in this topic I highly recommend the Business Intelligence course with Dalhousie University School of Information Management. If you are looking for an expert in this field for your company in NS drop me a line and I am happy to introduce you to someone that I highly recommend. The goal here is to articulate your organization's strategic direction backed up with data.

Data Infrastructure: I believe that the significant progress in this area is mostly the reason that Big Data has become a hot topic in recent years, Data Scientist's job became the sexiest job of the century and marketers need to learn about ML and AI today :)

From a technical stand, it wasn't feasible to capture, store and query petabyte and zetabyes of data, cheap and quick of course, about 8 or 10 years ago,. Today, we can do that without any major technical or financial barriers! Storing and querying data these days can be cheaper than your daily coffee expenses! With that I think I just implied that you don't have any excuse not to have a solid data infrastructure in place :). You can refer to my previous blog post for reference to some major cloud providers if you still need to invest in a solid data infrastructure for your organization. Keep on mind that the goal should be to build the infrastructure to get your data into one place and query it. Make it affordable. Make it accessible. Move on/up to the next phase.

Data Science and Analytics: With a solid data infrastructure in place comes the need to leverage the data for creating data models. Although analytics is still necessary, the true value lies in advanced analytics such as predictive modeling. Although PA deserves its own blog entry, for a definition of PA, please see the following.

At this phase, you should enable your team to learn about the best in class techniques in data modeling. The good news is that data modeling used to be expensive but it's much cheaper today which makes it easier for companies to invest in it. If you are in dire need to hire data modeling experts your best bet is to look for professionals with a degree in Information Management. Connect with your local universities, they sure can help :)

Machine Learning: I have a six part series on this topic. I highly recommend for those in management positions to learn the ML terminology and invest in learning what questions they can ask of their team. That is one of the keys to success with ML. Keep on mind that ML doesn't sit on top of Data Sci, ML is another tool in a Data Scientist's toolbox. But your organization has to climb up the maturity model to qualify to start investing in building an ML practice. For instance, you need a team who is well-versed in building data models. You can still gain business value from the data models to justify the cost and move on to building an ML practice which requires your team to use their models to collect more data, feed it back to those models to identify patterns and eventually prove value.

Artificial Intelligence: Once your team feels comfortable with building data models and leveraging the power of ML, it is time to push them out of their comfort zone yet again (that's what leaders do, right? :)) You must keep on mind that ML is a sub branch of AI. AI is the broader term with more opportunities. For instance another hot topic under the AI umbrella is deep learning which I find really exciting for marketers and digital analysts. Read my previous post for more practical examples on this topic. In summary Deep Learning has granted us access to even more exciting data sources such audio, image and video (unstructured data). While you are in this phase, you should look for opportunities with AI beyond machine learning. Deep learning is most likely a good place to start.

One Costly Mistake to Avoid

If you are looking for a recipe to fail in your upcoming data projects try to replicate what works for your software engineering projects!!!

This maybe an unpleasant news to many executives but data science projects are being initiated without a firm answer as to whether they will work or not (like any other scientific project). In contrary to a typical software development project, uncertainty in data sci projects is expected. The process is experimental which means that the end result may or may not work.

You may need to take a step back and analyze your organization's current culture. Is there room for testing, experimenting and working in various unknown unknown scenarios at your company? If not maybe you should avoid hiring a team of data scientists. Because the truth is that they will need time (plenty of time) and they also need permission to fail. If your senior leadership team doesn't have the patience and their frustration is going to be reflected in the performance review of your team members, you will be pushing the talent out of your organisation. Data Science is not supposed to be an addition to your current IT and dev portfolio.

Data Science will be a portfolio initiative with a variety of opportunities from high to low level risks. You need to analyze each opportunity and define your appetite for taking risk. While you think of the risk and cost of failure you may as well take into consideration that a win can also have a significant impact on your bottom line.

Summary:

In short:

A) No AI w/o ML > no ML w/o data sci & analytics > no analytics w/o infrastructure > no infrastructure w/o business objectives

B) Data to AI projects are not software development projects

C) Plan in for allocating resources to R&D (test, experiment, research)

Fuel Consumed to Produce This Blog Post:

In search for inspiration in the past few days, I stumbled upon La Finca while I was in Montreal. If you ever visit Montreal make sure to visit them and try their amazing menu of fresh homemade food :)

Maybe you could treat your team with inspiring food while moving your organization up the maturity model, from data to AI too.

Happy Business Developing! :)

*HiPPO: Highest Paid Person In the Organization

*w/o: without

*AI: Artificial Intelligene

*ML: Machine Learning

*Data Sci: Data Science



#ArtificialIntelligence #MachineLearning

Permalink

Most Recent Blogs

Log in to see this information

Either the content you're seeking doesn't exist or it requires proper authentication before viewing.