How should I manage AI expectations with my CEO?
Sort By:
Oldest
VP Of Engineering in Software6 years ago
When most people ask about AI (Artificial intelligence), they are really asking about ML(Machine learning)[1]. In any case, following are some suggestions on setting expectations with your CEOThis is a capability that is worth developing internally, as in, this is unlikely to be a fad that goes away. So ask for a separate team/budget for this. It does not have to be excessive. Most companies will do well to start with a team of around 5 data scientists/analysts.Educate your CEO and the e-staff on the reality vs. hype in this space. If needed get external consultant and spend several hours with your CEO on this. Go into the math a bit so it is demystified and does not remain something esoteric. Lead with business outcomes that you want to enable. ML/AI is not a panacea, it is not going to make you a business that you fundamentally are not. The onus should be on business leaders to think generatively about how ML/AI can help them. You will have to try large number of projects to develop this muscle. So ask your CEO to be prepared for several failed initiatives. So, set expectation that rapid iteration and quantity of projects is what you will optimize for in the first 6 months. Set expectation that ML is dependent on availability of data. If you don't have much data, you are unlikely to build a model that generates magic from nothing. Depending on the technical level of your company, set expectation that there needs to be a cultural change to make data based decisions part of your core DNA. The CEO (and e-staff) should lead by example. Few things for you to do, watch outLook hard at data you have but have not been able to use successfully. For instance, many companies have machine sensors that generate large amounts of data that go into a black hole. Revisit these large data stores and ask what new things can be done with them. ML is doomed to fail if the underlying data is bad. Pay attention to your data infrastructure and get that to a good place constantly. For clarity, I am not suggesting a big bang, "build a data lake" project. But ensure that all data sources needed for business outcomes in #3 are in a good place and then build from thatOften when people say they want ML, they mean they want better analytics as in they want more reliable quicker answers to questions they have. Test if this is the case and you can do really well by hiring strong analysts as opposed to data scientistsPair key business leaders with analysts who can answer any data related question for them and can educate them on how to interpret data. It is surprising that there are many leaders who do not understand terms like p-value and correlation coefficient. Getting leaders moderately savvy, goes a long way in changing the organization culture. [1] On definitions Artificial intelligence is the term used when machines do things that are usually considered the realm of humans. Machine Learning is when a machine gets better at performing a task with experience. ML is one of the ways to build AI. Deep learning is a subset of machine learning that uses neural networks. VP, Innovation and Technology Services, CIO & Chief Digital Officer in Education6 years ago
I think Arvind's answer is perfect. The only thing I can add are two examples from my organization with two different outcomes.Example 1: We have many of our classrooms equipped with monitors and video cams so that we can include remote participants and guest lecturers. Using Microsoft's [now deprecated] emotion API, we were able to build a system that could use the cameras to feed live reactions from students and create an "engagement meter" that reflected how students were reacting to the instructor at any given time. A timeline was created so that instructors could determine where peaks and valleys occurred during a given lecture. Problem: Most faculty (even around the country) were not fans. The few faculty that liked the idea were not the ones that would need help with student engagement. Further, there were concerns about privacy, etc. The outcome was a cool product that had no place in our organization, but might be marketable in other industries. However, we did learn a lot (to Arvind's point #1) and have developed new technical capabilities as a result.Example #2: Our advancement office (alumni donation team) spends many hours manually scouring obituaries to determine if the deceased are alumni of our institution. If they are, we send flowers, cards, and condolences to the survivors, and (later, eventually) ask for donations. Using natural language processing, the team was able to create an obituary reader that can scour obituaries in the local papers and national databases and determine with a high degree of accuracy whether 1) the deceased is one of our alumni, and 2) the surviving partner, spouse, children, and grandchildren of the deceased. This saves lots of labor hours and helps automate a manual process. Our student enrollment office became very excited because identifying legacy relatives could potentially help them meet recruitment goals more easily. So the outcome here was much more positive.Both of these examples were fairly simple to start and build on. It is interesting that the basic text reader might be the huge win for us at this time, rather than the video/emotion detector. A big part of that is understanding what your organization/field is ready for at the time.Good luck!Global CIO in Telecommunication6 years ago
Create a PoC and demonstrate with live business outcome (Roi) in real business environment VP of IT in Services (non-Government)6 years ago
I think it depends on the industry. Some industries are moving faster than others. If the company in question is in the financial/tech/telecom industry then they are behind. If the company is in healthcare/manufacturing/consumer then it’s not such a burning platform.The most important thing to do is to define what we mean by AI (see picture below). Once you start to break “AI” down into digestible pieces of what it is you can start to segment initiatives.Once you have clarity on what it is then you need to measure the level of disruption it is likely to have on your industry and how quickly.Chances are most companies are doing some sort of AI (broadly defined) today, it’s just buried in proof-of-concepts or other analytics work. So I it’s important to highlight those efforts.Then of course one must start to chart a strategy and roadmap towards enabling the business around AI (and digital more broadly).CIO, Managing Director IT - Retired in Finance (non-banking)6 years ago
AI is a buzz word for intelligent programming. Every decade we coin old terms and call them new. Object oriented programming, service oriented architecture, Cloud just to mention a few. What is important is to define the problem then leverage the best available tools to solve it. You can call it AI when it is done but what is important is that you solve the problem.