February 28, 2018 | Written by: Hemant Kumar
Categorized: Industry Insights | Upstream Petroleum
I have been tracking AI (Artificial Intelligence) initiatives, for the last few years, across the business sectors through conferences, readings, personal experiences and my network. There is a distinct pattern about the approaches firms adopt towards AI. Some firms jump headlong into their AI initiatives, letting the chips fall where they may. On the other end, some stay too long in limbo, trying to work out every possibility before they take their first step. Evidences suggest that both the approaches produce underwhelming results; the former leads to the false starts while the latter delivers the over-wrought initiatives, stumbling along, till they finally get killed.
But then there are quite a few firms that manage to find the sweet spot between over exuberance and over prudence and not surprisingly, they also tend to have, as it were, better running AI projects.
As AI fever grips more E&P (Exploration and Production) firms and more E&P executives approve AI initiatives, they need to find their own sweet spot to get the best out of their pursuits. This is not a trivial task considering a hyperactive market fueling the AI frenzy and blurring the line between hype and reality. Moreover, AI is still at an early stage in the sector and there is limited appreciation of the risks and challenges that go with its implementation.
If you are a manager or functional lead nominated to oversee an AI initiative in your firm, you may already be feeling the pressure of managing the high expectations while delivering a successful outcome that will jumpstart your firm’s AI journey.
Assuming that’s the case, you may consider some of the following lessons, from the battled-hardened AI warriors, that could help you put your project on a sure footing and improve the odds of success.
1- Get acquainted
Over the year, the term AI has taken a life of its own and now become a catch-all phrase for many things such as machine learning, machine intelligence, deep learning, various forms of analytics (descriptive, predictive, prescriptive etc.), Big Data, IOT or even Industry 4.0. While some of these are just different terms for AI, some are far from it. The fact that not many can tell the difference has created wrong notions about what qualifies as AI. It has also given opportunities to the vendors to package their traditional analytics and pass it off as AI. The risk it that creates is that firms may start a project believing it is AI when it’s not. Or they may target the wrong problem to solve with AI.
It’s good if you are already up on the learning curve, but if not, then it’s not too late to take a crash course in AI—not with an intent to build the neural networks but to familiarize yourself with how it works and what its limitations are. This will put you in better position to see through the hype and ask the right questions to the vendors.
There are plenty of avenues for you to tap into, such as academia (e.g. MIT, Stanford) courses in public domain, YouTube videos or free awareness workshops offered by the vendors. Firms with their own data scientists or digital officers will obviously be more enlightened, but if you don’t have those, you can form your own team of AI enthusiasts who can be the voice of reason for your firm.
2- Start quick
Any digital transformation consultant will tell you about the value of right strategy, process design and roadmap before you press forward. They are all important but can also slow you down if it takes the consultants months to produce a bunch of PPTs and word documents.
Truth is there will never be a perfect strategy or roadmap with AI as things will evolve constantly and new knowledge will emerge in terms of what technology can do (or what it can’t do). Instead of spending too much time on a months-long strategy piece, it’s more effective to hit the road with a couple of ideation workshops, involving the right stakeholders. Target 3-5 opportunities to get the fundamentals of a 2×2 matrix then press go.
Focus on identifying the potential use cases and a candidate for the pilot project. Build a high-level roadmap and refine it as you go along. It’s normal if you notice skepticism from some people about the quick and dirty approach. You are better off showing value through something that works than “costly” PPTs. So deliver “that something” quick and let the results speak for themselves…which brings me to the next point.
3- Start simple
A typical pitfall in AI initiatives is to try to take on too much or start with a complex problem. Every AI project is an experimentation and experiments fail sometimes. Nothing hurts like a bloated AI project that goes on for months and then fails to deliver.
Starting with a simple scope helps you quickly test the feasibility of the use case while minimizing the potential losses should there be a failure. On the other hand, success will get you on firm ground and help you sell your initiative better. It also will give your team the necessary confidence and momentum to keep going. If, at all, your use case must be complex, break it down into phases and make the first phase simple. If you can’t deliver your first phase in a few weeks, then it’s still complex.
4- Take emotions out of your business case
Watch out if you are feeling too high about your initiative or acting under unrealistic urgency. Blind spots can creep in when stakes (personal or organizational) are high.
Problems that may follow can range from force-fitting AI into the business problems that are better solved using alternative methods, to selectively picking up the data points to build a business case, to even overestimating the business benefits.
You may think none of the above apply to you and you have an airtight business case, but you never know. It wouldn’t harm you to get a cold-eye review by the people who are still not infected by your enthusiasm. Let them chip away at the business case till you have whittled it clean of all the unknown assumptions. If the numbers don’t add up, iterate and re-examine or reconsider.
5- Set the right expectations
With so much buzz around AI, your stakeholders may unwittingly form very high expectations, especially when they are not familiar with how the technology works. On top of that certain individuals may have different views on what a project will deliver. This creates potential for dissatisfaction with the project outcome.
You must educate your key stakeholders on the expected project outcome and what success will look like, and communicate it at regular intervals. Have an open discussion with your AI vendor and agree on the success criteria. Surface all the assumptions, risks and dependencies, in unambiguous terms. Insist on the vendors to bring in their subject matter experts so that you get informed opinions and not just a sales pitch.
6- Get your best SMEs on the job and listen to them
A common wisdom in the E&P sector is that involving SMEs (Subject Matter Experts) too early or too much slows down an initiative. It’s not completely unfounded. SMEs either are too busy with their daily jobs or don’t share the enthusiasm of the project sponsors.
There is no getting away from the SMEs when it comes to the AI initiatives, nor should there be. SMEs are not just the lifeblood of the firm, but also their best ambassadors. Make them part of you project from the word go. Only they can help you validate the use case, build the business case, assess data quality, train the AI system and “nurture” it once it’s live. And not just any SMEs but your best and most experienced ones—an algorithm is only as good as the training it gets. Although it may create some change management issues early on, once you clearly communicate the purpose and value of the initiative, your SMEs will come to the party.
Be open-minded when your SMEs have opposing viewpoints on a use case. If they don’t see much value, then probably there is none regardless of what you feel about it. It’s time then for you to move onto the next use case.
7- Secure leadership commitment
Enthusiastic talks and rousing speeches about the technology in the meetings or the conferences are necessary but not sufficient evidence of leadership commitment. If the senior executives are not on governance board of the project or actively involved in it, you may ultimately face challenges. While things may go well initially, you will invariably come up against the roadblocks such as not getting data, access to systems or resources. Having leadership behind you will help you clear the obstacles faster than without them and save yourself a lot of time.
As the cliché goes, well begun is half-done. A good start, although not a guarantee of success on its own, will give you a lot of confidence in navigating your AI initiative through the inevitable growing pains.
Visit https://www.ibm.com/industries/oil-gas/big-data-analytics to learn more.