Source: Deep Learning on Medium
Winning AI strategies for organizations — Some tweet thoughts on a MIT chat
On 23rd October 2019, I participated in a live Twitter chat (#MITSMRChat) t on ‘Winning Artificial Intelligence (AI) Strategies’ organized by MIT’s Sloan Management Review. I got an opportunity to interact with a diverse group of brilliant AI practitioners and answer some questions. These are shown below .Each answer below was shared as a tweet, thus prizing brevity & clarity of thought.
How does one handle data cleaning problems for AI — I would prefer to call these as data challenges rather than as data problems. The organizational data hoses will keep getting larger. Big Data is a reality. Generic frameworks are needed to be developed for easily mapping organizational data for AI usage.
Where does one start an AI project in an organization ? — One may look at the methodology suggested by Professor Ajay Agrawal to identify the organizational tasks & workflows where AI generated cost-effective predictions could have maximum impact. This could be a good starting point across business domains & functions in organizations.
Should organizations start with a Big Bang AI project ? Andrew Ng who is considered as one of the pioneers of modern AI suggests the exact opposite. A Big Bang AI project is a highly risky one. A tiny, yet strategic project — from which learnings can be scaled both horizontally & vertically — usually has a higher chance of organization-wide incorporation & success. Andrew Ng has referred to many examples based on his past Google AI experiences to illustrate the same.
Isn’t the hype around AI similar to the one around Y2K earlier ? Yes & No. The essential difference is that Y2K was a clearly defined problem domain with a fixed date. AI is an ever-expanding field at the forefront of cutting-edge research, leading to innovative usage across multiple domains and development of new benchmarks almost daily. However, as with any new technology, expectations and hype must be realistically managed.
How does one assess the competitive landscape for pursuing an AI strategy ? I would suggest NOT to look at competitors first. AI has many untapped opportunities for usage with the core organizational stakeholders — customers, suppliers, owners, employees & regulators. The fundamental strategy would be to observe and assimilate their pain points & leverage the same for an AI driven process approach.
AI should perhaps be like the air we breathe — pervasive in every strategy & thought process, yet working so silently that we are unconscious of it’s invisible presence.
There are many frameworks available from the pre-AI era that will work for any competition landscape analysis even today. AI is a technology intervention whose impact can also be easily measured in already available business metrics.
How to best assess whether AI in a business context, e.g. product/service or customer experience or operational efficiency, is a source for Competitive Advantage vs. approaching Basic Competitive Requirement ? With the fast pace of AI development, it will not really matter as competition eventually catches up. So, the only answer is to keep pushing AI both across scale & depth in a business context. Also, like any other IT project, there must be metrics to measure ROI (Return on Investment) success.
What are the organizational alignment, data sharing & collaboration challenges for AI ? Alignment reminds me of the story of 2 friends in the jungle being chased by a tiger. ‘I only have to outrun you’ says one to the other. A smart CEO know that some parts of her or his organization will be eaten up by the AI tiger if she or he does not act. Perhaps, fear can be the strongest alignment.
Functional or other organizational unit heads must be motivated to share their internal data hoards for the cross functional AI projects. Quantitative metrics which help measure these data sharing efforts across the organization are needed. This will help create organizational incentives for AI success.
As in any project involving people, collaboration is a huge challenge. Specially with AI, because at times, stakeholders may be nervous about the outcome as it may impact their future work routines. Communication as always is the key. Learning & unlearning too is crucial for success.
Does AI only impact knowledge workers ? It is only a matter of time when fast paced AI advances will also impact physical jobs by super fast development of knowledge wrappers around their processes & tasks. The recent OpenAI Rubrik Cube development is a case in point. This does lead to a challenging scenario on re-skilling & the jobs front in the short run.
What type of organizational leaders are needed to harness AI ? Going ahead, AI capabilities will get embedded in software systems & hardware devices. So, the key is to have organizational leaders who not only understand technology but also it’s interplay with the various business functions & domains and the empathy to carry along people at different stages of their respective individual AI learning curves.
Does AI play just a supportive function within the organization ? Not sure whether only a purely support role can be played by AI. Organizations need to have AI embedded capabilities within all business functions. So, AI becomes an integral part. Of course, the success of the same depends on the quality of data maturity within the different business units.
Is an organizational AI effort a marathon or a sprint ? I would call it as a marathon at the speed of a sprint. Global competition is relentless. As an example, the pace of implementation of AI projects in China both across the government & private sector is staggering. Any innovation spreads pervasively & efficiently. This is today’s brutal reality.
Postscript : While searching for suitable photographs for this article from my phone, I queried my phone for the word ‘Winning’ which is part of the title of this article. My phone’s AI assistant returned the following image.
In addition, my phone’s AI assistant offered to copy text from the above image which it did so successfully as shown below (saving my typing efforts) :
Winning Entrepreneurial Strategies
Responsive to Market
Fast Product Development
Great Teamwork-Delegate, trust
Fierce Competitor — Sun Tzu Art of War
Lose the Battle, not the War
Personalize your Customer relationships
Learn-Teach Team to learn
This capability of understanding text from images did not exist a few months back on my phone.
In conclusion, AI is learning and learning really really fast. We humans need to buck up for the challenge.