Where to Place Your Machine Learning and AI Bets

A few weeks ago, I published my 2020 predictions and stated explicitly that enterprises must demonstrate AI and Machine Learning wins. So, in this post, I survey several recent research surveys to determine where organizations should place their bets.

What machine learning and AI will you invest in?


I looked for three indicators:

  • Is the timing right for medium and late adopters to get in the game, and how much must they wager?
  • Should the bet go to red (cost savings) or black (customer experience and revenue)? Or both?
  • Where should organizations invest their skills, time, technology, and money?
My summary from six recent surveys and my takeaways follow.


1. Place big bets in machine learning now - or risk falling behind faster


AI/ML investments are increasing and so is the adoption:

  • Only 23% are mid-stage or sophisticated adopters of machine learning with models in production 2-4 years (mid-stage) or 5+ years (sophisticated). 38% are just getting started and are either evaluating use cases or starting to develop models.1
  • 70% increased their AI budgets from 2018 to 2019 with 28% of respondents stating a budget increase of 26% or more.1
  • In a survey primarily covering the financial services and healthcare industries, over 65% are investing over $50M/year in Big Data and AI.2
  • Salaries and sizes of data science teams are driving up the costs with over 50% of respondents of one survey reporting data science salaries between $100-200K and 25% reporting being on data science teams with twenty or more employees.3
  • Industries with the largest gains in embedded AI capabilities from 2018-19 are retail (+35%), travel/transport/logistics (+26%), and high tech(+17%).4
StarCIO Takeaway: Bet now and bet big. If you haven't started or in the early stages of big data, machine learning, or AI journeys, then you are probably behind your competition. That goes for small and medium-sized companies as well. And companies diving into this space are taking sizable investment bets, so limping in probably doesn't get you very far. 

2. Cost savings may be the low hanging fruit, but more considerable opportunities in customer experience and growth


  • Most significant revenue impacts are in marketing and sales, where 80% of survey respondents report increases, followed by product and service development at 71%. The two highest cost savings opportunities are in manufacturing (64%) and supply chain (61%) of respondents citing savings.4
  • Top machine learning use cases that have revenue potential focus on generating customer insights (37%), improving customer experience (34%), and retaining customers (29%).1
  • In a survey of large banks, 59% report using deep learning to improve customer satisfaction, but their ROI is more focused on cost with 80% looking to improve margins and 57% on reducing capital expenditures.5
StarCIO Takeaway: Organizations investing significantly in machine learning must look beyond cost-saving opportunities. This is particularly critical in competitive industries where other factors are contributing to digital disruption. In addition, organizations that overly focus on cost savings may see more internal resistance to change, especially when jobs are impacted. 

3. Follow the leaders when selecting technologies, machine learning algorithms, and tools


  • Jupyter is the overwhelming favorite IDE followed by RStudio and PyCharm. The same respondents site more traditional machine learning algorithms such as linear/logistic regression and decision trees/random forest as top methods. Their top platforms are AWS and GCL.3
  • Top machine learning libraries sites in a student survey are TensorFlow, Scikit-learn, Keras, and Pytorch in that order.6 But the top two cited in the Kaggle survey are Scikit-learn and Keras.3
  • Beyond machine learning, other top AI capabilities receiving investments include robotic process automation (RPA), computer vision, and natural language processing (NLP).4
StarCIO Takeaway: It's a messy playing field of platforms, technologies, and algorithms, and while there are some leaders, we're not down to a two or three-horse race yet. AI/ML leaders should expect significant ML Debt when consolidation and simplified implementations are warranted, while laggards should probably stick with mainstream approaches.

If you are just getting started with ML/AI, consider reading my post on 10 questions before starting a machine learning POC and 3 must-read AI books. Experts may want to review my InfoWorld article on 5 takeaways on scaling machine learning.

And if you want to chat about AI/ML or becoming a data-driven organization then please reach out to me!

References to AI and Machine Learning Surveys Sited


Please see the research below for more details!


No comments:

Post a Comment

Comments on this blog are moderated and we do not accept comments that have links to other websites.

Share