Business opportunities & challenges posed by rapid advances in Machine Learning

By Pritam Dodeja / Mar 20, 2023

Managing routine rapid advances in Machine Learning

It seems that with each passing day, some new advance is celebrated in the mainstream media about what Machine Learning can now do.  Even before the ink has dried on the press release, the next best thing is released.  It is natural to experience a “fear of missing out” moment, to get lost in the hype.  However, to fully take advantage of the opportunities and surmount the challenges posed by the era of Artificial Intelligence, in our opinion, it is the “first principles” based approach that will differentiate the successful organizations of the future.

Core Principles:

1. Your data is now more valuable than it was before as more insights are now reachable.

2. The value proposition of your organization can be amplified by the application of machine learning only if:

a. Your data is well organized and governed.

b. Business cases have direct line of sight into outcomes that can be enabled by Machine Learning.  You are not getting lost in the hype cycle.

c. You apply sound engineering principles to build out Machine Learning capabilities using best practices borrowed from a variety of more mature disciplines.

3. Your organizational culture needs to recalibrate around the new paradigm wherever applicable.

4. Human creativity and expertise will continue to be highly valuable and sought after.  There will be productivity gains at every level of knowledge.

Data: An Appreciating Asset

It used to be that business transactions would occur, captured in some system, moved to another system, analyzed by humans, some insight derived, and an action taken in response.  Now the say activities, with better algorithms, can be performed as the transactions are occurring.  In some cases, they are informing the type of transaction that should occur in the first place.

The reason for this is manifold; data processing capabilities have matured greatly due to the advent of cloud.  Machine Learning algorithms can easily fit into those data processing pipelines and provide real-time actionable insights.  However, the insights they can provide are limited by the quality of data that they can be trained on.  Data Governance is required for high quality inputs into Machine Learning algorithms.  The enrichment of this data into higher fidelity signals is a joint responsibility of human engineers as well as neural networks that can spot patterns that humans cannot.  If the data quality is lacking, then the insights produced are no longer applicable in the real world.  If there isn’t enough data to train the algorithm, then the insights are not reliable.  The reason for this is that Machine Learning algorithms are just fancy mathematical functions that predict future outcomes by interpolating on the data they have been trained on.  This also means that protecting your organizations data becomes even more valuable, as a competitor, with access to your data, can create the product that your R&D team would have created.

Amplifying the Value Proposition of Organizations

Whether for profit or not, each organization serves a mission, which in some form provides a service to society.  This involves some value generating activity where there is a creator and a receiver.  One common way to use Machine Learning is to understand the customer better.  There are a variety of Machine Learning algorithms that can spot patterns in user behaviors that can help with creating the right offerings for that customer.

Another common pattern is intelligent automation, to remove the drudgery from work by having Machine Learning algorithms create and process documents as part of a business process.  Now algorithms can understand spoken language, translate it to another language.  Algorithms can see images and make sense of them, take actions based on them.

However, the key to success will always come down to first principles: Are we, as an organization, understanding our customer well, creating the right types of products and services, offering them at prices that are competitive, and are we able to sustainably do this.  A knowledge of the domain and the right application of that knowledge will still be a core human activity, but the activities below this can be intelligently automated to drive efficiencies, and expand the reach of products and services.

To implement these Machine Learning algorithms at scale is an Engineering challenge that involves many different aspects.  Engineering data quality, enriching features with domain knowledge that helps the models to learn faster and better, finding the best set of parameters for a particular domain, training the models, verifying they work as intended on unseen data, monitoring them on an ongoing basis as the business landscape changes.  Doing this well requires rigor across multiple complex disciplines.

Strategy: What Culture’s having for breakfast

Just as it is important to substitute the Engineering cycle for the hype cycle, it is important to create an organizational culture where the concept of failing doesn’t exist.  Only learning exists.  Formulating a culture that takes calculated risks in Machine Learning projects, that looks at data as a precious asset, and one that appreciates how complex and new this all is key to achieving differentiated business outcomes.  If projects, by definition, must always succeed, then organizations do not do difficult projects; they do the safe ones.

Business and domain knowledge shaping the business cases and contingency plans, coupled with Product Managers and Engineering resources that formulate and translate the problems into ones that can be solved by Machine Learning will be critical in creating a culture where Machine Learning is possible.

Creating a community of practitioners that span the entire business landscape will accelerate the application of lessons learned in one part of the organization to the other.  Some Machine Learning algorithms do this internally and applying it at the organizational and cultural level will reshape the types of business use cases that are pursued via programs and projects.  To think big, but act with rigor and discipline will be key.

Human Creativity: Will always be highly valued

There is a lot of creativity required in the creation and application of Machine Learning.  And so, it is also with business problems.  It is the humans that currently know the business domain the best, and can come up with novel application of systems to solve human problems.  That creativity will need to be nurtured in organizations and it can lead to the products and services of the future.  Machine Learning will reduce the drudgery through intelligent automation, but it will not replace intelligence altogether in the foreseeable future.

Conclusion

Machine Learning will dramatically transform industries through intelligent automation and augmenting human intelligence.  As a discipline, it is young, however, it has the potential to help organizations understand customers better, create more compelling products, reduce the wastage in producing those products, and drive profitability, efficiency, and effectiveness.  The potential can only be realized if old fashioned engineering applied to the latest advances is combined with organizational and cultural transformation.

There are plenty of industry specific and industry agnostic use cases possible today to help accelerate your Machine Learning journey.  Let’s talk about how we can separate the wheat from the chaff for you.