The Realities of Machine Learning for the Benefit of Tech Leaders

January 14, 2019 Tyler Warden

In the cycle of technology buzzwords, Artificial Intelligence (AI) and Machine Learning have become part of the regular vernacular used to make technology seem more relevant, modern and innovative. In reality, these tools can either make solutions better or complicate and frustrate well working processes. In order to empower a business with Machine Learning, the technology needs to transform from buzzword into reality – this journey can be achieved with the right investments in targeted areas with ROI measurements attached.

Why Machine Learning Matters
Organizations are generating too much data, from too many systems, which moves too fast for humans to keep up amongst other priorities like budgets, career ambition and business goals. Also, it isn’t sustainable for business analysts, Centers of Excellence or business users to correlate and extract value from all of this data given limited time and skill set constraints. And with typical user bases wanting to manage by exception in order to prioritize high-value, business-moving activities, machines can help lead the way. Machine Learning can provide more brain horsepower to organizations at scale to augment existing user volume and skills. By combining machine brains and human brains, insights, business processes and innovations can expand like never before.

Machine Learning algorithms and investments can also provide tangible, differentiated value. With all of the talk about “Digital Transformation”, organizations are looking for ways to leverage computing technology to create competitive advantage. It is long established that bringing the right talent into organizations can create real competitive advantage as well as both top line and bottom line benefits. The same can be said about Machine Learning algorithms, training sets and processes. With the right investments, talent acquisition can now come in the form of machines.

How Machine Learning Becomes Real
In order to reap the benefits of Machine Learning in the real world, teams will face the common “build, buy or partner” decision. Many technology platforms on the market have Machine Learning capabilities built into them, which are designed to solve a variety of use cases. Generally. these solutions fall into two categories:

  1. Solutions that use Machine Learning to solve targeted problems
  2. Technology that provides the tooling and backbone to create and curate custom Machine Learning algorithms, models, and processes

In the first category, the best solutions will claim to have Machine Learning capabilities as well as offer embedded tactical uses cases demonstrating how Machine Learning makes their solution better. Spend the time to drill into how Machine Learning makes the processes executed by the solution better and how those Machine Learning capabilities are evaluated over time to ensure the proper training is put into the algorithms and that bias is worked out of the models over time.

In the second category, investing in tools and a platform that facilitate the creation and stewarding of tailored Machine Learning processes allows for the creation of powerful, differentiated Intellectual Property for an organization. These platforms can range from powerful Data Science workbenches that require highly skilled data scientists to code and deploy algorithms to more business-focused applications for the design of tools like chatbots. When going this route, the most successful implementations are ones that improve the current behavior of users in business processes by applying targeted Machine Learning to resolve specific user challenges.

Trusted Data Leads to Trusted Machines
When bringing Machine Learning into an organization, the level of quality and trust of the data used to train the machine has a direct correlation to the trust that can be put in the output of those Machine Learning processes. Machine Learning processes and pipelines typically execute with little to no human interaction until late in the process – and often only at the time of output. This reality is quite different from more traditional transactional processes and analytics where humans play a hands-on role during the processes’ execution and therefore have the ability to manually apply high reasoning abilities against the data and almost naturally deal with data inconsistencies, faults and confusions. When machines are working with the data, they function best when the data is already well understood and high quality – both from a semantic and technical perspective.

As Machine Learning becomes embedded throughout enterprise solutions, the ability to enable trusted data from inception through removal becomes even more important. When working with Machine Learning in a Data Science construct, a common approach is to dig a metaphorical “moat” around the Data Science laboratory in order to establish a strong Data Governance process around what data goes into the lab and who can access the results outside. While this approach works well in the case of Machine Learning developed by dedicated teams in house, it quickly becomes infeasible to draw the same kind of hearty boundary around all data across all systems. Instead, a pragmatic approach is recommended: as data evolves along its journey, the policies and rules that govern it are strategically enforced to ensure the right investment is made the right time based on business impact. For example, the techniques used to ensure high quality and protected patient data in a hospital setting will likely be more robust than the enforcement strategy around data trust for real estate data for corporate offices.

Many organizations are investing in major transactional systems, ERP systems, CRM, HCM, PLM, and analytics environments. In these cases, performing a data migration can help ensure the data that goes into these systems is of the highest trust and quality so that any embedded Machine Learning functionality will yield the greatest benefits possible. Once live on these new systems, organizations need to maintain the data quality and trust levels via ongoing Data Quality and Data Governance initiatives.

Machine Learning presents opportunities for real innovation both today and well into the future. In many ways, it offers the promise of technology delivery catching up to the ideas of the innovators and meeting the needs of users. By strategically implementing Machine Learning and empowering those machines with well trusted data, enterprises can drive meaningful competitive advantage from their technology organizations as a center of innovation.

> Click HERE to read the original article on DATAVERSITY.

About the Author

Tyler Warden

As VP, Product Management Tyler is responsible for the product strategy and roadmap of BackOffice Associates' software products. These software products include a platform for data stewardship, Data Migration, Enterprise Data Quality, Master Data Management, and Information Governance. He is a member of both the Strategic and Product leadership teams and is instrumental in the development and driving of the product strategy and vision. In his time at BackOffice, Tyler has worked in all areas of R&D as well as at customer sites giving him a unique perspective on the needs of the Information Governance market and the current and future needs of customers. A leading voice in the data management and governance industry, Tyler is a common contributor to industry publications as well as a speaker at conferences around the world.

Follow on Twitter More Content by Tyler Warden
Previous Article
Trend-Setting Products in Data and Information Management for 2019
Trend-Setting Products in Data and Information Management for 2019

Information Governance Cloud, included on the DBTA list of Trend-Setting Products for 2019, combines AI and...

Next Article
3 Keys to Data Modernization
3 Keys to Data Modernization

Three strategies can help with data modernization efforts, says Rex Ahlstrom of BackOffice Associates. Lear...