• Ben Lowe

Review of Using AI to Enhance Business Operations from MIT Sloan Management Review, Summer 2019

Enterprise Cognitive Computing (ECC) is a relatively new capability that enables an organization’s business processes and combines it with cognitive computing capability more commonly called machine learning, artificial intelligence (AI) and deep learning. [1] For businesses embarking on using ECC, the main issue is how can it be applied to business operations successfully and create value. In the Summer issue of MIT Sloan Management Review, 2019 issue, the question is posed in which five capabilities and four practices are described needed for success. [2]

Five Crucial Capabilities

1. Data Science Competence – this encompasses a wide range of skills involving cleaning, curating, tagging, collecting and analyzing internal and external data from multiple sources. An ECC group needs to have competency in AI, machine learning and various other analytical techniques. This organization is either created internally and/or externally through hiring directly or by employing a consulting agency.

2. Business Domain Proficiencies – to actually get value from the analytics you need people who have a clear understanding of the relationships between the data and the business processes in order to curate, tag and analyze the data correctly. That merely prepares the data to be used. Additionally, the implementation of the analytics requires an understanding of the tasks, workflows and logic of the business processes so that the AI algorithms and analytics incorporate the regulatory rules, business/process constraints and practices to ensure they drive business value.

3. Enterprise Architecture Expertise - as the ECC application is put into place there may need to be changes in the organization. Depending on the scope and scale of the change, you may need expertise from change management people, IT professionals who are knowledgeable in streamlining processes from a systems point-of-view and human resource professionals are exposed to organizational roles, role design and skills training to drive the change in the organization.

4. Operational IT Backbone – this encompasses the company’s technology and data foundation where you need a group of people who have the responsibility to support the ECC application. This will consist of the capability to store and access the critical data, integrate with other applications and ensure reliability, security and privacy for the application.

5. Digital Inquisitiveness - the ECC group has to have a natural inclination toward data and digital inquisitiveness and have their AI algorithms evaluated by the business, as well as, further enhanced by human judgement, which may take time. Equally as important is to level set the other parts of the organization--the leaders, managers and others who are impacted by the AI algorithms. They need to understand the output of the ECC solution and know the impact of poor data quality, the building of decision trees, and how the algorithms can be taught and improved over time.

Four Key Practices

1. Develop clear, realistic use cases – a use case needs to be developed with a clear scope of the ECC application and clearly defined what the AI application will do, what expected enhancements the application will need and the outcomes of the business process. The use case should also describe future benefits the AI application can provide, if further developed. The use case should facilitate the identification of the overall cost and benefits of the application.

2. Manage ECC application learning – there is a potential that the business application will change in which there needs to be a “drift” factor of the ECC application requiring monitoring throughout the life cycle. Built into the application should be a reporting mechanism that can generate alerts to ensure the outputs are aligned with the business goals. If the application drifts, then domain experts and data scientists along with their IT counterparts need to work together to identify, access, clean, tag and architect the system to improve results.

3. Cocreate throughout the application life cycle – AI applications are created jointly between the data scientists, IT professionals and business domain experts. Ongoing improvements involves all groups to maintain, monitor and improve.

4. Think “cognitive” – once the ECC application is implemented it needs champions to create excitement and to promote use. The response to a new AI application can be varied throughout the organization from fear to exaggerated expectations so an effort to level set is important to gain the “buzz” and adoption needed, and also ensure the organization has digital inquisitiveness.

Just some comments on this article from a practitioner point-of-view that may help an organization ensure AI application success:

1. Data in organizations is unmanaged or not clean – the implementation of AI applications assumes the data is clean; otherwise, there is the all too common, “garbage in, garbage out” leading to spurious results and resulting in low adoption. So many organizations that do not have a master data management program, will be ill prepared to even undertake an AI program or form an ECC group. Data clean up and management can mean restructuring the organization to make sure the data is clean and useable. Establishment of a data governance group is a good start which involves master data governance.

2. Data is non-existent or in silos – in many cases as a data scientist begins their project they find that the data they need has never been collected in which they may have to work with the business to collect the data for the work to begin. This may require the inclusion of sensors or creating new interfaces. In other cases, if the data is in existence, the data is found in silos (in people’s computers, in factories, in business units…) where retrieval can be tough. The article implies that there is a dependence on an IT organization to retrieve the data and do most of the data preparation work. A better solution is to create a dedicated data engineering team that caters to the data scientist to do the preliminary setup of data acquisition, clean up, storage, business model development and create the ability to display the data. This expedites the data preparation phase and the likelihood of success. As the ECC application has been fully tested, then it can be moved to IT to put into production.

3. Business intelligence, storage, data retrieval is expensive – With the advent of big data, the cloud, AI, machine learning and deep learning, there has been a radical improvement in our ability to analyze and do great things with data, but there is a cost to this and one that should not be underestimated. Development of AI algorithms assumes the ability to store, retrieve, program, and test to deliver results and put into production. This requires a development, testing and production platforms. IT needs to be involved along with a data engineer to scope out this effort and execute. Do not underestimate this effort and can be more costly in time and money than expected.

4. Talent acquisition – as more and more companies are creating ECC groups, the talent pool is limited as data analytics programs have a finite capacity and have only been around for a few years. Demand is high and people are scarce so expect finding qualified data professionals difficult with protracted hiring and difficulty in retaining data science employees.


[1] M. Tarafdar, C. M. Beath and J. W. Ross, "Enterprise Cognitive Computing Applications: Opportunities and Challenges," IT Professional, vol. 19, no. 4, pp. 21-27, 2017.

[2] M. Tarafdar, C. M. Beath and J. W. Ross, "Using AI to Enhance Business Operations," MIT Sloan Management Review, vol. 60, no. 4, pp. 45-51, 2019.

26 views0 comments