AI Strategy and Planning

Utilizing AI for revenue growth, automation, efficiency improvements, customer service enhancements, competitive advantage by offering groundbreaking products and services are some of the key reasons why AI has become critical in the digital transformation journeys of many organizations. In the following sections, I review some core AI digital transformation strategy factors including vision, ROI, how to build and implement, the overall impact to the organization, privacy/security, and the required AI tools & services.

According to an Accenture report 75% of C-suite executives believe that if they don’t scale artificial intelligence in the next five years, they risk going out of business entirely.

Define the vision, ROI, and success metrics

An organization’s top leadership needs to have a clear vision of why and how AI will be used to transform the organization. This vision is typically driven by the need to enhance operational agility, automate, or compete in the market. To build a well-thought-of and informed AI vision requires a thorough analysis of the current business use cases and additional uses cases that can be unlocked by using AI. At the end of the day, the vision should lead to a strategy that is aligned with the overall goals of the organization.

According to Cognizant report 73% of executives believe that AI will help them evolve their business model by enhancing the ability to introduce new products/services or by creating new businesses in the next three years.

Creating a clear vision typically involves the CEO or the CIO working closely with all the business units in an organization to get alignment across the whole organization. IT unit usually takes the lead in technology selection, installation, configuration, and maintenance. Most of the time it is the IT unit that takes on bold technical initiatives, thus the role of a CIO is important in understanding how any technology, including AI, will help the organization in meeting its business goals.

A clear vision along with an ROI defined upfront is a recipe for success. Any digital transformation initiatives which involve AI should also have success metrics defined upfront. These AI-based systems, with telemetry configured properly, can provide insights, and required metrics to quantify the ROI. Once the vision is clear, it is important to communicate that to all the layers within the organization. AI-based systems can be quite disruptive to the usual way of how work is done in any organization, and it is important that everyone in the organization understands why AI is being used and how it is going to benefit the internal employees, customers, or the organization.??

According to a McKinsey report 22% of the businesses in 2020 reported at least 5% EBIT (Earnings Before Interest and Taxes) because of AI

Build, buy, outsource, or partner

After the vision and ROI have been clearly defined and communicated through all the layers of the organization, a decision needs to be made if you should build, build, partner, or outsource the AI system.

Build – Customized

If the proposed initiative will be unique to the organization’s needs and there is already an in-house data science, machine learning/AI developer, and other required resources available, then you may want to build your own customized AI system utilizing these resources. Depending on the scope of the AI initiative the AI team will have to collaborate across various units of the organization. Building a customized in-house solution is the most flexible option available, but usually is the most expensive to maintain for the long-term since you must retain the data scientists, ML/AI development engineers, and other resources.

Build – No-code & Low-code

All the top cloud providers are offering many no-code/low-code AI building tools and services, such as the Azure Machine Learning designer, as part of the Machine Learning Studio. These tools can be used to quickly start a prototype and build a production AI/ML-based system. These tools typically don’t require deeper machine learning expertise and, in some cases, no coding experience either. This ease of use, however, comes with minimal flexibility in building a fully customized system. While utilizing these tools, machine learning and AI developers can tweak the code, a data scientist may still be required to build a fully customized AI system.

Buy

There are certain common business use cases AI applications and services available that can be purchased and easily customized to suit an organization’s needs. These out-of-the-box AI solutions can be purchased and with minimum configuration, you are ready to go. Although such solutions may work initially, as you or your customers use them, there may be a need to customize them in the long run. This is the least flexible solution since there is very little customization that is allowed for such AI-based systems. For any extensive customization, if it is allowed, you may have to work with the original vendor who created the AI system/service or bring in an external consulting vendor who is experienced with this AI system.?

Outsource

Outsourcing will be an option if the organization lacks both the availability of data science and ML/AI development talent and other necessary skills to, build, deploy, and maintain AI systems. This is the costliest of all the options for creating AI systems and applications. This is the best option to get started quickly while you can work on hiring data science, ML/AI developers, and other resources to help with the long-term management and maintenance of the system.?

Partnering

This is another option to consider if you are part of a large organization. You may be able to partner across the units in an organization and combine your resources to build an AI system. One unit may be able to help with developer resources, while another unit can help with data science requirements and yet another unit can help with the overall program and project management of the system. This kind of cross-team and cross-organization collaboration is very common in a large enterprise where resources and talent are spread across various units of the organization. This option, however, is hard to manage and sustain because you have to work across different units for the long run.

Assess the impact on people, processes, and technology

Before you bring in an AI-based system, either internally or to your customers, you should carefully assess the impact on the three key pillars of IT: people, processes, and technology.

People

AI systems are generally perceived as something which will replace all jobs and that is not true at all. While most of the AI systems look to automate most of the manual tasks within an organization and improve operational efficiencies, they are also responsible for creating many jobs for managing and maintaining these new systems. An AI strategy should carefully examine the impact of AI on people, including both internal employees and external customers. Such an assessment should include jobs losses that may occur, jobs creation, and any retraining that needs to be done. Retraining your current workforce is a good way to minimize the number of job losses that may result from the introduction of an AI system within an organization. The retrained workforce can be utilized to manage and maintain the AI systems along with possibly building the next version of the current AI system.

According to a McKinsey report, the survey respondents at high AI performing companies are 2.3X more likely than others to consider their C-suite leadership very effective.

Processes

Two of the key reasons why AI-based systems are built are to improve operational efficiencies and to bring agility to the current manual processes. One of the fundamental advantages of AI is the automation of manual processes which may have existed before the AI system was built. With AI, there will be significant gains in efficiency in processes leading to faster business results. This will also lead to a shift in business culture because AI-based systems will bring in new ways of doing older tasks. There may be some resistance to this new way of accomplishing tasks. Communicating to the workforce on how the AI system will make their jobs easier and faster along with training them to use it properly will reduce any friction in adopting the new AI system. Externally, the customers will appreciate the agility and automation you have built using AI in your products, services, and digital experiences.??

Technology & tools

For building and maintaining an AI-based system, you will need to decide on tools, software, and any necessary technology. Traditionally, AI systems use machine learning and deep learning, under the covers. For building, training, and deploying the underlying machine learning and deep learning models in an AI-based system, you will have to decide on the tools and services required for those tasks. All this can add to a huge upfront capital cost to procure all this required software and hardware. An alternative to setting up your own infrastructure and getting a disparate set of tools and software from different vendors is to use cloud providers for all your AI needs. Major cloud providers offer machine learning and AI tools, services, and compute infrastructure to build, deploy, and maintain these AI systems.?Azure Machine Learning, Azure Cognitive Services, Azure Bot Services, and Azure Cognitive Search Services are examples of such tools and services from Microsoft. Going with one cloud vendor for all your ML/deep learning/AI needs will reduce the administrative effort of managing relationships with multiple vendors.????

Privacy and security

Security and privacy of data used by an AI system are critical parts of any AI system and essential elements to be considered in building, deploying, and using an AI system. Depending on where globally the AI system is being built and will be utilized, consider the local privacy and data protection laws such as General Data Protection Regulation (GDPR) and other regulations. Other strict US regulations such as California’s Consumer Privacy Act (CCPA) call for anonymizing the data and removing other personal information. While the data is still useful and can be used in training a machine learning model, it may not be that helpful in training the AI system because of missing personal information data.

Most AI systems use machine learning and deep learning models which require data to be continuously fed into these systems to help the models to become more and more intelligent in accurately responding to queries or making predictions. Some of the data is private in nature, such as medical diagnosis data, for an AI-based healthcare system. It is important that you have proper consent from the users, from whom this data has been obtained. The personal nature of this data also makes it more susceptible to theft and breaches. Once proper usage consent is obtained, make sure that this data is properly encrypted while at rest or in transit as it gets utilized for training and building the machine learning models. When the AI system is in production and is being utilized with additional input from users and producing output, all that data is still required to be secured via encryption and other cybersecurity techniques.?

AI tools and services

Many companies are offering AI services and tools. There are very few companies out there that might offer end-to-end tools and services to fulfill all your requirements. Cloud providers are examples of companies that can offer end-to-end AI tools and services. It is worth considering them, so you have one vendor for all your tooling and AI services rather than going with multiple vendors. Vendor, tools, and AI services management will become an issue if you go with multiple AI vendors.

Cloud providers like Microsoft, Google, IBM, and Amazon are trying their best to commoditize AI and give developers the tools, services, and ability to build and deploy AI-based systems quickly and easily. Each cloud provider has its machine learning offerings, AI services, and tools. ?????

0