How can an organization develop its Artificial Intelligence or AI strategy? What are the various facets of developing an organization’s AI strategy? What is the CIO’s role in developing that strategy?
In an earlier episode, I had covered the basics of AI, its related technologies and the impact that AI is having on organizations in general and touched upon certain business use cases. So, if you are looking to get a quick review on the basics of AI, then please look for the episode published before this one.
In this episode, we will take the discussion forward and go deeper and start discussing the overall strategy or strategies that CIOs and CTOs need to pursue to start building an effective AI-enabled organization. We will also discuss topics that should be of importance for any CIO or technology executive when defining an organization’s AI strategy.
So, let’s get started.
It’s no secret that CIOs and CTOs across all industries are under pressure to accelerate the adoption of AI within their organizations to give them the needed technological and operational advantages and boost over their competition. Strong AI-capabilities, similar to digital transformation efforts are becoming quite essential for survival. Not having an AI strategy and corresponding capabilities can be very risky. I mean not building the needed AI capabilities will leave organizations stuck with old systems and processes, and as those legacy systems and processes lose their relevance fast in this digital economy, it could put the organization at a great peril. On the other hand, if applied properly, we know having the right AI capabilities can provide organizations with the needed competitive advantage due to AI’s ability to deliver deep business insights from its data, automation of operations, ability to forecast business demand, and creating better products and other capabilities that can help position organizations to start competing effectively in the new AI and digitally enabled marketplace.
As AI is fairly new, CIOs and CTOs should also ensure that there is collaboration and cohesion amongst various departments of the organization wherever there may be any AI related work and experimentation going on. This collaboration will ensure that organizations as a whole can learn from each other faster in turn accelerating AI’s adoption within the enterprise. Having shadow activities throughout the organization can prevent executives from formulating a cohesive strategy, which can deliver more impactful results than standalone or siloed initiatives.
CIO’s role in the adoption of ML and AI capabilities
Before we get into the various dimensions of developing an AI strategy, let’s ensure that we understand a CIO’s role and responsibility in the development of such a strategy. As CIO’s are increasingly become part of the organization’s top executive suite, we should remember that CIOs are also responsible to reinvent their organization’s business models as being demanded by the external market forces. CIOs can do so more effectively by using AI and its technologies to extract relevant insights and knowledge from its operational data that can help in business executive decision making.
CIOs are also responsible to work with the business to identify various business problems and use cases where ML can benefit the enterprise and to deploy relevant ML solutions across the enterprise. In this context, CIOs are also responsible to select the right tools and technologies, architectures, and patterns along with identifying the right AI models and institute a process where deeper data analysis can yield insights from data. In a nutshell, CIOs need all this to help them accelerate the creation of customer value and increase business growth.
Keeping this in mind, we will review 7 key dimensions that contribute to the creation of an organization-wide AI strategy. They are the following:
- Getting clear on an organization’s overall strategy and objectives
- Investment in AI research including the need for experimentation and democratization of AI in the organization
- Identifying the business use cases for AI within organizations
- Considerations in building the AI-enabled technology platform
- AI-enabling various business applications
- Developing an AI focused Data Strategy
- Building the right set of skills in the organization
In the rest of this episode, we will now cover each of these in greater detail.
Get clear on overall strategy and objectives
The first dimension we will look at has to do with getting clear on the organization’s overall strategy and objectives. As always, when adopting any type of strategy, executives such as CIOs and CTOs should be focused on the overall business strategy of their organizations, and stay focused on specific business outcomes as they consider developing their AI strategy. As we will note later in the episode, implementing an AI technology infrastructure and processes can be an expensive proposition and before embarking upon any such initiatives it’s important that business executives get an idea on how AI is going to benefit their organization and their commitment on pursuing such initiatives. As we had noted in an earlier episode, implementation of AI within an organization can happen pretty much at all levels of the organization – from automation of simple processes to bringing in robotics and implementing real-time Deep Learning algorithms and processing against large volumes of disparate data collected from various sources. As a business, you can’t afford to just buy a set of tools and technologies and then hope to get the right ROI from that investment. It’s important to spend time and understand the business outcomes that you desire, discuss with your stakeholders, understand your organization’s readiness in terms of the data and the underlying platform, and then you can start to formulate strategies and invest in those capabilities.
It’s therefore essential to ensure that an organization takes a methodical and systematic approach towards AI’s adoption while ensuring that their overall approach is aligned to the organization’s overall strategy. Perhaps at this stage, an organization can also develop some sort of a maturity model to get a roadmap that shows the adoption of various AI capabilities over time within the organization.
Taking a strategic approach doesn’t necessarily mean that enterprises should stop themselves for addressing some obvious problems that they could resolve readily or to ignore the low hanging fruit, but it means that keeping the perspective of the overall organization strategy is a must. AI should, therefore, be a core component of an organization’s overall digital transformation strategy.
Invest in AI research
The second dimension of developing an organization’s AI strategy is to invest in the right research as a precursor to building larger AI capabilities within the organization. In general, we know from industry surveys that a number of organizations have been investing billions in AI research for the past 3 years. This is also confirmed by a number of research organizations including a study by Mckinsey. Depending on the potential business cases that you as a technology executive have identified in the organization, you should look to start experimenting to gain basic capabilities in ML, DL, natural language processing, speech recognition, computer vision, etc.
When investing in research and experimentation, a CIO’s focus should be to democratize AI and its capabilities in the organization. When carried out properly, eventually this can result in a change of the organization’s overall mindset. As an example, analytics is widely used in all parts of the organization. An organization can take its analytics progressively from non-AI to AI-enabled analytics to give them more insights from that data. Most organizations are used to descriptive analytics, which essentially informs end users on historical information and explains ‘what happened’. The next step up from that type of analytics is “diagnostic analytics”, which delves into the root causes of the events and provides information on ‘why did it happen’. The third step up from that touches on predicting future outcomes from existing information and data and the next level after that is for the system to recommend actions – something which is also referred to as “prescriptive analytics’. Finally, we enter the domain of advanced AI, where systems become self-healing and find solutions to various problems. Although a number of such self-healing solutions have been built-in in various computing and telecom hardware, their use is also showing up in other user applications as well. So, we see that by democratizing the principles of AI and the insights they can provide the business can enable an organization’s users to start thinking at a totally different level and augment an organization’s overall intelligence.
As part of the research, an organization should also start experimentation with building of a digital platform that can have AI applications covering scenarios across various devices, the cloud and the edge. Starting the experimentation process by embarking on various Proof of Concepts and pilots related to the various use cases based on a coherent strategy can help accelerate AI’s adoption within the enterprise.
Leading the research and experimentation process can thus trigger the needed conversations at all levels of the organization on the possibilities for the organization and help an organization rollout AI and augmented intelligence across all the organization’s business processes.
Identify the Business Need for AI
Another key input to developing an organization’s strategy is for the CIO along with other key business stakeholders to start the conversation and identify the specific business use cases where an organization can benefit from AI. This is an important point that we should reiterate and that is that a CIO should not get too bogged down in the tools and algorithms but rather in the specific business use cases that your organization can benefit from. In that context, it doesn’t hurt to discuss the use cases in the context of the available tools and technologies but the focus should be on specific business use cases and the overall business case for piloting AI solutions for those use cases. CIOs should also start the conversation by broadly including staff from IT, LOBs, industry players, competition, and other business stakeholders. As a CIO, it’s useful that you start by asking the big questions, identify the real business problems, and discuss opportunities where AI can bring the most impact to the business. As you have these conversations, you will begin to see your strategy start taking shape and as that happens ensure that the strategy stays aligned to the overall goals and objectives of the organization. As you get clarity on that, eventually you can start a deeper discussion related to the relevant AI tools and frameworks.
More commonly, CIOs start by implementing AI in their departments and usually the data center. AI is being used in the data center for the past few years and can be used in a number of cases such as forecasting hardware and computing requirements, optimizing space and energy requirements, and in dealing with infrastructure management issues such as detecting or even preventing hardware failure. We also see AI being used to the extent that a majority of data center tasks are not only being completely automated but AI solutions provide recommendations on further optimizing various aspects of a data center’s operations. AI is also extensively being used in the realm of cybersecurity to fend off online attacks. AI technologies allow not only for detection of attacks but based on historical incidents and other data across the net, can provide various prevention strategies.
Before we leave the topic of use cases, you should keep in mind that the AI platform or infrastructure that you end up building and its corresponding costs will depend largely on your business uses cases. For example, if you are implementing an AI system to provide you real-time insights related to the health of a manufacturing operations or let’s say a trading operation, you will need to invest appropriately in the right kind of storage and processor hardware solutions to make up for the latency and scalability issues associated with regular network attached storage systems. Similarly, if you are planning to implement Deep Learning based solutions, they sometimes require more processing power than the traditional CPUs. In that case, you will need GPUs or better processing capability to handle those AI workloads and services.
Building the AI-enabled technology platform
Once an organization identifies opportunities where AI can potentially help the organization, CIOs should start focusing on building the right technical infrastructure that can deliver those capabilities. Again, depending on the specific uses cases, this can have wide-ranging implications. Decisions needed in the building of the right technological infrastructure include the choice of tools, service providers, ML architectures that may be needed for the various business areas of the organization, and other such decisions.
Similar to putting together any other digital or technology solution and infrastructure, implementing an enterprise wide AI system and infrastructure requires a well thought-out and cohesive strategy. After all, implementing such a system that touches enterprise wide data elements, requires massive processing power and large and continually growing storage, numerous interfaces and pipelines that bring together data from various sources, high bandwidth networks, algorithms that are far sophisticated in logic than the other workloads that an enterprise may be running, etc. needs strategic thinking and planning and a clear implementation roadmap.
Accordingly, as you build sophisticated Al systems and applications, you will need to ensure that the surrounding systems management and processes are upgraded as well to handle the upgraded capabilities of your systems and applications.
As for building the AI applications and services, one obvious choice is to start with AI service providers, which offer various ready-made AI services. These include services from the large service providers such as Microsoft, Amazon, Google, IBM, and others.
For example, if we look at Microsoft, Microsoft’s Azure ML enables the building of ML and DL models. They also offer Pre-Built AI by providing access to finished services on the cloud, which organizations can start consuming with little effort. If your organization wants to build a chatbot, for example, they have a framework for building bots and provide tools for the development of conversational AI. In fact, Azure provides a number of solution templates, reference architectures and design patterns on their AI gallery.
Amazon’s AWS, IBM’s Watson, Google and other service providers too, provide a number of pre-built AI services as well as tools to build AI applications. Depending on your business situation, you can explore those service providers and their tools and see which one would fit your situation more than the others. Perhaps in subsequent episodes, we will have a more in-depth conversation about the various types of tools and some of their pros and cons. If you are interested, ensure that you subscribe to this podcast so you are notified of our future episodes.
AI-enable existing applications
Next, let’s take a look at another dimension related to developing an organization’s AI strategy and that has to do with how to AI-enable various applications within the enterprise. In this context, organizations can start using the capabilities embedded in their internal enterprise systems along with developing their own AI applications for their sophisticated business use cases. I discussed this topic briefly in an online meet-up when answering a specific question. Here is that clip.
For example, a number of enterprise systems have already started to incorporate AI capabilities augmenting the capabilities of those tools and essentially providing the users with augmented intelligence. We see AI functionality being built in systems such as SalesForce, MS Dynamics, Workday, and others. When used in these systems, AI turbo charges capabilities of those tools, providing organizations better insight into their customers, suppliers, operations, and other business processes. So, one strategy might be to use AI in that context. More specifically, let’s consider SalesForce, an established CRM tool in the market. This tool has added a layer of artificial intelligence through a product called Einstein, which brings various AI functionality to its users. Einstein in SalesForce allows users to perform sophisticated data analysis and allows users to predict business outcomes, allows an organization to build chatbots that can be trained with the organization’s CRM data making chats with customers more personable and other such capabilities. Microsoft’s SharePoint is another example in which Microsoft has packed a number of AI features, especially related to image recognition and text extraction. Users can essentially teach SharePoint a number of things such as recognizing a certain image and when the tool scans documents and detects that image, it can treat them in a certain way using the rules defined in the system. Such capabilities can help in managing contracts, invoices, and more.
So, we see that many tools have already incorporated AI under the hood and while it may not be very obvious to end users, a lot of reports and analytics that is presented to users already make use of machine learning algorithms to report on more sophisticated analytics.
Organizations can also start developing AI applications of their own by incorporating specific AI technologies that are more relevant for their business use cases and needs. For example, technologies are available from companies like Second Spectrum that use ML to watch every second of video of a basketball game, understand it, and then deliver insights about the game and potential improvements to teams and coaches. Earlier this was accomplished by interns spending hours and hours to watch games, tagging videos, and making them available for coaches to watch. Now all that can be squeezed into minutes and seconds. Organizations who see any relevance in such technology, can deploy appropriate ML models to get insights from all types of data including video, images, text, and others and use those insights to improve their business processes.
Organizations can also enable their other apps and cloud workloads with AI and ML technologies to enhance their customers’ experience. AWS, for example, provides what they call “pre-trained” AI services through AI APIs, which can provide a basic set of ready-made intelligence for use cases, which include but are not limited to personalized recommendations, increasing customer engagement, getting security recommendations, image and video analysis, advanced text analysis, and more.
Developing an AI-focused Data Strategy
The sixth dimension in developing an organization’s AI strategy has to do with data. Organizations have now realized the massive insights and intelligence that is trapped in their processes and data. CIOs, therefore, must immediately adopt a comprehensive data strategy to help them get that data organized and then to start leveraging the right tools to provide them the insights and enact processes to act on that intelligence and derived insights. A CIO has the responsibility to establish a platform that takes the data across its lifecycle from its initial creation and allow for democratization and easy consumption and analysis of that data. Considering the complexity of the process and the vast stores of enterprise’s data that’s easier said than done and thus requires a solid data strategy. As we know, data is the fuel that powers AI and its applications. The problem that we have at the enterprise level related to making data available for AI is manifold. First, we need to ensure that we prepare and organize the data assets appropriately to make it AI-ready. Second, at an enterprise level we usually are dealing with large volumes and multiple stores of data. Digital transformation initiatives are already resulting in solutions that are inundating enterprises with large volumes of data. The nature of AI system that you decide upon may further increase the volume of data that your organization may handle and process. Third, we need to ensure that the various types of data are organized for organizations to run ML and AI to extract the needed insights. Finally, the digital transformation is making enterprises generate huge amounts of data constantly and an enterprise’s AI infrastructure needs to ensure that it caters for this non-stop data streaming as well. In lots of cases, the amount of data that an organization is collecting is faster than the organization can run insights on that data.
So, with these challenges at hand necessitates that an enterprise and its senior leadership develop a relevant data strategy. Having the right data architecture can make the development of AI applications faster and easier and can also pave the way for future AI applications as they get identified depending on the organization’s business needs. So, in the next few minutes, I will touch upon various points that a CIO should consider when developing an organization’s data strategy for AI.
- First, we should point out that the data strategy for AI is part of an enterprise’s overall data strategy. An enterprise’s data strategy is usually created with a number of requirements in consideration, among which AI should be one of them. Some organizations who don’t have a data strategy may find themselves scrambling for one as they venture to bring AI into the enterprise. While they may have gotten away with not having a data strategy until this point but it can be difficult to establish a coherent and well aligned AI strategy without having an underlying data strategy. So, a number of organizations find themselves scrambling to develop one, if they didn’t have one before.
- The second part of developing a data strategy is to identify potential data sources across the enterprise and even external to the enterprise. These data sources are those that an organization intends to join and marry as part of their AI efforts to derive the needed business insights. These data sources could be both structured and unstructured and could be from sources such as ERP systems, mainframes, legacy databases, IoT devices and sensors, and other sources. The analyzed datasets could also span a number of business areas such as product complaints, literature reviews, external research, product malfunctions, social media and other data. So, if you are preparing and organizing a data model for marketing, you could get data from various omnichannel sources, internal platforms, media partners, agencies, online transaction data, and data from other sources.
- Another aspect of the data strategy is integrating all that data into repositories that can be used for AI. This may involve integrating various data sources into a data lake or to integrate data in traditional data warehouses. Data aggregation involves integrating data lakes and other sources, building data connectors for data access, data from IoT and other sensors, etc. Consolidating their enterprise data to give them better insights into their business operations such as bottlenecks in their processes, errors in their manufacturing operations, and so on can be the first step toward AI. ML can then be brought in and used against that data to identify patterns in that data and to provide enhanced forecasts and predictions.
- As part of developing the right data strategy, it’s also important for an organization to establish the right data flows. An organization needs to ensure that the organization has the right and repeatable data flows that pull in data from an organization’s business processes, covering all devices, sensors, etc. These data flows may need to be established to bring data into the enterprise’s analytics or overall AI platform.
- Another point related to preparing and organizing data for AI is that due to the nonlinearity of data and the various types of structured and unstructured data that must be integrated, developing a data strategy for AI may be more complex than traditional approaches. Also, data preparation in AI usually involves statistical and mathematical processing of aggregated data including normalization of data, and data cleansing. CIOs therefore should understand that appropriate skills related to data science and AI may be needed to develop the right data strategy for the organization.
- Data quality is also an important factor that should be considered when developing an organization’s data strategy. Low quality data into your AI system will result in corresponding poor quality insights, no matter how sophisticated your AI algorithms. Techniques such as data scrubbing and data cleansing therefore become quite essential in this exercise. Various quality checks ensure that data is fit for ML training and testing and also allow for addressing security, privacy, and governance issues.
Build the right set of skills in the organization
Another dimension of building an organization-wide AI strategy is to ensure that the organization has the right types of skills to help it build its AI capabilities. The skills can vary quite a bit depending on your organization’s needs and mostly would depend on the types of capabilities that you as a CIO decide to get within the organization. CIOs and technology executives also should not overlook the need for formal training for its key staff. This training can address the basics of AI. And once the organization has a basic strategy in place, more advanced training can be provided in line with that strategy. So, for example, if the organization plans to use AI services from an external service provider such as Amazon, then it can focus its training on that front.
The various type of skills that an organization may need include the following:
- Skills related to Image Processing, Image Analysis, and/or Computer Vision
- People with skills and training in machine learning data mining and other quantitative research analytics such as: Non-Linear Regression Analysis, Multivariate Analysis, Bayesian Methods, Generalized Linear Models, Decision Trees and Random Forest, Non Parametric estimations, Neural Networks, Ensemble Models, etc.
- Strong background in statistical languages technologies (e.g. R, SAS) and deep experience in Hadoop technology and open source languages (Python, Spark)
- Experience with computer development languages (Java, C++).
- As most of these solutions are built on the cloud, skills are also needed related to delivering solutions pertaining to a mix of Cloud, Hybrid cloud,
- Skills are also needed in the area of data techniques such as Data Warehouse, Data Engineering, Advanced Analytics, and Data Science
- And there are other such skills needed.
With this we come to the end of this podcast. I will have more topics on AI and how to overcome various challenges in bringing AI in organizations in a later episode. Again, please subscribe to this podcast to ensure you are notified of those episodes. This is Wasim Rajput and thanks for listening.
