Building A Cognitive Enterprise Architecture

Josh Sutton, Global Head, Data & Artificial Intelligence, Publicis.Sapient
49
74
21

Josh Sutton, Global Head, Data & Artificial Intelligence, Publicis.Sapient

You can’t turn around today without somebody talking to you about the role of AI and cognitive in society, media, and specifically your business. It seems like everybody has an opinion about how you should be deploying AI solutions, from the most senior leadership of your company through to the intern that just joined. They are correct in their belief that these technologies could have a profound impact on your business, but transforming your enterprise architecture to leverage them is not nearly as simple as most people might like to believe.

I have had the opportunity to work with a number of Fortune 1000 companies and help them to develop their strategies related to cognitive capabilities. In this article I will share some of the best practices that I have seen work successfully as well as some of the lessons that I have learned along the way.

Start with the Full Set of Use Cases

One of the most common mistakes that I see firms make is that they attempt to start by either selecting a single platform (Watson, Amelia, etc) or a single use case (chatbot, etc). The rationale is often that they will learn from this exercise and be able to make better decisions going forward. This misses the key point that the combination of artificial intelligence and big data assets are transformative technologies that build off of themselves. The right first step is to understand all of the use cases that you believe might be impacted, or even created, as a result of leveraging these new capabilities. They don’t need to be completely accurate or comprehensive, but they need to cover the full range of solutions that you believe might be impacted by AI. These can then be segmented into categories of use cases that will build off of one another.

  ‚ÄčThe opportunities provided by cognitive technologies are compelling, however the execution of solutions isn’t as easy or simple as many would like 

The most common categories of use cases that I have seen tend to fall into the following categories:

• Insight generation: Machine learning, inclusive of deep learning tools, is enabling us to generate insights from our data assets faster and more reliably than ever before. The applications of this range from optimization of marketing spend through to analysis of call center trouble areas. Some of the latest technologies even enable this benefit to be derived from unstructured data such as images, call center conversations, and social media.

• Conversational engagement: Engaging with technology in the same way that we engage with one another, whether by voice or text, is becoming more and more of a reality every day. This ability to use technology to handle basic interaction is central to a number of customer and employee empowerment use cases, not to mention the allure of conversational commerce platforms.

• Knowledge work acceleration: I personally haven’t seen a great deal of use cases that are focused on full-scale automation, however I have seen a tremendous amount of opportunity that firms have identified related to automating individual tasks and components of a person’s job. The net result is a more effective and efficient workforce that is better enabled to deliver value on behalf of your company.

Map Your Use Cases to Services and Technology Platforms

Once you have a full set of use cases that you have summarized, it becomes a fairly straight forward exercise to extract out the set of core services that will be needed to enable the implementation of these transformative capabilities. Typically, I would expect to see high level services identified around a number of areas such as machine learning, natural language processing, vision, voice to text, data ingestion, reasoning and deductive engines, and the like. Some firms like a lower level of granularity and others a higher level–I personally don’t have a bias one way or another as long as you develop a clean and consistent framework to build enterprise services that can be leveraged across multiple use cases.

Many vendors will tell you that their AI platform can handle all of your needs. I have had the luxury of partnering with most of the largest firms in the market, as well as many of the smaller ones, and I can tell you that while there are many amazing technology platforms in this space, I have yet to meet the firm that can deliver on that promise. The very term, whether it is AI or Cognitive, is what creates a great deal of the problem. There are a multitude of different technical capabilities that are grouped underneath the umbrella of those terms. I have found it best to identify the top two firms per service that you are seeking to develop. Often the same platform will be in play for multiple services, which is a good thing, but keeping at least two in the mix ensures that you maintain negotiating leverage and design your services with an eye to being able to add and remove vendors as the landscape evolves.

Execute Against Uses Cases that Demonstrate Value Quickly

Now that you have a clear picture of all of the services that need to be built and the technology platforms that you plan to leverage, it’s time to go back to your use cases. Pick a small set of use cases that you can execute against that require different services, enabling you to build your architecture piece by piece while still creating real value along the way. I have found that three key criteria that I like to use as part of my prioritization process are 1) ability to materially improve either customer or employee experiences, 2) ability to validate technology components that you believe could be risky, and 3) those use cases which have compelling financial business cases in either the form of incremental revenue generation or cost reduction. By deploying use cases in this measured fashion, you are able to develop the enterprise architecture that you need while delivering real value in the timeframe of months instead of years.

Some Final Lessons Learned

There are a few final thoughts that I would like to share. I hope that you keep these in mind as you embark on your transformative journey:

• Experience design matters: In many cases, the experience design is actually more important than the technology implementation. People are excited by new technology, but also apt to reject it if their usability expectations are not met.

• The technology and data providers are plentiful (and will change): This industry is evolving faster than almost any other technology in history. It is safe to say that there will be new players before long that you will want to include in your architecture, and some firms that you will want to replace. Design your architecture (and your mindset) accordingly.

• Under promise and over deliver: People in your organization have wildly incorrect (and varied) beliefs about what is possible with cognitive technologies. Set their expectations at a level that is possible with the tools you have at your disposal today – don’t trust that they actually understand what is and isn’t possible.

The opportunities provided by cognitive technologies are compelling, however the execution of solutions isn’t as easy or simple as many would like. A well-executed enterprise strategy, however, may very well be what separates tomorrow’s market leaders from laggards.

Read Also

AI: How Do We Market to Artificial Intelligence?

Douglas Turk, CMO, JLT Group [LON:JLT]

Legal Knowledge Management and the Rise of Artificial Intelligence

Christopher Zegers, CIO, Lowenstein Sandler LLP.

Why Every Employee in the Enterprise Doesn't Need to be a Data Expert

Kumar Srivastava, VP, Product & Strategy, BNY Mellon