Raz Mitache

How to Get AI Right Using Enterprise Architecture

Artificial intelligence (AI) is probably the most important new technology today. It has clear use cases, and the value that it’s produced so far is indisputable – just think of the digital assistant on your phone, driverless cars, even Gmail uses it. But it’s no longer the sole remit of huge tech companies. With AI becoming more established, many organizations are starting to get access to and try their hand at running artificial intelligence initiatives. The business world is after all similar to an arms race, and having the latest ‘weapon’ to help you get ahead of competitors is an irresistible prospect. The forecast? A large wave of new AI deployments in the near future… and with it, a lot of heartache.

Are mainstream organizations ready for AI?

The reason is that while successful, well-known AI projects may be capturing headlines (along with CIOs’ dreams of digital transformation), the technology remains challenging. Technological know-how is scant, and so are resources. Therefore a short answer to the question above is no. A more nuanced one is that some organizations, namely those who have a pedigree of delivering advanced IT projects, as well as relatively deep coffers, are better positioned to try and leverage this technology to their advantage.

Of course, this is cautious talk, hence anathema to the brave, ambitious, or foolhardy. If we learned anything in the last few decades is that if there’s any chance, not matter how small, that something might give a company the upper hand over everybody else, they’ll try it. They’ll buy into the hype, key people will get excited at unrealistic prospects, and they’ll try it.

And then they’ll probably fail, wasting time, money and resources in the process (maybe some image capital, too). So, in light of this wave of unprepared, overly-enthusiastic organizations forming on the horizon, all of them getting ready to wrestle with artificial intelligence, we thought we’d share our thoughts on how to make the most of this technology. The key to success? We’re confident it’s enterprise architecture.

Powering AI with EA

Smooth AI deployments need Enterprise Architecture

The first thing organizations should understand about AI is that it’s a very specialized technology, requiring a lot of contextualization. Talking to Forbes, for instance, Andrew Ng, one of the world’s leading AI experts, said: “AI technology in isolation is not useful. It needs a lot of customization to figure out exactly how it fits into your business concept. Doing that requires a broad understanding of your company and a reasonably deep understanding of AI. Exploiting the value of AI today requires a team that understands the business context and has cross-functional knowledge of things like how to fit AI into your hospital, or how to use AI in your logistics network. Without cross-functional knowledge of how your business runs, it is difficult to customize AI appropriately to drive specific business results.”

That’s coming from someone who headed both Google and Baidu’s artificial intelligence programs. He’s talking about understanding the business concept, first and foremost; having clarity over the business context and larger organizational picture; he highlights the need for cross-functional teams and collaboration, he even mentions to aim for specific business results(!). Now, if a minute ago you had any doubts when we said enterprise architecture was key to a successful artificial intelligence deployment, those doubts had better be gone now.

How Enterprise Architecture supports AI initiatives

The fact is, you’re simply not going to manage a complex, expensive initiative in a most likely already intricate landscape of dependencies without being able to capture the organization’s ‘truth’, i.e. have a good understanding of the business logic, strategy, stakeholders, processes, as well as the technology and data landscapes. Well, great news – enterprise architecture is your friend across the board there.

What better means to evaluate and determine the right AI use case for your organization; carefully plan and allocate resources to support the project; or properly mitigate against undesired repercussions across the wider enterprise environment? When you leverage architecture, you bring in the structure and transparency that’s required for planning and executing complex change across such a vast and diversified area as a modern enterprise.

A practical example of EA supporting AI

Let’s consider a real-world example. Say this year a telecoms company finished last in an industry-wide customer satisfaction survey. Given the poor result and, perhaps more importantly, the notoriously low level of customer loyalty in this field, the company decides to improve its customer satisfaction level. Here’s one way of going about this seen through the lens of enterprise architecture.

Step 1 – identify the elements/capabilities that support customer satisfaction

The first step the company takes is to create a customer journey map and run an analysis in order to identify the elements/capabilities that are currently contributing towards a great customer experience. This also allows them to carefully consider whether there any distinct negative aspects they might’ve missed before. Already, employing a customer journey map places the customer at the center, which is the very point of the whole effort.

Once this is completed and the capabilities identified, the team in charge can then proceed to highlight the cross-architectural elements that make up these capabilities. So, for instance, let’s say three capabilities – Customer Forum, Cinema Chain Partnership, and Telephone Customer Support Service – are identified. According to internal surveys, the first two are performing at a high standard and people who engage with them are on average happy with the level of service.

The third one, however, it turns out is constantly being slammed by customers who complain about the long waiting times and staff rudeness. Improving this capability would elevate the customer experience significantly. Just like that the opportunity for introducing a virtual chatbot becomes apparent.

Step 2 – assess the impact of introducing an AI solution on your environment

So now the case can be made for introducing AI technology in the organization. But what about the present capability? Well, EA makes it possible to precisely pinpoint the applications that enable customer support staff to carry out their work, and analyze whether retiring them to make room for a digital assistant would be preferable or not. Leveraging the technology and application architectures, in this case, with effective impact and dependency analyses will ensure that at the end of the day decisions taken are based on facts.

Step 3 – plan the implementation of the new system ensuring adequate mitigation where necessary

Up until this point, the telecoms company has decided they want to improve the customer satisfaction level. By looking at the customer journey, they identified an area of poor performance that is responsible for the vast majority of their customers’ complaints – their telephone customer support service. This is an area where incidentally AI has had substantial success recently. An analysis is carried out to determine whether the risk, cost, and overall requirements of introducing such a solution would be outweighed by the benefits of having a cutting-edge customer service offering. The organization decides to move forward with this idea.

Now this is the part where actual timetables are drawn and changes to the process, technology, and data landscapes are mapped out. By establishing a baseline and then mapping out the future state together with all the necessary intermediary plateaus, the organization can get a solid handle on the difficult task of bringing one capability offline, while also launching its replacement. This constitutes a core EA activity and it’s what makes it such an effective practice for managing complexity and organizational change. The result of this stage should be a clear understanding (via diagrams, charts and graphs) of the upcoming changes, with all stakeholders that are involved able to identify their role in the process.

Step 4 – execute and manage the implementation

Finally, as the project advances, unneeded supporting applications and technology are retired in a secure and orderly manner. The AI solution starts to come online, with its cast of supporting capabilities preceding it, just as planned. Should any unexpected development to do with the technology or regulatory framework occur, EA would once again be the best bet for coming up with a solution and ensuring the delivery of the project.

Step 5 – analyze feedback and optimize

Once it goes live, the organization ought to monitor the customers’ response and adjust accordingly. This learning and feedback loop is vital to continuously optimize the investment. Nothing is ever perfect, so keeping a close eye on the solution to try and find new ways to make it run even smoother – that sort of incremental improvement compounds nicely over a longer period of time. Not only that, but by assessing its performance regularly and testing various scenarios the telecoms company might very well identify new uses for the solution, broadening its scope and increasing the return on their investment. For example, they might realize after a while that their internal knowledge base could itself be significantly improved by introducing an AI-powered assistant.

Conclusion

In this scenario, EA powered the process of improving customer satisfaction levels, end to end. The bottom line is that in any situation, whether a problem needs immediate addressing or not, enterprise/business architects undoubtedly have a crucial role to play in testing future scenarios that involve AI. Their cross-functional knowledge of enterprise value streams, processes, technologies and data is a powerful enabler of transformation. As we move forward towards an age of generalized adoption of artificial intelligence technologies, the transparency and accountability afforded by EA position it as an essential tool for identifying great use cases and successful deployments.

Hopefully this overview managed to give you an idea of how enterprise architecture can be leveraged effectively to support artificial intelligence programs and similar digital transformation initiatives. If you found it interesting then you might also want to take a look at our Business Intelligence 2.0 post from a little while back. To learn more about capability-based planning, simply download our free Capability-based Planning with ArchiMate white paper. Happy reading!


Subscribe to our Blog

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *

1 × two =