Navigating AI Implementation: The Case for an Enterprise AI Architect

Unlocking the Future of Business: How an Enterprise AI Architect Can Guide Companies Through the AI Revolution

Even to the casual observer, Artificial Intelligence (AI) is clearly one of the most discussed digitalization topics of 2023. No wonder, since AI has the hallmarks of a real game changer, potentially revolutionizing the way the world economy works – and more. Large and fast-growing investments by companies around the globe reflect that.


AI implementations impact most business areas and require coordinated delivery across multiple units and all technological layers. A plea for establishing a new Enterprise AI Architect role.

The ongoing global digital transformation is prone to creating technology hypes regularly, some short-lived, some with a lasting impact. But in all probability, nothing in this century has been more of a game changer than Artificial Intelligence. It will potentially completely alter the way we do business with each other and may even fundamentally change our lifestyle.


Many organizations have realized the importance of AI. BCG surveyed 1,400 C-level executives across 50 markets about artificial intelligence and GenAI. Remarkably, 89% of them placed these technologies among their top three priorities for 2024. In 2022 alone, the worldwide corporate investments in research, development, and applications amounted to 92 billion US Dollars. With Generative AI having added fuel to the fire at the turn of 2022/2023, this number was undoubtedly higher in 2023.



Hasty implementations will backfire

As with most new technologies, there is no known gold standard for implementation and utilization of AI. Different business units within the same organization can tackle similar topics using various solutions from different vendors. This creates a considerable space for inefficiencies, especially if AI is implemented without looking at a broader picture of the entire business & IT ecosystem.


Having said that, poor data quality, wrong integration mechanisms, and suboptimal exposure to the business will all have a detrimental impact on results and performance of an AI application, and may even render it useless. This will be especially the case within a decentralized IT-landscape, for example, in enterprises organized in independent communities or domains.


Deploying AI models at scale and making them useful for the normal operational field is proving to be difficult at times, but it is a necessary step to ensure value generation.



First, sort out the basics

So, what is needed to get a tangible ROI from money spent on AI solutions? Any company contemplating an investment like that should be able to answer some basic questions. The following are some key examples:





  • Value assurance: Where in the organization is AI most useful? For which business processes or business capabilities?


  • Model governance: Which solutions should be used to address specific business needs? GenAI, Machine Learning, or just conventional algorithms, e.g., decision trees? How can we ensure that AI is used responsibly and that models do not drift over time?


  • Platform foundations: Are any solutions already implemented in the organization that can be utilized to accelerate the whole process and save money? Should the company leverage vendors’ proprietary cloud platforms or develop their own inhouse?





  • Data governance: Is the relevant data for model training available?

    Is it in the required format and of good quality? How can we ensure data privacy and avoid data leakage?


  • Delivery approach: What is the best way to integrate AI solutions into the business processes within the organization?

    Building new SPA applications or, e.g., embedding AI in Business Process Management or existing domain solutions? How can they be tested?

It is apparent that the aforementioned questions cover a vast range of subject areas, from AI itself and data to cloud solutions and integration. To get meaningful answers, a holistic view of the organization, its business and IT environment is needed. And it must be complemented by comprehensive knowledge of AI, e.g., its capabilities, limitations, or the pros and cons of different approaches and technologies.



The new Enterprise AI Architect Role

In our opinion, the sheer number of interdependencies call for a new role to coordinate AI-related efforts at an enterprise level, the Enterprise AI Architect. This person’s job is to keep track of, and continuously manage, requirements from the business side and match them with technical capabilities and relevant governance policies.


The objective is to implement an effective AI application in a cost-efficient manner by leveraging existing solutions while also adhering to architectural principles. Alternatively, if the necessity arises, the Enterprise AI Architect could decide to discontinue an AI project if it hits insurmountable economic, technical, or legal obstacles.


The role of an Enterprise AI Architect therefore requires someone with a multitude of different hard and soft skills. The person filling that role needs to have a solid grasp of AI concepts, enterprise architecture, and data science, as well as a comprehensive understanding of the broad business ecosystem and the legal ramifications, such as:

Knowledge of the AI solutions market; key vendors and their products,

Proven integration approaches, relevant for specific AI solutions,

Management of AI models over time,

Delivery and licensing models, their pros and cons,

Data management from requirements to acquisition, transformation, and creation of adequate data sets,

Deployment at scale options – on-site vs. cloud,

Continuous observability of AI model effectiveness to avoid any drift in quality over time,

Security and responsible AI by design.




Imperative: Broad business and technical background

Knowing the developments in the evolving AI world is, as already stated, just one side of the coin. The Enterprise AI Architect also needs to know the company’s strategic mid- to long-term goals, business processes, and IT environment. She/He is a key stakeholder in helping define the end-to-end transformation process and ensure the target architecture will be defined and met.


This should, of course, include awareness of the low-hanging fruits, i.e., where an AI-based improvement is needed most. This makes it possible to determine if a suggested AI solution is useful for the business or is an experiment in disguise.


This is just the beginning of the required skills, as the Enterprise AI Architect is expected to be responsible for advising on data preparation, integration, and rollout. This includes addressing conceivable compatibility and data quality issues due to outdated systems or datasets. To top all this off, there are internal as well as external governance principles and guidelines to consider. Some of these are mandatory, and alignment of the AI solution is a must.


Finally, the Enterprise AI architect must enhance current architecture principles, patterns, guidelines, and technical standards to meet the new reality, while preserving freedom for innovation and value creation.

Desirable: Good mediation skills

Dependable soft skills are also needed, as the person will have to inspire the business owners, but also needs to be able to talk to IT about the proverbial ones and zeros. This is already indicative of the necessity for a very open-minded character, who is willing to constantly learn and experiment. After all, AI is mostly terra incognita, at least for now.


The Enterprise AI architect must also ensure proper collaboration with AI-related internal stakeholders, like data engineers, data scientists, and data governance representatives, but also external ones, like third-party AI solution vendors and external consultancy experts, if required.





Enterprise AI Architect – a future must have

All things considered, the Enterprise AI Architect is a very challenging role that requires management of a quickly changing solution from all perspectives – code, data, and model wise.


Where it is located within the organization – with the enterprise architecture function, in an independent AI realm, or at an entirely different unit – depends on the enterprise’s individual characteristics.


There is only one certainty: The Enterprise AI Architect is a must.






Corporations that are first in establishing such a role the right way will gain a significant head start in the AI race. They will be better positioned to shape their AI strategies and invest their portion of the 92 billion dollars spent on AI in a more optimized and reasonable manner.


About the Authors

William El Kaim

IT Architecture Director
Paris, France

William El Kaim is a BCG Platinion IT Architecture Director and has more than twenty five years of experience in planning, designing and governing complex Information Systems. Before joining BCG Platinion, he specialized in agile IT transformation, digital innovation, new technical architecture (Big data, Cloud, IOT, Microservice) and data governance. Over the years, he has worked with international companies in Finance, HealthCare, Energy, Travel and Transport. William also created a startup, contributed to UML V2 and Model Driven Architecture at the OMG and published the Enterprise Architecture Digital Codex web site. William holds a doctoral thesis in Computer Science Engineering from the Pierre and Marie Curie University (Paris, France). In his spare time, he enjoys creating digital games with his kids using Unreal engine and stencil, blogging about technology and watching anime and SF movies.

Przemysław Krajewski

Lead IT Architect
Warsaw, Poland

Przemysław is an experienced IT Architect working in the BCG Platinion Warsaw office. He has over 10 years of experience in conducting consulting projects related to the analysis and optimization of business processes, planning changes in the area of corporate architecture, IT and IoT architecture.