A Day in Life of an Enterprise Architect Augmented by GenAI

A new model of work for Enterprise and Solution Architects can already be foreseen and will become fully workable with support from GenAI type of tools within 2 years. It will require necessary adaptation from our community as it implies new additional skills that do not appear yet in SFIA 8.0 competency model or in TOGAF’s Architecture Skills Framework. When I am reviewing TOGAF’s current Architecture Skills Framework and a number of older TOGAF Series Guides focusing on World Class EA, I get the impression that a good architecture practice, according to them, can only be implemented in a very large enterprise with an army of architects performing in a variety of roles.

This is not a winning value proposition for most of our clients that would like to be able to set up an EA practice promptly, quickly reach the 3rd level of EA maturity to deliver business value-based services. These services would be quickly appreciated and respected by various stakeholders dealing with a wide swathe of unanswered questions, such as about how to make progress in digital transformation, achieve systems compliance with new regulations, tackle technical debt in our architectural landscape, or how to create architectural runways. These would consist of leveraging building blocks for changes/parametrization/refactoring to enable prompt change implementation. Overall, such questions need to be better answered for more responsive and nimble system/service design and solution delivery.

From our recent contacts with several global clients, we can hear that in a number of cases centralized enterprise architecture teams have been disbanded and more federated approaches are being adapted to move the architects closer to their business clients with an expectation that these reassigned architects could better oversee the implementation of solutions instead of just defining standards and good practices.

I am not suggesting that standards and good practices are less important than before, but we don’t need a very expensive architecture team in the head office just for that, a team far from the action and the agile teams at the tip of the spear. These agile teams will also be impacted by GenAI and tools like LLMs (large language models) and KG (knowledge graphs), so EA expertise must penetrate deeper into the whole service/product life cycle.

Architects or Designers must become fully fledged Prompt Engineers. Why prompt engineering? Even if you use GenAI agents to assist you in some conversational activities and these GenAI agents do the prompting for you, you should always have the possibility to refine a prompt because the use of different patterns can lead you to a better outcome. From time-to-time, LLM hallucinations can even be welcome in the sense of supporting design thinking practices aimed at delivering more innovative thinking. LLMs are good storytellers and if we have good prompting skills we can better exploit the full potential of LLMs.

Several leading companies like IBM, Google, or AWS with its Bedrock AI Architecture, or even several start-ups that are using open source LLMs like LLAMA2, have already started to offer various specialized AI agents built on top of foundation models (LLMs). These foundation models can also be pre-trained by competent users and can support enterprise specific requirements and skills needed, i.e. for migrating from a Cobol-based legacy system to an alternative open-source platform based on Java, or for building an integrated information solution using currently existing enterprise data repositories or knowledge bases.

So how can such a day look like in the life of an architect being augmented in his/her practice with GenAI agents? Does enterprise architecture go away? Do fine-tuned metamodels go away? Does modeling go away?

I wager that in 2 years, architects and designers will be able to take EA or Solution Architecture from the Vision to the Delivery Phase. In this accelerated life cycle, they will be interacting iteratively with various stakeholders on the way to deliver POCs, then Pilots, and finally solutions ready for implementation after a business acceptance testing stage. Architects who acquire these types of skills to tap full GenAI potential will be very much sought after. However, the journey will be an intense one in which architects must be able to thrive in a very dynamic business/technology landscape. But they will be investing on the way in gaining and using the knowledge and skills they need anyway to bring continuous value-added support as modern EA SMEs!

 Authored by Alex Wyka, EA Principals Senior Consultant and Principal