AI and enterprise architecture, two years on

Was haggling about agents with a colleague and dug out a thing I wrote for a customer back in May 2024. Things are moving so fast, setting realistic timelines to anything seems weirder by the day. In any case, this foresight work from two years ago seems to be roughly on track…

Note: Potential impacts of AI on enterprise architecture choices
Three main issues appear relevant to (redacted-org-name) future operations:

1. Artificial intelligence and the use of large language models (LLMs) like ChatGPT will impact what is considered best practise in enterprise knowledge management. We have standards we operate to today around accuracy and findability of data. With LLMs in knowledge management, these accepted standards will be upset, if not now, then certainly inside of two years. This has an impact on how we scope any review of our content management systems and how we design for data accuracy in our services such as (redacted-product-name). Going forward, any investment in KM solutions that doesn’t factor in AI will probably be a poor use of funds.

2. AI in “code co-pilot” mode (whether Microsoft, Github Copilot or others) will extend the powers of workers like (developer-name-redacted). This is already happening today and needs to be factored into what we ask developers to do.

3. We are already seeing the use of AI in healthcare provision (for example https://beam.ai/solutions/healthcare). While not yet broadly established, the use of agents by citizens now seems almost inevitable (as it’s cheap and offers significant benefits). If this is true, it’s not unrealistic to assume that inside of 24 months, services will need to be designed not simply to interact with patients, but to interact with patients’ agents. This fundamentally changes the interaction design paradigm as the needs of the AI agent are very different from those of the human patient. Planning for this uncertainty means designing responsiveness and flexibility into our budget and roadmap. This is a challenge and a risk when framing the challenge in terms of complex architecture developments.

#architecture #agents