The Sticky Tentacles of AI

"AI likes to get it's sticky tentacles into all your systems"

Talking with Rosie as part of the Into the Motaverse podcast, I made a passing comment on the use of AI that really stuck with the both of us. Like many in QA, I've been skeptical of AI in the QA space specifically. I've struggled to find use cases for it, and question the validity of it's output. I believe this is fair and valid - my whole career is based on scepticism: of requirements, of code, of the system itself, etc. That's how I test. 

On reflection of my comment, I do have one potentially great use case for AI: internal systems integration for information gathering. Part of the major gaps in system quality is ensuring alignment. That the system is build according to specifications, that the specifications are updated regularly, that key  decisions are not lost in Slack threads or huddles, customer feedback and support tickets make their way into the backlog, and that the test cases and product documentation accurately reflect the intended state of the system. This is large-scale systems integration: Slack, Github, Jira, Google Docs, Hubspot, AWS logs (or whatever application stack you use) - all different products built by different companies that do not integrate with each other at all. Thus, I spend far too much time chasing this data and attempting to collate it into something meaningful. A worthwhile endeavour, but tedious and time consuming to say the least. 

The issue is that these systems CAN integrate with each other. They simply DONT. They all have external facing APIs that technically allow them to share data. I say this with love as a former API developer, these APIs are merely an afterthought for many companies. They are cumbersome and poorly designed and documented. Any human attempt at creating an integration middleware is often met with frustration and failure. 

AI on the other hand suffers no such fatigue. It is designed with sticky tentacles that eagerly seek to consume all your data. It's a foundational part of how AI functions. In my theory, give it the correct API Keys and it will greedily sync all your data sources and ensure nothing is lost into the ephemeral void of a Slack huddle. The AI can provide you with what was said or messaged by whom at what date and what pull requests are related to that discussion. This creates a timeline of decisions and actions. Meaningful information. 

Eventually it may stand that the AI itself will become a single source of truth. The AI wont need to update Confluence pages for people to review, but rather you'll just ask the AI to give you the information directly, on demand. 

We still have to be skeptical. AI will always hallucinate, fabricate data, or make inferences that are not true. It would be prudent to ensure the AI is not just giving you a summary but can cite its sources. But the scope of data being made available in one place from many sources is an incredible innovation that in my opinion should be explored.