Analytici obchodní inteligence

AI will NEVER be able to understand the difference between correlation and causation. Bayesian inference, causal inference, experimental modeling are all areas where AI will always be inferior to humans

I've been working as a Business Intelligence Analyst for 12 years now. From what I've seen over the years I don't expect full time BI roles to be around much longer. I see the space evolving so that BI will gradually be absorbed by other roles. Initially as part of data engineering or data analyst/scientist (this has been happening for the last year or two), but then over time when AI becomes better and data engineering have set up the required data infrastructure, it'll become a part of less technical roles like marketing managers.

It's free, Less margin for error, and is already been taking peoples jobs for years now. This isn't "new", AI has been taking these jobs for a while now.

There's so much diversity in this role that it depends massively. In some roles you have a robust data model to work with and clear instructions from stakeholders on what do with it, so you spend your time building and maintaining dashboards. Those roles would be very easily automated, but they're quite rare in my experience.

The bulk of my time as a BI analyst is spent on 2 things. The first is working with domain stakeholders without technical knowledge (marketers in my case). This can often be like herding cats. Lots of people will have lots of different opinions on what KPIs they want to track or how to track them. And often they don't really understand the data limitations of what we can and can't report on, so I need to be there to provide guidance. Much of my job is spent guiding these people along, often massaging their egos along the way, so that the wider group of people arrives at a consensus.

The second is ETL. Even with whole teams of data engineers and operations managers, data is very rarely centralised into a single and easy to understand model. I work with about half a dozen different types of data sources (from AWS to Google Sheets). Each of these have hundreds of different indexes and many of those indexes have hundreds of fields. A tiny fraction of these fields have any kind of documentation and so all you have to go on is the metadata and the name of the db managers who put it together. Actually tracking down the data you need requires getting really into the weeds and following up with multiple people to try to track down who actually knows where to find the data you're looking for. That's just nowhere near enough data for an AI to get a hold of the data it needs.

This is uncertain because a mindset, an observation that a person can have, may be impossible for AI to take over.

It is a field where people usually believe in other people rather than AI.

Employers may want to replace current data workers but this may never come to be if the current offerings are anything to go by, AI tools in the analysis field struggle to produce satisfactory results...don't believe me? go ahead and try out CoPilot with even something as simple as Excel.

The ETL process is also a complicated, one which most AI is not nor ever may be able to handle, data needs to be cleaned and standardized before AI can take a crack at it, the "AI" and yes I have to put that in quotes does not understand the context of anything, it is a prediction model using gradient boosters that performs quite well under controlled circumstances, thrown into any critical thought role it starts to lose pace. Furthermore nobody who works in the AI space authoring models ands understands the inner workings of "AI" treats this as anything more than a highly sophisticated toy...maybe in another 10 years we can come back to this question and see if we should start to worry.

It's 2023 now and chatGPT4 can already do a lot of the components of a BA. Won't be long now..

Once a data model is available, a lot of the tasks can be automated. However, I don't see how computers will be able to bypass data quality issues and still give complete and correct data.

Because business analysis may not be very reliant on human operation, it may be analyzed by machine using a sequence and pattern of human behaviour

Still requires human elements like critical thinking.

As a BI analyst, I automate my own job every chance I get. Data, business requirements, and technology are constantly changing.

I spend half of my time maintaining, tweaking, and fixing automation jobs - these include dashboards, data sets, and database tables.

It requires a technical person who is also an expert in their business domain to translate business requirements into data or reports that others need.

A business analyst is all about identifying the critical areas in which it needs improvement which a manual or human-dependent work system.