Holistics vs Looker: AI-Powered BI Comparison

Generative AI has gotten pretty good at descriptive analytics. i recently tried Gemini in Looker and as long as the tables had descriptive column names, it did a great job answering business questions that my stakeholders usually will go to a dashboard for.

I've been in analytics for 16 years, I have used most top models for interacting with data on the analytics maturity spectrum by now. Any job where the core KRA is building dashboards/ reports is already at risk as long as there's an appetite in your company to use AI.

It's garbage in garbage out process. If the underlying table has false/no data description or column names or misleading column names, it's set to come up with incorrect insights. No surprises there and it's more of a data problem than a Gen AI problem.

Reg: first layer, that's what I meant with the descriptive analytics. I've had a bit of success having AI explain what contributed to a spike or drop in a kpi but then I made sure proactively that it had access to data needed to derive that.

It's not able to do exploratory analysis...yet but we've barely crossed 2 years and the capabilities multiplied rapidly. It's just a matter of time it does a decent job on exploratory analysis and once that is accomplished, predictive analytics wouldn't be a far-fetched dream.

Having said this, it's a lot dependent on human governance (prompting, supplying credible data etc.) but those wouldn't be limited to just analysts. Anyone would be able to use natural language to interact with data.

Yes, mostly I use it for refining my SQL queries or to see what are the different ways in which I can solve the same problem. It's quite helpful in that sense.

I also use it to review queries of my juinor team members. To see if I may have missed out on anything. It helps with that extra set of eyes at times.

And to create documentation for dashboards/reports.

I also use it to ask questions which I can use in stakeholder meetings for requirement gathering. For eg, I explain the context of the meeting and then ask the AI to roleplay with regards to what questions I should be asking in that specific meeting. Helps me prepare and also helps understand perspectives from different POVs.

My guess is that the bulk of the companies are just bull$#!7ing with the buzzword. To successfully introduce AI into your BI, you need clean orderly data.


Go ahead and tell me the last time everyone cheered when the data governance team came through the door.

People mostly rool their eyes and crawl into a ball.

So - C suite seems to think AI can “answer all their questions”.
So I respond with “what are the questions you are wondering? I can pull data, schedule reports, to your inbox, or build live dashboards with graphs in any color of the rainbow!” - which is usually met with blank stares.
That’s why I know AI isn’t worth the trouble. It’s a solution to a problem we don’t have.

Some of the tools are getting better, but I can't help but think they still aren't really solving a problem.

If you're technically minded with some experience in data, then none of them are doing anything better/faster than what you can do with SQL or a BI tool.

If you're on the business side, they still aren't good enough because as the other poster said, you are reliant on a semantic layer so it's not that much better/faster than asking someone for a new dashboard.

The only way these tools can be halfway effective is if they sit on top of a well manicured semantic layer. I also think that the real winner will be the platform that figures out how to invoke an action from the insight. I.e. the analysis picks up on repeat customers and be able to recommend an action to take for those customers and then kick off the process with a simple push of a button …or if the action is low risk enough to do it automatically.

I seriously think that at the end of the day maybe 1% or less of companies at any given point are at a data maturity stage where they could truly leverage cookie-cutter BI-AI solutions. The rest of us are still cleaning up messy data and figuring out what the proper business logic is.