Agentic AI for analytics. AI tools for data viz. AI, AI, AI.
The AI revolution is here, and it's dominating every discussion on every stage in business and even in our personal lives. I'm a frequent user of many AI tools myself, and I think it's pretty powerful. I have been able to build things faster, debug with my AI buddy, refine ideas. It's absolutely a game changer.
When I arrived at the Tableau Conference, I expected AI to completely steal the show too. It was going to dominate every conversation, every presentation. Partially, I was right. In the keynote & many of the smaller Theater conversations were around Tableau Next and overall agentic AI topics.
However, the truly meaningful conversations were centered around the strength that humans continue to bring to analytics. I have been a believer for some time now that we will need to possess strong AI skills in order to thrive in the changing analytics field, but that AI is not going to replace analysts.
Still, I witnessed several products and services that appeared to be trying to create tools that would "replace" parts of the analysis workflow, rather than augment or optimize with a goal to improve. It felt very focused on purely cost reduction and efficiency, not a quality increase which does often come with some efficiency boosts.
That didn't feel right to me. I spent the remainder of the first day, as well as the second and third days, going into smaller breakouts, hands-on trainings and roundtables where there could be more of a discussion between analysts and leaders so I could learn more about how industry feels about this.
During these conversations, it started to become clear that this isn't what people were looking for. Leaders were obviously trying to increase the ROI within their analytics teams, but that's not just an efficiency play. You have to have confidence in the data, and you have to enable humans to bring what they are uniquely good at to each analysis problem.
I think by now we are all aware that hallucinations are a concern with any LLM. If the AI doesn't know something, often it will just make something up. This presents a problem that (rightly) concerns leaders and analysts. How do I trust the results, and more importantly, can I trust them enough to make a decision based on those results?
As a side note, my company's product does help solve some of these issues by providing data quality rankings and providing those in the results, which you can find out about if you're interested here, but even with that, it doesn't replace analysts' human perspective and validation. It would be like if every leader just made decisions off of "vibes" rather than using data. That doesn't usually work out too well.
Much of my time both at the conference and the Serving Data + AI event was spent discussing confidence in data and the duality of bringing AI into the fold. While you can get to insights much faster, you absolutely must be able to trust those insights. Without that, the entire industry runs the risk of discarding AI tools because they cause your organization to make bad decisions based on bad insights.
If you've never attended Iron Viz, it really is a sight to behold. The energy in the room is huge, but the real excitement comes from the competitors and the amazing dashboards they create live on stage.
What actually gripped me from this contest wasn't actually the visuals, though they truly were incredible, it's no surprise these three were in the finals. It was the storytelling that won the day. I was particularly moved by Ryan Soares' focus. The dataset chosen for the contest was the FAA's wildlife strikes database, showing incidents in which birds or other wildlife were struck by aircraft and the resulting incidents. Ryan chose to highlight the data from the perspective of the bird, rather than from the perspective of the aircraft. To check out Ryan's entry, head over to Tableau Public.
While this wasn't the original intention of the dataset, which was to improve air travel safety, this perfectly highlights why a human's creativity is so valuable. Ryan was able to paint a different picture with the same set of data, one that could solve a problem for the birds too.
This is yet another reason why humans excel in analytics. AI is fantastic at recognizing patterns. While that's great for lots of use cases, there are many times in my analytics roles where I had to think outside of the box to find something that we had not thought of before. AI struggles with novel concepts since it's trained on patterns, on what we already know. It's not that it can't create something imaginative, but it doesn't innovate like humans.
The business world is bullish on AI, as they should be. There is a ton of value in getting to useful insights and outcomes much faster. You can beat your competition to the next customer problem. You can create something new before somebody else thinks of it. The possibilities are limitless.
I may be biased, but I believe the analytics industry will be one that continues to need humans. Should you use AI? Absolutely yes, if it's a good tool that has good validation and security, provides consistent and trustworthy results, and serves your business objectives. You would be foolish not to.
With that though, you must consider the role of the human in the analytics workflow. This is not an ethical consideration, though I think that's important too. You have to think about the context, creativity, and storytelling that you will be missing if you try to automate the entire workflow of an analyst. You will miss the nuances of your unique business. You'll underrepresent a customer segment. You would miss a blindspot that your team member finds because of something they heard a stakeholder bring up in a meeting anecdotally.
The human perspective, especially in analytics, is simply irreplaceable.
Like this article? Share it with a friend, on your LinkedIn feed, or discuss it in the GOATs analyst community here.
