Tableau's CPO on the impact of AI, and how data analysts need to change

Jon Reed Profile picture for user jreed November 2, 2017
My last piece on Tableau news left out a big honking question: what's Tableau's stance on AI, and how should data analysts respond? Here's the three point advice for analysts from Tableau CPO Francois Ajenstat.

In my my one-on-one with Tableau CPO Francois Ajenstat, we hit on the big news stories from Tableau Conference in Las Vegas (Tableau CPO - bring on big data analytics - and the data competition). My analysis focused on Tableau's push for enterprise scale, so I honed in on the pesky issues of data certification and governance.

From a conference angle, I left out two big pieces: Tableau's position on AI, and their view on the turbulent cloud BI market. The "AI" piece is of particular import because many experts have forecasted a big downturn in the need for human analysts as automated analysis matures.

Tableau's take - AI will augment - not replace - data analysts

A commenter on The Artificial Intelligence Opportunity: A Camel to Cars Moment summed up the "I'm concerned" angle for data analysts:

It would seem the "business analytics" I learned back in Business School and along with my Ad degree is rapidly losing its utility. The "storytelling" skills are still relevant in my day-to-day, but I am absolutely seeing AI's tendrils creep steadily into my employer's marketing organization.

Right now, it's automating roles we haven't filled and solving problems we haven't been able to solve, but I think that's just a buffer in between deep learning and most of our jobs. Advance warning, rather than solace. Being able to see the imminent transformation is a gift. We get to be scared s***less, but we also get a head start on retooling ourselves.

Ajenstat says Tableau takes a different view. They don't deny the AI change afoot - except to poke fun at the overhype. Tableau sees AI as an augmentation, not a replacement for human analysts:

Right now we live in the hype of AI. Five years ago was the hype of big data, and everything was big data. Now, it's the hype of AI, and how it's going to change everything. And the reality is, it will change everything. But, I think it'll augment people. That's our point of view.

The impact will be pervasive:

[AI] will be fused across everything that we do. From how we connect to data, to how we clean data, to how we analyze it, how we deliver it, and how we alert it. It is this new super power that gets added across the product line.

That puts the onus on Tableau to infuse their software with "smart automation" and so forth:

One insight that we had in the past year is that when we started looking at all the ways that AI could help our users, we started realizing that at the end of the day, AI means smart software. It means doing things automatically that are hard. It means removing challenges... or helping the user solve the problem they had.

Then this vertbatim excchange:

Jon Reed: At one of your customer presentations, a customer was talking about moving from chasing reports from behind, to turning people into true business analysts. I think you could make the argument that if AI's done properly - whatever we call it - that's kind of the end result of that, right? Ideally, it should free analysts up for "higher value roles."

Ajenstat in Las Vegas

Francois Ajenstat: That's exactly right. Because if you have more time, that either means you solve your questions faster. Great. You have more time to answer more questions, and add more rigor, and more depth. Or answer new kinds of questions. And it just adds value throughout.

So what advice does Ajenstat have for business analysts who remain concerned about being automated out of their role? What should they be doing to keep their skills sharp in the context of working with machines and automation?

Ajenstat's three point AI skills plan for data analysts

1. Embrace technology change - "Don't fear it, just embrace it."

2. Know your industry - "Be a master of your domain. So if you're in supply chain, understand how supply chain works. Figure out: are there things that could be automated by machine learning? What are those business roles? And add value in that way. Figure out what are all the patterns that you can analyze, and whether some of them could be operationalized. Credit-card risk monitoring is a good example. Risk is something that you do a million times a day, for every single credit card transaction. Is there fraud here? What's the impact of this thing? That has to be operationalized."

3. Master the aggregate view, and help refine the algorithm - "If you step it up one level, and you look at the aggregate of all those things, that's where the analyst adds a lot of value. Being able to see all these transactions, and see what these patterns are. To help refine the algorithms. So, they're partners in crime. They shouldn't be fearful of that."

During the keynote, Tableau brought out an important point that isn't stressed enough: analysts can play an ethical role, ensuring that algorithms are not just optimized, but fair - striving to remove the bias and the algorithmic weaknesses. Or, as I put it, the unfortunate consequences of brute force automation. Ajenstat:

You need to add human intuition and experience into the equation. You might realize that you need additional data, or additional context that didn't exist. Oh, yeah, sales are down. Well, why? It turns out that there was a horrific hurricane that impacted Houston. That's why they're down. The data just said sales are down.

My take - data analysts must change

I thought Ajenstat's three step view for analysts to evolve was on the mark. We could probably break out part three into a couple subtopics. If I did, it would be:

  • Learn to derive difference-making insights from patterns identified by machines - machines are terrific at identifying patterns in data faster and more efficiently than humans, and they're getting better at recommending actions. But an analyst who can take the patterns and find meaning in them that the business can act on, while ruling out false positives and handling exceptions that need customer-facing skills, are useful analyts indeed.
  • Don't be afraid to improve the algorithms - keep pushing to make the machines smarter, more predictive, more automated, and keep seeking higher ground. The water will rise up regardless so keep pushing yourself.

Now that I'm thinking about it, we could probably add one more step which is: become adept at visual data storytelling. Learn how to capture the imagination of executives or the attention of lines of business matters with data.

Those steps are sufficient in the near-term (four year time frame). After that, I'd be wary of speculating just how sophisticated machines might get. Our AI discussion was short to ensure we could cover all the relevant news, but if I had time, I would have pressed Ajenstat on the urgency of these changes.

I'd go so far as to say those analysts who don't aggressively follow his steps are going to be out of business. Those who are content to port data from Excel, or compile reports without recommending actions. are not going to be employable for too much longer. I suspect Ajenstat's views aren't too far from that, regardless - his skills advisory is worth heeding.

Humans that get ahead of the curve will fare better. They'll do that by excelling beyond machines and, to Ajenstat's point, working with them. Analysts will need both virtues - something I have not emphasized enough.

As for AI-flavored customer offerings, I suspect Tableau will have a lot more to say about that next year. From an automation standpoint, several of the news items this year were designed to automate or ease manual headaches.

For now, it's nice we didn't have to hear about yet another chatbot named after a historical figure, though there are a few names left. I don't think Pocahontas is taken, or, perhaps more appropriate for Tableau, Magellan. For that last topic, the ups and downs cloud BI, I'll have to save those thoughts for another day.

Updated, 7am UK time Friday October 3 with the data storytelling addition to the conclusion.

A grey colored placeholder image