Essential Guide

Browse Sections


This content is part of the Essential Guide: Guide to using advanced analytics and AI in business applications
News Stay informed about the latest enterprise technology news and product updates.

AI analytics expected to rise, along with management complexity

Growing use of artificial intelligence tools by businesses was the focal point of analytics and data management trends predicted by a group of IT consultants and vendor execs.

GRANTS PASS, Ore. -- Business uses of AI analytics applications will grow over the next 12 months, but so will the need to better manage and orchestrate the algorithms that drive them.

Those were among the predictions made at the 2017 Pacific Northwest BI & Analytics Summit, which brought together a group of IT consultants and vendor executives here this month to discuss -- and forecast -- business intelligence, advanced analytics and data management trends. Other topics they eyed included edge analytics and new data privacy rules taking effect in the European Union next year. But artificial intelligence (AI) was top of mind during the predictions part of the proceedings.

Organizations increasingly will find that the diverse pools of big data they're collecting can't be effectively analyzed with traditional tools and human brainpower, said Yves de Montcheuil, a France-based consultant who works with technology startups on marketing and strategy. As a result, he thinks AI software will become more and more crucial to getting real business value from big data applications.

"AI is becoming the new black," de Montcheuil said -- a statement amended shortly thereafter by IBM executive Harriet Fryman, who proclaimed AI to be "the new bacon."

More to think about on AI management

But, as AI analytics becomes more common in corporate enterprises, managing the process is expected to get more important -- and more complex.

Analytics teams will have to pay more attention to "the composition of AI systems," said Donald Farmer, principal of consultancy TreeHive Strategy in Woodinville, Wash. They'll also need to implement detailed governance and oversight procedures "as companies start to put hundreds and thousands of algorithms in place," chimed in Shawn Rogers, senior director of analytic strategy at vendor Tibco Software Inc.

If we're going to live in a world where things are going to be driven by algorithms, we have to be able to convey their ambiguity.
Donald Farmerprincipal, TreeHive Strategy

Gartner analyst Merv Adrian foresees networks of AI-powered tools and devices that can communicate with one another and have the ability to ingest data on their own -- developments Farmer said would make it more clear that data scientists and other analysts are "participants in AI systems" as opposed to users of the technology in a traditional sense.

Another issue to contend with is the level of uncertainty in what AI algorithms predict. Farmer said AI-based analytical models tend to be accurate if they're well designed, but there's almost never a 100% probability that their findings are correct -- something that needs to be made clear to business executives so they don't expect infallibility from the technology. "If we're going to live in a world where things are going to be driven by algorithms, we have to be able to convey their ambiguity," he said.

Edge analytics set to take flight?

In addition to increased use of AI analytics, big data environments are likely to push deployments of "in-flight" analytics applications at the edge of corporate networks, said Mike Ferguson, managing director of U.K.-based consultancy Intelligent Business Strategies Ltd.

As data continuously streams from devices on the internet of things (IoT), mobile apps, stock-trading systems and the like, trying to funnel it all into a centralized data repository for processing and analysis becomes a tall order, Ferguson said. He envisions wider development of event-driven data architectures with edge analytics systems that can trigger automated actions on the fly. "This is a world where the data never stops, and it's completely challenging the way we've done things in the past," he noted.

Companies face a different kind of challenge in complying with the EU's General Data Protection Regulation (GDPR), which establishes stricter rules on data privacy and security for companies that operate in Europe or do business with organizations that handle the personal data of EU residents.

Due to become law in May 2018, the GDPR will require new data governance processes in many companies -- a step that Farmer said could end up contributing to the law's undoing because of the potential for added business costs. "The GDPR could collapse," he said. "The way it forces you to change how you do business will be unacceptable [to a lot of organizations]."

If the law doesn't fall apart, EU regulators likely will find it hard to broadly enforce the new rules because of the large number of companies that are affected, Adrian said. But IT and business executives shouldn't get complacent about their GDPR compliance efforts, he cautioned, saying that he expects EU officials to try to "make an example of someone" to scare other organizations into adhering to the rules.

Next Steps

Check analytics and data management predictions from the 2016 event

Essential Guide: AI technology comes of age for business uses, at long last

AI applications must be free from data bias to produce accurate results

Dig Deeper on AI business strategies

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

The idea of many AI bots communicating with each other, as mentioned by Merv Adrian, has run into a difficult problem. Facebook just shut down an AI problem to do just that when it turned out the bots started to develop their own language that was unintelligible to humans. Unintended consequence of which there will be many.
As Donald said, "If we're going to live in a world where things are going to be driven by algorithms, we have to be able to convey their ambiguity," This is clearly true, but the question is how? That's why I prefer Bayesian nets because, unlike most AI algorithms, particularly machine learning, Bayes models start with the assumption that you actually know something and refine that as new evidence flows in. Insights are always rendered as probabilities, not simple answers. In addition, AI has to become a system of intelligence with Unattended Discovery, Prediction, Justification, Action and Learning. No offering of AI software is capable of this yet. 
Hi, Neil. Thanks for the comments -- interesting points. If AI software can't provide all the functions you listed at the end, are those things that user organizations can piece together themselves? Or is that a wish list of integrated capabilities for the future?
If you thought ERP was hard, wait until you try to integrate AI into your operations. We will need vendors to provide these capabilities, or at least orchestrate and support them as a whole, as a "System of Intelligence." 
For sure. As DL/ML-driven functionality is embedded in ERP apps,  the ERP and AI DevOps pipelines need to be brought into greater alignment. Bayesian logic will form the foundation for more AI-driven ERP apps that "explore then exploit" predictive statistical patterns within complex business processes (e.g, supply chain, logistics, materials mgt) to realize steady-state outcomes (e.g, ensuring on-time deliveries, ensuring zero-defect production runs, etc.). A big part of that converged ERP/AI pipeline will involve continual feedback of process metrics to adjust the Bayesian logic, the neural networks, and other machine-learning algorithms that power the new generation of intelligent/autonomous ERP.
ERP hard? No way! (Ha ha.) So, when do you think integrated AI capabilities of that kind will become reality? One year? Two? Five? Who knows?
It's already begun, but it will be a long process. Right now, a great deal of AI R&D is not aimed at business processes and decision-making, Instead, you see it in cybersecurity, intelligence, defense, robotics, consumer products. To the extent that these breakthroughs bleed into business, it will be the result of fundamental "general AI" whose applications can be applied to running businesses. Cognitive computing, as we know it today, isn't much more than parlor tricks. Look no further than the Watson/M.D. Anderson oncology project. $50 million and they could't get it to work. 

Please see the article on AI in the June issue of Scientific American ("Making AI More Human").  This article is a great rendition from the real cutting edge of research in AI.  Great hope did emerge after neural networks and Bayesian methods were combined with the unevolved state AI had been in for some time.  But, consider closely how well AI can support corporate strategy at this point in time, by itself, without the "man behind the curtain" (Wizard of Oz, 1939).  Recognizing a cat and counting to three are impressive achievements for a machine working by itself only.  But, much more R&D is needed for AI to send business decision makers skyward.

Thanks for the comment, Mike. How long do you think it will be before the "man behind the curtain" isn't needed, if ever?
What kind of uses do you see for AI applications in your organization?
I think we still have a ways to go in actually collecting the data for input into AI processes.  Text mining or NLP, which goes hand in hand with AI, is itself, from a business perspective, still very crude and far from what  AI eventually needs.  My own organization focuses on new ways to describe, measure and leverage the complex data spaces of interest, but still for very narrow industry verticals.  I do see sharper AI applications IF they do not need to apply to all kinds of markets and business processes at the same time.
That's great perspective -- further evidence of the need to separate the hype from the reality on AI. You could say the same thing for just about every technology, of course, but AI is probably pushing the limits of the hype-ometer at this point.
AI is still an uphill task, as it demands so much  resource to compute to get a final result or decision like in the case of IBM Watson. Also as highlighted in few discussions, it can not drill down to reasons and reasons and reasons. DL , ML are helping a lot in Analytics, if properly design.
Good info. Can you add anything about the kinds of analytics your organization is doing or planning with DL, ML and AI technologies?
At the 2017 Pacific Northwest BI & Analytics Summit we also discussed that one of the implementation failures of AI is the artificial component of it and the inability to rationalize the result options. As Donald suggested sometimes the 60% decision is good enough and the added 40% becomes noise that prevents action rather than encourages it. There are other human factors that help with the decision process and we need to keep that in mind. There is immediate value from NLP and machine learning, but less so for a broad implementation of AI in Business Intelligence.
Good points, Suzanne -- thanks for commenting. And let's hope people always have a place in the analytics process, at least to keep an eye on the machines. :-)