AI means Business

Smart Business

AI means Business

Around the world, artificial intelligence (AI) is almost everywhere. In just a few years, it has spread beyond something only technology nerds get excited about, to permeate every realm of business, from supply chains to hiring, manufacturing to marketing, and customer services to medicine. Established companies are spending big to acquire the technology and pundits are not only offering impressive growth predictions but also deployment models for those aiming to seize upon the potential of AI.

by Alan Earls

Applications for artifcial intelligence (AI) are popping up all over the place. For example, French home – improvement retailer Leroy Merlin is running algorithms against historic sales data and other information, such as weather forecasts, to help drive what appears on shelves, and in what quantities, allowing the company to cut inventory costs and improve sales in the process.
Or take Merantix, an AI research organization in Germany, which has spun out a company called MX Healthcare that can analyze mammograms and find indications of cancer with startling accuracy.
“People are sometimes skeptical about the growth of AI just because not all AI is exotic deep learning or extremely complex, a lot of it involves familiar but very useful things, like bots and virtual assistants,” says Greg Schulz, senior advisory analyst at StorageIO. Clearly, there are also plenty of examples of much more advanced deployments, he adds, such as autonomous road vehicles and automated drone delivery undergoing testing around the world. Alex Bekker, head of the data analytics department at ScienceSoft, an IT consulting company, comments, “Nowadays, companies from different business spheres can make use of different types of AI.” He notes the case of a manufacturer which could have robots on its assembly lines with automated visual inspection systems ensuring quality control. The company could also employ deep neural networks to assess the risks associated with their Tier 1 and Tier 2 suppliers – in other words, a mix of functional, visual, and analytic AI. But that’s only the beginning, he says.

A lot of AI involves familiar but very useful things like bots and virtual assistants.
Greg Schulz, Senior Advisory Analyst, StorageIO


In Europe alone, researchers from Ernst & Young, working on behalf of Microsoft, identify hundreds of major companies benefiting from AI (see “Microsoft Study Looks at AI Adoption in Europe”). Likewise, industries, including telecommunications service providers, are poised for transformation and likely to spend more than $11bn on AI by 2025, according to analyst from Tractica in its report, Artificial Intelligence for Telecommunications Applications. The authors outline likely use cases such as network operations monitoring and management, predictive maintenance, fraud mitigation, cybersecurity, customer service, and virtual digital assistants (VDAs) for marketing. According to the Global Artificial Intelligence (AI) Market Report 2019– 2024: Trends, Forecast and Competitive Analysis from market insight frm Lucintel and recently made available through the Research and Markets store, the worldwide AI market is expected to be worth around $71bn by 2024 with a compound annual growth rate (CAGR) of 26 percent from 2019 to 2024. The researchers predict that machine-learning technology will remain the largest segment and likely to see the greatest growth but the impact of this spending on the global economy as a whole is truly eye-popping. An earlier report from Price Waterhouse Coopers, Sizing the Prize: What’s the Real Value of AI for Your Business and How Can You Capitalize?, global gross domestic product (GDP) could rise by up to 14 percent in 2030 as a result of AI investments, or some $15.7tn in net gain. China alone could see a boost of up to 26 percent in GDP, while North America would likely grow at the global rate of 14 percent, it said.

The nature of AI

AI is a very broad phenomenon, says Anne-Laure Thieullent, AI and analytics group offer leader at consulting from Capgemini. “We see three big building blocks of AI technologies with great traction at the moment,” she says. First, and probably the most recent in terms of use cases and uptake, is computer vision which, coupled with deep-learning techniques, can enable features like image classification, object detection, facial recognition, or even emotion recognition. “This has great applications in manufacturing but also, interestingly, in entertainment to get real-time feedback about content from an audience,” she adds. Second, natural language processing is now getting more widespread adoption, Thieullent notes, with cognitive document processing and intelligent content recognition, semantic search for enterprise knowledge, summarization solutions, and even intelligent agents for conversational interfaces. Third is automatic speech recognition (ASR) solutions based on deep learning. These are also ramping up for a variety of applications around sentiment detection, tonal analysis from voice data, keyword identification, text-to-speech, and translation. “All of these can have great applications in customer service centers,” she believes.

Expert Assessment

Despite encouraging beginnings in Europe, illustrated by Leroy Merlin and MX Healthcare, some say the region has been too slow in launching, and too limited in sustaining, its AI initiatives. A 2019 report from the McKinsey Global Institute, Notes from the AI Frontier: Tackling Europe’s Gap in Digital and Artificial Intelligence, says that early digital companies have been the first to develop strong positions in AI, yet only two European companies are in the worldwide digital top 30. Encouragingly, though, Europe has about 25 percent of all AI start-ups.

If Europe scaled up its efforts, the authors note, AI could potentially add up to €2.7tn in GDP to the €13.5tn European economy, dependent on its current set of skills, state of digitization, and other factors. This would translate into a 1.4 percent compound annual growth through 2030. However, the report notes, realizing that potential will depend on achieving a diffusion of skills and knowledge.

AI means bussiness - Merantix Healthcare built Vara

Reducing the Workload: Radiologists are held back by an increasing workload of examinations without any fndings. So Merantix Healthcare built Vara, a platform powered by machine learning which reduces repetitive work for radiologists and enables them to focus on cases which really matter.

According to Thieullent, organizations in Europe are already ramping up their investments and deployments of AI technologies. “We see various interesting use cases deployed at scale in manufacturing, where not only can machine learning help detect failures in production lines or optimize overall equipment efficiency but also computer vision is used to assist in quality defects detection,” she says. For consumer products, the focus is more about using AI to support marketing efficiency, by anticipating market trends from other regions and allowing the trends to modify product launches to fit market specificities in a much more proactive manner, she adds. For retailers, she sees a great uptake of using AI to improve sales-forecast accuracy and decrease inventory costs, as well as progressing on a demand-driven supply chain. “That also helps them with their sustainability agenda,” she says. Public sector and government agencies in Europe are also strongly ramping up their AI investments to improve their services, says Thieullent. Here, the goal is to optimize administrative processes and offer a more digital experience to people. On the positive side of the ledger, Thieullent says European business is “definitely past the early-adopters phase.” However, she warns, a good number of organizations are now in the phase she calls the “AI Death Valley.” Thieullent explains that her phrase stems from how AI has gone through several “winters” in the last decades, where storage and compute power or advances in deep learning research were not yet adequate to fulfll the promises implicit in the mathematical theory behind AI. The Death Valley situation today is similar but not caused by technical shortfalls. Some organizations, she explains, are in a situation where a lot of different pilots, proofs of concept (POCs), and minimum viable products (MVPs) have been launched but often with neither a clear business strategy nor a strong operating model – something one expert at the recent Industry of Things Expo in Berlin referred to as “proof-of-concept hell.”

In Europe, a good number of organizations are now in the AI Death Valley phase.
Anne-Laure Thieullent, AI and analytics group offer leader, Capgemini


Thieullent observes that many AI deployments require cloud computing from the very beginning but many companies are still at the start of their cloud journey. On the other hand, many initiatives are currently underway globally to boost AI. For example, within the European Union, the “France is AI” initiative is now gathering together many companies and start-ups, and the AI4EU collaborative platform got going earlier this year. One area where Europeans may be ahead of the curve is in their focus on the ethical aspects of AI (see “Can AI Be Evil?”) – proactively addressing potential biases in data sets or algorithms, building explainability and visibility into AI solutions, and adopting a more transparent approach about the finality and intent of AI applications. “Companies like Telia have published clear ethical AI guidelines to provide a framework for these applications,” says Thieullent. With greater sensitivity toward data privacy as well as the trust and consent of the general public, especially after the Cambridge Analytica incident, this may turn into a competitive advantage in the long run for European organizations that will implement a human-centric approach to AI – or an AI that makes sense to humans, she says While conditions for AI adoption and expansion may not be perfect, companies and organizations around the world are moving ahead. Cogito, a young US company with roots in the Massachusetts Institute of Technology’s (MIT) Human Dynamics Lab, trains machines to detect and interpret the social signals in human communication. The company now offers in-call guidance to call center agents for every phone conversation.

AI Meets IoT

Hong Kong-based Orient Overseas Container Line (OOCL) provides shipping containers for the world market and has been applying AI extensively in its operations. The company recently upgraded its MyOOCLReefer (MOR) service for refrigerated containers by combining AI, Internet of Things (IoT), and mobility to provide transparency, visibility, and convenience to shippers when monitoring their cargoes.

AI means bussiness - failure is no option

When Failure Is Not an Option: Machine learning can help to detect failures in production lines and to optimize overall equipment effciency. Already, computer vision is being used to assist in automated quality control systems.

In Vietnam, agriculture is get ting a boost through Sero’s crop monitoring which uses AI to analyze photographs and identify likely diseases or infestations. The system will have the ability to diagnose and recommend treatments to farmers. BP, the British multinational oil and gas company, has invested $20m in Beyond Limits, an AI company with roots in NASA’s Jet Propulsion Laboratory, to help it accelerate the delivery of AI software that shows promise of offering the energy sector new levels of operational insight, business optimization, and process automation.

AI allows us to improve the safety and reliability of our infrastructure and production process.
Meghan Sharp, Managing Director, BP Ventures


A hope of the partnership, according to BP, is that it could enable a change in the way it locates and develops reservoirs, produces and refines crude oil, and markets and supplies refined products. Beyond Limits’ software will help support improvements in the speed and quality of decision-making and manage operational risks by harnessing the collective knowledge and experience of BP’s experts.

Beyond Oil and Gas

The company says it hopes the software will also allow the oil company to improve the safety and reliability of its infrastructure and production processes. Meghan Sharp, managing director of BP Ventures, believes the investment is an example of BP’s ongoing support of entrepreneurs and innovators that goes beyond the traditional world of oil and gas. Watson, IBM’s famous cognitive supercomputer, has been applied extensively in health-care management both within the US and globally. For example, it supports a Care Manager function that can sift through both structured and unstructured data to help tailor care programs in conjunction with human medical professionals. Watson has also been applied to the hunt for new therapeutic drugs and in optimizing cancer treatment and care using historic data and patient information to fine-tune regimens to the needs of the individual. Dreamstime is a European stock photography company which has started to use AI to improve the experience of its website users, for instance during the photo-vetting process. Horia Beschea, an AI specialist working with the company from Bucharest, Romania, explains: “Before photos are posted onto the website, AI is used to sort through them. That allows us to distribute content to our users at a much faster rate than ever before.

AI allows us to distribute content to our users much faster than ever before.
Horia Beschea , Dreamstime


Our AI models recognize human models in images, image type, and content that should be filtered [e.g. adult/health/violence] and run on all new images at once.” Applying AI allows Dreamstime to get an automated understanding of the image content and its potential value as stock photography. Freed from the onerous and timeconsuming tasks of sorting images, editors can focus on quality issues. In Barcelona, ForceManager, which specializes in mobile CRM, says it is the first in Europe to incorporate machine learning and conversational AI technology (along the lines of Siri and Alexa) to help field sales representatives working away from the office. The system delivers insights on upcoming deals, recalls data from previous visits, and even recommends certain products or services for promotion to specific customers. “We’re seeing many of the consumer AI trends carry over to business to cut out menial tasks and drive efficiency,” says ForceManager’s cofounder and CEO Oscar Macia.

AI is a huge win for an industry that relies on heavy reporting.
Oscar Macia, ForceManager


One of the company’s creations is a virtual, AIbased sales assistant called Dana. On average, according to Macia, field sales reps spend 63 percent of their time on non-selling activities. With Dana’s help, they can use their commuting time to report on a meeting in real time and stay up to date on their pipeline. “It’s a huge win for an industry that relies on a heavy reporting funnel to survive,” he says. AI-enabled artificial assistants are in many ways similar to what’s on offer from LivePerson, a provider of conversational commerce software that can work semi-autonomously or in concert with employees. “We are changing the very nature of brand–customer interaction,” says Moritz Fischaleck, a product evangelist at LivePerson in Berlin.

Conversation starters

LivePerson has access to data that helps inform the brand–consumer relationship and interaction, he explains. This data can be leveraged to reveal within each product category and subcategory why people are reaching out to customer care. The system is built on 24 years of customer call center data. According to Fischaleck, more than 18,000 companies currently use LivePerson. Finally, Stephane Rion, senior deep learning scientist for Teradata in France, says his company is delivering AI customized solutions “built from the ground up” and based on the client’s requirements, using a blend of the latest open-source technologies and Teradata’s Vantage analytics platform. “We deliver fraud detection solutions based on deep recurrent neural networks, financial products recommendation systems, and document processing and automatic validation for the back office,” Rion says. Teradata is currently working with Abanca and other major banks in Spain on the implementation of a solution to accelerate the loan acceptance process for bank customers.

there needs to be a single point of truth to avoid duplication, stale data, and silos.
Stephane Rion, Teradata


The product is up and running and enables the validation of hundreds of loan requests per day, Rion claims. It processes and classifies necessary client documents (proof of address, pay slips, etc.) for a loan request using natural language processing techniques and machine learning. It also extracts and validates specific information such as national insurance numbers and signatures from the documents using optical character recognition and deep-learning models. Beyond those adoption stories and the range of views on AI’s progress, Thieullent at Capgemini sees three big impediments to success. The first is that the different AI initiatives aren’t necessarily focused on the right use cases or that these use cases aren’t aligned to the organization’s strategic objectives and therefore cannot scale from a business perspective. The second problem is that it is still too difficult for IT departments to put AI solutions into production, often either because their data landscape is not managed well enough for the right data to be used in a recurrent manner or because their cloud strategy hasn’t been fully implemented (see “Top Down: Why AI Initiatives Need CIO Support”).

Microsoft study looks at AI Adoption in Europe

A recently completed study conducted by Ernst & Young for Microsoft ex amined the outlook for AI – in Europe for 2019 and beyond, as well as current practices. Of the 307 companies surveyed, 59 percent say they expect AI to have a major impact on aspects of business that are “entirely unknown Strong Commitment to AI Microsoft study looks At AI Adoption In europe to the company today” – though only about four percent indicated that their own use of AI is currently making a large contribution to operations or could be considered to be “advanced.” More than a quarter of the respondents say they have already put AI to use and over 60 percent claim to be in the planning stages. Overall, the researchers found a strong, and apparently permanent, commitment to AI spending and innovation and the report concluded, “It is no understatement to suggest that AI will be a chief protagonist in the change transcending all elements of business in what has been labelled the Fourth Industrial Revolution.”

Microsoft study looks at AI Adoption in Europe

AI solutions are sometimes doomed to stay in proof-of-concept hell for an unreasonable amount of time, Thieullent explains, before making it into production, where they are fully integrated into the IT landscape in a recurrent manner to serve business users. The third impediment, she says, is that organizations still haven’t completely figured out what the right operating model is for AI to “work” at scale for their teams. For instance, she notes, they need to think through how to program manage the initiatives, how to fund them, and how to properly measure and recognize tangible success and business outcomes. Implementors also need to think about how to prioritize the next set of use cases, how to build the right skills, and how to harness the necessary transformation of the workforce. Solving these three big aspects of scale – business adoption, production-grade technology, and operating model for AI at scale – will be conditional for AI to become fully mainstream in the next 12 to 18 months; and organizations that will succeed in this will become clear winners in their market, Thieullent believes.

A Single Point of Truth

Rion at Teradata warns that, as a practical matter, a strong data foundation is vital to take advantage of AI. “This means the data needs to be in the right format so that it can be exploitable by the data science or other analytics team. There needs to be a single point of truth to avoid data duplication, stale data, and silos, and there also needs to be enough of it. Volume and variety of data is crucial to build performant machine-learning-based solutions,” he says. In addition to this, a clear AI strategy roadmap defining some simple use cases to start with is vital.

AI means Business - Finding the Right Framework

Finding the Right Framework: A clear AI strategy roadmap defined by the business stakeholders and the data science team is crucial for the smooth deployment of successful AI prototypes into production. “Don’t just jump into AI for the sake of jumping into AI,” says Teradata’s Stephane Rion (click to resize).

“These should quickly provide returns on investment and are typically defined by the business stakeholders and the data science team through workshops and discussions,” he maintains. Finally, a robust methodology designed by data architects and the development team to ensure the smooth deployment of successful AI prototypes into production is also key to avoid delays or even potential failure, Rion adds. Above all: “Don’t just jump into AI for the sake of jumping into AI. First consider where it could have a real impact in your organization,” Schulz of StorageIO advises. “Look at some of the easier entry points. The leading cloud providers such as Azure and AWS [Amazon Web Services] offer many powerful cognitive AI and machine-learning tools that can give an organization a good start without making a huge investment,” he concludes.

Tags: /

Leave a Reply

Your email address will not be published. Required fields are marked *