The following is the main content of the article:
Although the overall popularity of artificial intelligence in the enterprise is still low (about 20% in our last study), executives know that artificial intelligence is not just hype. Organizations in various industries are paying close attention to this technology. Look at what it can do for their business. They should do the same. According to our estimation, 40% of the potential value created by today's analytical technology comes from what is called 'deep learning' (using multi-layer artificial neural networks) Artificial intelligence technology. In general, we estimate that the value of deep learning can be between $3.5 trillion and $5.8 trillion a year.
However, many business leaders are still not sure where they should apply artificial intelligence to get the most out of it. After all, embedding artificial intelligence into the entire business requires huge investments in talent recruitment and technology stack upgrades, and radical changes are needed. Initiatives to ensure that artificial intelligence delivers substantial value, whether it helps make better decisions or improve consumer-facing applications.
Through an in-depth study of more than 400 artificial intelligence use cases across 19 industries and 9 business functions, we found that it is most appropriate to use an old adage to answer the question of where to deploy artificial intelligence. Money goes '.
Business areas that traditionally bring the most value to the company are often areas where artificial intelligence can have the greatest impact. For example, in retail organizations, marketing and sales often bring great value. Our research shows that only customer data With artificial intelligence for personalized promotions, the incremental sales of brick-and-mortar retailers will increase by 1-2%. By contrast, in high-end manufacturing, operations often bring the greatest value. Here, labor Intelligence can be predicted based on potential causal drivers of demand rather than previous results, increasing forecast accuracy by 10-20%. This means that inventory costs may be reduced by 5% and revenue may increase by 2-3%.
Although the application of artificial intelligence covers a wide range of functional areas, in fact, in these two cross-cutting areas - supply chain management / manufacturing and marketing and sales - we believe that artificial intelligence can play the most in several industries The power of this, at least for now, is combined. In summary, we estimate that these use cases account for more than two-thirds of the total artificial intelligence opportunities.
Artificial intelligence can create a value of $1.4-$2.6 trillion in marketing and sales for global companies, creating $1.2-$2 trillion in value in supply chain management and manufacturing (some of which are owned by the business and some of which are owned by the customer) In manufacturing, the greatest value from artificial intelligence comes from using it for predictive maintenance (about $0.5-0.7 trillion in global companies). Artificial intelligence can handle audio and video, including audio and video. A large amount of data means that it can quickly identify anomalies to prevent malfunctions, whether it is a strange sound from an aircraft engine or an assembly line fault detected by a sensor.
Another way for business leaders to determine where to deploy artificial intelligence is to look at those functional departments that are already leveraging traditional analytics. We have found that in the use cases where artificial intelligence can create the greatest potential value, neural network technology Can perform better than existing analytical techniques, or generate additional insights and applications. In our research, 69% of artificial intelligence use cases are the same.
In only 16% of the use cases, we have found a 'greenfield' artificial intelligence solution that works well for other analytical techniques. (As algorithms become more versatile, the various data they become feasible become Easier to obtain, the number of deep learning use cases may increase rapidly, and the proportion of 'greenfield' deep learning use cases may not increase significantly, as more mature machine learning techniques have the potential to become better and more common.
Even if we see the economic potential of the use of artificial intelligence technology, we recognize the practical obstacles and limitations of artificial intelligence implementation. It is a great appetite to obtain a large enough and comprehensive data set to meet the profound appetite of deep learning for training data. Challenges. People are increasingly worried about the use of such data, so this is also a challenge that companies need to address. Security, privacy, and the possibility of passing human biases to artificial intelligence algorithms need to be addressed. In some industries, such as health care and insurance, companies must also try to explain the analytical results of artificial intelligence in a simple language: Why does this machine come to this answer? The good news is that these technologies are progressing and starting Solve some of these limitations.
In addition to these limitations, companies may face even more difficult challenges in adopting artificial intelligence. Mastering technology requires new levels of expertise, and processes may become a major barrier to successful adoption of technology. Will have to develop a robust data maintenance and governance process, and focus on the 'first kilometer' – how to acquire and organize data and efforts – and the more difficult 'last mile', ie how to output the artificial intelligence model Integration into the entire workflow, from clinical trial managers and sales managers to purchasing staff.
Although companies must be vigilant and responsible when deploying artificial intelligence, given the scale of the technology and its beneficial impact on businesses, consumers and society, it is well worth investigating. This pursuit is not simple, but it can be Follow a simple concept: Follow the money.