Shulou(Shulou.com)06/03 Report--

Https://www.toutiao.com/a6637266246809158158/

2018-12-21 10:22:52

From building an isolated proof of concept to a lack of defined criteria for success, a series of pitfalls can undermine the business prospects of your artificial intelligence project.

In recent years, the business enthusiasm of enterprises for artificial intelligence has been increasing. Global corporate spending on cognitive and artificial intelligence systems, from chatbots to deep learning, plus the infrastructure that powers them, will more than triple from $24 billion this year to $77.6 billion in 2022, according to IDC's latest forecasts.

More obviously, artificial intelligence has changed from an early adopter to a mainstream business use case, with a wide range of organizations exploring pilot projects and putting artificial intelligence into production in almost every industry. But this does not mean that the process is foolproof and risk-free. If you don't want to waste the money you're going to spend on artificial intelligence, here are some common mistakes to avoid.

grasp all, lose all

"Don't try to boil the ocean on the first day," Lance Olsen, head of Microsoft's cloud artificial intelligence team, told CIO.com. You can't use artificial intelligence to change your entire business decision-making process overnight, so it's best to start small and take gradual steps while you gain expertise.

Look for the fruit that you can easily pick. Before you can deal with the most important system, you need to develop a system to experiment and verify the experimental results. "Don't make the biggest investment in the first place," he warned. "

Establish an independent proof-of-concept system

Building an one-off artificial intelligence system does not help you create the overall flow of artificial intelligence, nor is it part of your existing data pipeline, which will not take you far forward. You need to create a sustainable AI asset for each project. Here, sustainability means a system that generates a sufficient return on investment, and you can continue to invest in the system for further development and expansion. Every time you do this, it will help to create AI capabilities for the entire business, not just a new tool for a particular team.

Based on the business analysis you are already doing, turn these historical systems into predictive capabilities. "start with investment optimization, use your existing pipeline and optimize what you've already done," Olsen said. " Then you can move on to more revolutionary projects and make bigger changes to the way your process works.

Lack of a suitable technical infrastructure

According to a recent McKinsey report, you need to invest in core and more advanced digital technologies before you start artificial intelligence. Companies that already have expertise in cloud computing, mobile and web development, big data and analytics are three times as likely to adopt artificial intelligence tools. 3/4 organizations that use artificial intelligence say they rely on what they have learned from existing digital capabilities. In other words: if your business is not ready to take advantage of cloud computing and data analytics, you may not be ready for artificial intelligence.

Lack of data

The vast majority of artificial intelligence systems, including those built by enterprises themselves, are machine learning systems, and machine learning requires data. As Julia White, vice president of Microsoft, said at the company's recent business AI event, "where will our new robots be? or where will the new robots learn?" In fact, without good data, AI will hurt you rather than help you, because you are actually confident in something that has no actual evidence.

In addition, if you only have the same public data as your competitors, you will only get the same insight as your competitors, so you need to use your own unique data. Assuming you have collected the correct data, it also needs to be cleaned up and standardized.

Don't underestimate the investment needed in this area; collecting and cleaning up data usually accounts for about 80% of the work of data scientists. Starting with the data you already use for business intelligence and analysis, you can also make it easier to ensure that your AI system will support key business processes, making it more likely to work. This will also help you define the tools and processes for data preparation that you can use to deal with unused data.

Lack of criteria for assessing and measuring success

Data science is science. You need to have an assumption about how to improve business decisions, sales, customer support, or anything else you want to do with AI, and you must test and evaluate the results in action.

This means that you need to design how to measure the success of a project-both in terms of adoption and results. This can translate into aligning the project with employee performance deadlines, such as a 90-day outlook for sales and marketing teams, or an hour quota for contact centers. This also means having a control group that does not use the new system, because if you invest a lot of money in developing a new system, you may come to an overly subjective conclusion. You need to make sure that people are making data-driven decisions rather than relying on intuition; if they habitually ignore data, it won't help even if artificial intelligence tools present it to them. You also need to decide in advance what success looks like, because this is the hypothesis you are testing. Do you want more customer orders or bigger orders? Do you want to reduce customer support calls, or do you need less time to deal with calling customers?

At first, I don't know what problems artificial intelligence can solve for you.

The problem with the word "artificial intelligence" is that it makes people feel that anything is possible. The artificial intelligence industry has made great strides in the past few years, but you still need to know what artificial intelligence can actually provide and how it will be integrated into your existing systems and business processes. Then you need to know what's wrong with your organization and what artificial intelligence can help you solve. You can't adopt artificial intelligence just because other companies are using it.

"executives need to think about two things before turning to artificial intelligence," Jacob Davis, senior director of analytical services at Cheetah Digital, told CIO.com. " First, what do we really want to solve? How do we solve this problem now and grasp the data at hand? If you can't come up with something, even in theory, then AI may not be able to help you to the extent possible in your current state. Second: do I think about artificial intelligence because I hear a lot of hype about it? You have to really assess your desire for this kind of solution, otherwise you may invest a lot of money in something that doesn't add real value. "

There is neither the right talent nor the right project.

You will need professional knowledge of data science, and if you do not have a dedicated data science team, then this expertise will usually be built in the IT team. Wherever it is, it is important not to isolate it in a center of excellence. A recent Ovum study of 2000 organizations that produce artificial intelligence projects around the world for data science software provider Dataiku shows that for a project to be successful, these experts need to participate in the business team they are solving, as well as project management and development teams. Because the central team may miss out on some cultural differences in local business units.

"again and again, we see companies and cross-industry teams around the world unable to start their data work because they can't help these people in different regions work together-let alone different skills and different types of people," said Florian Douetteau, chief executive of Dataiku. If you can't keep data science experts in key positions permanently, you can use the collaboration and knowledge transfer of central experts to help develop local data science skills.

Build your own artificial intelligence

While IBM Watson is well known, even pre-built artificial intelligence services require time and expertise to integrate with your own systems and processes and must be carefully evaluated, but few companies have the expertise to build everything from scratch. Today, AI tools are increasingly built into the marketing cloud of SaaS products such as Salesforce, Dynamics and Adobe, although they are likely to be plug-ins for which you have to pay extra. Azure, AWS and Google also provide cloud computing machine learning services, which can provide specific "cognitive services", such as machine vision and speech recognition, which you can customize and build into your own tools and services, or you can provide a library of general solutions that you need and can adapt to. Use these tools to get started quickly, and then consider what other models and tools need to be built from scratch, as employees are more adapted to the productivity benefits of artificial intelligence.

Expect artificial intelligence to replace manpower completely.

Like automation, artificial intelligence will bring you the greatest performance and productivity improvement when humans and artificial intelligence systems work together. A recent study in the Harvard Business Review shows that performance has improved four to seven times as companies adopt more and more human-computer collaboration. To achieve this collaboration, business teams need to be involved in assessing what artificial intelligence systems can actually do for them. Artificial intelligence tools that provide advice, options, decision support, and upgrades to experts are obviously more useful than those that give simple yes/no answers without any human involvement.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.