Ssas：AI势头、成熟和模型.pdf

TABLE OF CONTENTSINTRODUCTION .4AI IS WORKING .6THOSE WHO HAVE SUCCESSFULLY DEPLOYED AI SEE A STRONG CONNECTION BETWEEN AI AND ANALYTICS .8FULL DEPLOYMENT LEVELS SIGNAL HEALTHY MOMENTUM .10AI OVERSIGHT IS NOT OPTIONAL .12ETHICAL USE OF AI IS TOP OF MIND .14WHILE EMPLOYERS DONT ANTICIPATE A BIG IMPACT ON JOBS FROM AI, EMPLOYEES MAY BE NERVOUS .16LOOKING AHEAD .18ABOUT THIS SURVEY .204If youve ever assembled a large jigsaw puzzle, you know that for most of the time it takes to put it together, it looks like an incoherent mess. Small sections, made up of only a few connected pieces, sit isolated from one another, slowly gaining size as you find pieces here and there. But eventually the momentum shifts, and the glacial pace turns into a race to the finish large chunks of the puzzle are united quickly, and it all snaps into shape.When examining the survey responses of more than 300 executives across industries focused on artificial intelligence (AI) at large companies around the world, it appears likely that we are on the verge of a similar momentum shift. Critical parts of the AI puzzle specific technologies, capabilities, processes and more are being assembled in dif_ferent parts of businesses, in dif_ferent industries, addressing dif_ferent challenges. We are rapidly approaching a “critical mass” moment in which the entire picture comes into view. The responses submitted by many of these executives show a level of enthusiasm and AI-focused activity that point to an explosion of AI adoption just around the corner, even as gaps in capabilities and strategy are revealed. Already, 72 percent of the organizations we surveyed have either deployed AI-based technology or are in the process of doing so.AI IS WORKINGSurvey respondents report that they are encountering real success with AI, on a number of important fronts.FULL DEPLOYMENT LEVELS SIGNAL HEALTHY MOMENTUMMany respondents have moved beyond the experimental phase of AI deployment into more widespread applications.THOSE WHO HAVE SUCCESSFULLY DEPLOYED AI SEE A STRONG CONNECTION BETWEEN AI AND ANALYTICSThose who have deployed AI with more success1were more likely than others to view AI as being strongly connected to analytics.AI OVERSIGHT IS NOT OPTIONALLeaders in AI those who report having successfully deployed AI already are putting processes in place for reviewing the outputs of AI-enabled systems, overriding questionable results and more pointing to new levels of AI maturity.ETHICAL USE OF AI IS TOP OF MINDMost respondents report having ethics-focused processes already in place.WHILE EMPLOYERS DONT ANTICIPATE A BIG IMPACT ON JOBS FROM AI, EMPLOYEES MAY BE NERVOUSConsistent with many reports suggesting that the threat of job loss from AI may be greatly overestimated, these respondents do not believe large-scale job loss is likely. Some, however, are concerned about employee perceptions of the impact of AI.While our survey findings revealed many unexpected insights, here are some of the most important findings, all of which this report examines in detail:1This survey asked respondents “Are you or your team developing or deploying artificial intelligence-based technology?”, with the option of answering “Yes, it has been deployed,” “It is under development,” “We are considering/planning deployment,” “Not at this time,” or “Unsure.” Separately, we asked respondents to “Rate the impact of deployment of AI-based technologies on your operations,” with the option of answering “Unsuccessful,” “Slightly successful,” “Mixed results,” “Successful,” “Highly successful,” or “Too early to tell.” For the purposes of this survey report, we have chosen to define “leaders who have successfully deployed AI” as those who answered “Yes, it has been deployed” or “Under deployment” and also indicated that their ef_forts were either “Successful” or “Highly successful.” 5It is clear from our survey findings that leaders and early adopters in AI are making important advances in key areas, identifying and expanding on what works as they put AI to work in more ways, in more parts of their organizations. Best practices are already emerging. No one can predict the exact shape and trajectory of our AI future. But this survey provides a clearer view than ever of exactly how key elements are combining today to create an AI future. Heres an overview of some of the most important takeaways from the survey, conducted by Forbes Insights in conjunction with Accenture Applied Intelligence, Intel and SAS. AI: A WORKING DEFINITIONAnyone working in artificial intelligence knows that dif_ferent stakeholders define it dif_ferently. When we talk about AI in this report, we mean the science of training systems to emulate human tasks through learning and automation.6AI IS WORKINGPeel away the hype, and the big question is how will AI contribute to solving real business problems.A large percentage of survey respondents report having real success with AI. When looking at only those who have reported having deployed AI, 51 percent say the impact of deployment of AI-based technologies on their operations has been“successful” or “highly successful.” Survey data shows that C-level executives seem less likely to report success than their non-C-suite counterparts 45 percent of those in the C-suite say their AI ef_forts have been “successful” or “highly successful,” with 59 percent of non-C-level executives reporting the same. For some, the scale of their ef_forts is likely to influence their reported success. Regardless, these are encouraging results for a promising technology that is still widely considered to be in its early stages.“Its encouraging to see this level of success being reported by those who are actually using AI,” says Melvin Greer, Chief Data Scientist Americas at Intel Corporation. “And if we want this trajectory to continue, we have to take steps to make sure users feel that they can trust AI. In the early stages of the adoption of advanced technologies, we tend to see that people are very trusting but when that trust is broken, it can set back adoption significantly. We can build trust through improved data literacy, greater transparency, and a sustained focus on the ethics of AI.”As weve seen with many other leaps in technology over the years, greater familiarity is likely to lead to greater trust. “Think about your first ride in a car sharing service, or the first time you used online banking,” says Oliver Schabenberger, Chief Operating Of_f_icer and Chief Technology Of_f_icer for SAS. “In a sense, those represented a leap of faith in newer technologies. That is where we are with AI right now. But even for many sophisticated users, AI still is a black box they put data in, they get an output, and they do not understand the connections between the inputs and the outputs of AI systems. That is a fundamental challenge that has implications on everything from regulatory compliance to the customer experience it even af_fects how we respond to examining biases in our models. Organizations that have adopted AI can illuminate the black box by observing how the model responds to variations in the inputs, and adjusting accordingly.”Rate the impact of deployment of AI-based technologies on your operationsHIGHLYSUCCESSFUL29%30%21%8%13%SUCCESSFUL MIXEDRESULTSSLIGHTLYSUCCESSFULTOO EARLYTO TELL051015202530357What benefits are you seeing, or do you expect to see, as a result of using AI technologies in your organization?WHAT ARE THE BENEFITS?Current Benefits Expected Benefits0 20 40 60 80 9070503010GREATER REAL-TIME PERSONALIZATIONGREATER EMPLOYEE EFFICIENCYFASTER TIME TO INSIGHTS FROM DATABETTER ANOMALY DETECTIONIMPROVED PRODUCT QUALITYIMPROVED CUSTOMER SATISFACTION/RETENTIONINCREASED INNOVATIONFASTER RESPONSE TIMES TO CUSTOMERSREDUCED OPERATING COSTSBETTER RESOURCE UTILIZATIONREDUCTION OR ELIMINATION OF MANUAL TASKSHIGHER ORGANIZATIONAL PRODUCTIVITYIMPROVED CUSTOMER ACQUISITIONMORE ACCURATE FORECASTING roughly a third say analytics will play a “minor” role in AI or no role at all. To what extent do you expect to see analytics having a role in your organizations artificial intelligence?22%33%MINOR ROLE22%MAJOR ROLEMODERATE ROLE6%NO ROLE AT ALL16%FULL ANDCENTRAL ROLE9It seems clear that the more experienced you are in AI, the more likely you are to appreciate the central role analytics will play in your ef_forts. “Among those who have deployed AI, they recognize that success in AI is success in analytics,” says Schabenberger. “For them, analytics has achieved a front and center role in AI. In fact, in many ways, AI is analytics.” Perhaps that is why 66 percent of respondents agree that “AI will enable us to mine massive volumes of data faster to inform business decisions.” “In many ways,AI is analytics.”Oliver SchabenbergerChief Operating Of_f_icer & Chief Technology Of_f_icer for SAS66%AGREE AI WILL ENABLE US TO MINE MASSIVE AMOUNTS OF DATA FASTER TO INFORM BUSINESS DECISIONS To what extent do you expect to see analytics having a role in your organizations artificial intelligence, based on whether you have deployed or considering deploying AI?MINOR ROLEMODERATE ROLEMAJOR ROLEMAJOR ROLEMODERATE ROLEMINOR ROLEFULL & CRITICALROLEFULL & CRITICALROLEDEPLOYEDUNDER DEVELOPMENT/CONSIDERATION5%18%30%47%33%20%46%1%10FULL DEPLOYMENT LEVELS SIGNAL HEALTHY MOMENTUMIf you, like many observers, assume that AI is still primarily a sandbox exercise, these survey results hold some surprises.To be sure, some respondents 28 percent of those who have deployed or are deploying AI indicated that they have AI “partially deployed as an experiment.” No surprises there. Interestingly, though, slightly more respondents (32 percent) report that their AI capabilities are “fully deployed, in multiple use cases or lines of business.” Whats going on here? There are a number of possible interpretations. Rumman Chowdhury, Accentures Responsible AI Lead, suggests that its important to understand adoption rates through the lens of industry. “There are pockets of real progress,” she says. “Just look at the financial services industry, where theres widespread use of AI, both internally and customer-facing. In others, not so much. In retail, for example, only 9 percent report that AI is fully deployed. Among those Im talking to every day, I tend to see much more tentative deployment in the overall marketplace but its accompanied by a ton of interest.”“Generally speaking, the progress of AI is unparalleled,” says Intels Greer. “Weve seen some very positive first impressions regarding how AI can actually be used and we have a long way to go. Weve seen more sophistication from our customers, who are looking for us to be much clearer in our explanations of AI, and in illuminating important dif_ferences between dif_ferent types of AI and analytics technologies from augmented reality and machine learning to deep learning, automated forecasting and many more so we dont treat AI as the hammer to every nail.”How would you characterize the status of your organizations deployment of AI?0 5 10 15 20 25 30 35FULLY DEPLOYED: MULTIPLEUSE CASES/LINES OF BUSINESS32%FULLY DEPLOYED FOR ASINGLE USE CASE14%PARTIALLY DEPLOYEDAS AN EXPERIMENT28%DEPLOYED AS APROTOTYPE OR SANDBOX14%NOT YET DEPLOYEDANYWHERE, BUT IT IS CLOSE13%