Enterprises know they need AI, but they have concerns about how to safely implement it.
In a Recent Survey of Enterprise Business Executives, 98% of Them Revealed They Were Exploring, Piloting, Rolling Out, In Early Production, or Looking to Scale AI in Their Enterprise. In a White Paper Published by Opaque and a Passel of Professors from MIT and The University of California at Berkeley…
In a white paper published by Opaque and a passel of professors from MIT and the University of California at Berkeley, and Bret Greenstein, partner, data analytics and AI with PwC US, explained it this way: “People realize there’s just simply so much upside here. If they can figure out how to do it safely and do it in a well-governed way which addresses data privacy and security, they can drive disruption in their sector.”
So, what’s holding them back? Three things:
1: Making sure their AI systems are built on a high-quality data foundation
An early IT meme (before memes were even a thing) went “Garbage In, Garbage Out.” Another way to put it might be: Your enterprise AI implementation will only be as good as the data you give it.
2: Set a clear use case and a strategy to implement it rapidly
Different business units will have different reasons for putting AI to work for your enterprise; IT will cite competitiveness; HR will advocate for the best possible employee experience. Regardless, the reasons fall into three broad categories: revenue, customer/employee experience, and staying competitive.
3: Bridging the skills gap
While 80% of respondents to the Opaque survey acknowledged that AI and ML improves employees’ efficiency, 74% said that senior management lacked the broad knowledge to leverage AI at their organizations. It will take time to build, or even hire, those skills to bring them in-house. A temporary solution, perhaps even a permanent one, would be to contract out to a vendor with proven skills and success at implementing and operating enterprise-level AI.
OK, it’s not going to be as easy as 1-2-3 to make your enterprise exponentially more ‘intelligent.’ But those who have made the leap suggest the ROI is anything but artificial.
AI could be the most transformative technology of our lifetime. Unsurprisingly, people still aren’t quite sure what to do with it. This is especially true at the enterprise level, where AI and machine learning (ML) – a subset of AI – have long been on the horizon but benefits from their practical use have remained elusive due to difficulties related to existing software stacks, workflows, organisational cultures and budgets. Questions about responsible AI and ML make these conversations even trickier. Many decision-makers lack direction and understanding as they attempt to envision, justify, deploy and leverage AI and ML. As this report finds, there are three specific challenges they must overcome if they hope to get the most from these technologies: 1 Ensure AI is built on a high-quality data foundation. 2 Establish clear use cases and a strategy to move quickly. 3 Overcome the skills gap. A guide for a pivotal time Workday partnered with the international research specialist Vanson Bourne for an independent assessment of AI’s use so far. The goal: capture the sentiment around this remarkable technology and explain what people are really doing, thinking and feeling about the most significant disruptor many have ever seen – and then provide a way forwards. The challenges mentioned above were direct results of this research, and you’ll see many more insights below. But despite the challenges, there is also cause for optimism. Four out of five respondents agree that AI and ML are necessary in order to keep their business competitive, and two-thirds say that AI and ML have already increased productivity and operational efficiencies at their organisation. This report begins with a detailed summary of key findings and continues with thorough discussions of those three key challenges. It ends by acknowledging another key takeaway from the data: almost nobody feels ready to go it alone with AI and ML. Fortunately, they don’t have to.
Findings at a glance The research has generated a wide range of takeaways – some of them rather surprising. To summarise: • Many organisations are already using AI and ML, though they are likely hindered from achieving full implementation. • Decision-makers are facing pressure – often from above – to do more with AI and ML. • Organisations clearly intend to invest more in AI and ML in the coming years. • Those who have deployed AI and ML are already experiencing tangible ROI across many use cases. • Those leveraging AI and ML have reservations around data quality and privacy, but few clear answers. • Many teams worry about ensuring that personnel have the right AI-related skills
Facts and figures Seeing the challenges • Three-quarters of decision-makers agree that there are “many hindrances” preventing their organisation from fully implementing AI and ML. • 77% are concerned that their organisation’s data is neither timely nor reliable enough to use with AI and ML. • 72% say that their organisation lacks the skills to fully implement AI and ML. Acknowledging pressure and risk • 80% of decision-makers agree that AI is required to keep their business competitive. • 76% say their knowledge of AI and ML applications needs improvement. • 77% say that uptake of AI and ML at their company would increase if risk could be reduced. • The biggest organisational risks of implementing AI and ML: 1. Data security and privacy 2. Concerns on accountability 3. Inability to measure ROI 4. Decision-making errors Anticipating the workforce of tomorrow • 45% of decision-makers say AI and ML will benefit workers by augmenting workloads and creating new career paths. • 43% warn that AI and ML will replace some tasks, causing unemployment among some workers. • 12% are more doubtful, saying that AI and ML will replace humans completely and have a negative impact on workers. Addressing responsible AI • More than 9 in 10 (93%) of decision-makers believe it is important for a human to assist AI or ML when making significant decisions, rather than allowing the technologies to do it alone. • Only 29% are “very confident” that AI and ML are currently applied ethically to businesses; 52% are “very confident” it will be applied ethically within five years
Figuring out use cases Almost identical proportions of respondents said they have implemented AI and ML to assist with the following finance-related tasks, the exceptions being procurement and sourcing, which have greater levels of usage currently in place: • Improved forecasting • Automating non-strategic tasks • Scenario planning • Risk and fraud detection Feeling optimistic Strong proportions of respondents say their existing AI and ML deployments have improved key business indicators: • Communication and data visualisation • Procurement • Sourcing Consistency was also evident among HR professionals discussing HR-related tasks: • Recruiting and applicant tracking • Business analytics • Skills assessment tools • Skills management • Learning management • Talent management • Payroll • Time and absence management
Feeling optimistic Strong proportions of respondents say their existing AI and ML deployments have improved key business indicators: • Communication and data visualisation • Procurement • Sourcing Consistency was also evident among HR professionals discussing HR-related tasks: • Recruiting and applicant tracking • Business analytics • Skills assessment tools • Skills management • Learning management • Talent management • Payroll • Time and absence management 20% 40% 60% 80% Employee 100% experience Customer experience Company revenue Risk, fraud and compliance 65% 67
The ROI of AI Among companies measuring the ROI of their AI and ML deployments, the results are favourable. Poor – below expected ROI Good – at or near expected ROI Excellent – exceeded expected ROI Reaping the benefits already All respondents (n=1,000): What do you consider to be the main business benefits from investing in AI and ML? (Combination of responses ranked first, second and third.) 41% 38% 35% 34% 34% 33% 32% 31% Better decision-making and next best action (NBA) business insights Automating business processes Reskilling or upskilling employees Improved employee retention / experience Enabling higher levels of productivity Increased revenue and profits Reduced headcount and hiring costs Lowering costs 8% 42% 37
Reaping the benefits already All respondents (n=1,000): What do you consider to be the main business benefits from investing in AI and ML? (Combination of responses ranked first, second and third.) 41% 38% 35% 34% 34% 33% 32% 31% Better decision-making and next best action (NBA) business insights Automating business processes Reskilling or upskilling employees Improved employee retention / experience
% Better decision-making and next best action (NBA) business insights Automating business processes Reskilling or upskilling employees Improved employee retention / experience Enabling higher levels of productivity Increased revenue and profits Reduced headcount and hiring costs Lowering costs 31%
Companies are on the way All respondents: What most closely describes the stage your organisation is at when it comes to implementing AI and ML? AI is here to stay of respondents’ organisations are investing in AI, according to IT decision-makers. 94% of respondents expect their level of investment in AI and ML will stay the same or increase in the next year. 83% 16% Exploring Piloting Rolling out In early production Looking to scale No plans Don’t know
16% Exploring
16% Piloting
26% Rolling out
26% in early production
14% looking to scale
2% no plans
0% don’t know
AI is here to stay of respondents’ organisations are investing in AI, according to IT decision-makers. 94% of respondents expect their level of investment in AI and ML will stay the same or increase in the next year. 83% of respondents expect their level of investment in AI and ML will stay the same or increase in the next year.
Challenge 1: ensure AI is built on a high-quality data foundation If you’ve paid attention to the range of publicly accessible AI tools recently, you’ve probably noticed that their outputs are only ever as good as the information fed into them. AI-generated content, for example, relies on a discrete sample of data – today this is often a portion of the internet – for its processing. You can already guess that bad data can make even the most sophisticated AI and ML tools go wrong. In fact, models built on high-volume, largely unstructured data with varying degrees of quality provide less value and accuracy than models built on smaller datasets of high quality. The ideal, of course, is high quantity and high quality. As a result, business leaders have questions. Among them: • Do we have the right volume of data for AI and ML? • Is the data of the right quality? • Is the data structured in a way that can be easily used? • Can we keep it secure and maintain privacy?
77% of respondents are concerned that their organisation’s data 77% is neither timely nor reliable enough to use with
Among AI IQ respondents, 77% are concerned that their organisation’s data is neither timely nor reliable enough to use with AI and ML. Similarly, insufficient data volume or quality was the top reason (29%) for their AI and ML deployments falling short of expectations. And, deep within these responses was another fascinating insight: only 4% of the sample said that their deployments had not fallen short of expectations. We find it interesting that when it comes to pure ROI (see chart above), AI and ML implementations are performing very well, but when it comes to expectations (which are more subjective), it’s a different matter. Kim Morick, global talent data technology leader for IBM Consulting, puts the data issue this way: “Every AI solution learns from data. And, you can only see the data that you have access to.” This is an organisation’s first task. Before they can realise the benefits of AI and ML, leaders must ensure that their data – and their ability to validate, access and move it – will meet the challenge. They also need to consider leveraging existing solutions that draw on data from existing sources, and whether or not their software architecture can successfully accommodate these things. How good is your data foundation . . . How good is your data foundation . . . really? For many, a shaky data foundation stands in their way. When data exists in various silos, each with different structures, those silos can’t talk to each other, and there’s no single source of truth for AI to learn from, leading to questionable results, reduced user confidence and failed AI and ML implementations. With AI and ML continuously learning from all available data, it is also important that any new data is available to the AI as soon as it is created. This helps explain why so many respondents are concerned about the timeliness of their data – they know that’s a vulnerability especially relevant to AI and ML. A key opportunity with generative AI is unlocking the unstructured data of your business. Bret Greenstein, Partner, Data, Analytics and AI, PwC US, summarised this clearly: “People realise there’s just simply so much upside here. If they can figure out how to do it safely and do it in a well governed way which addresses data privacy and security, they can drive disruption in their sector.”
Challenge 2: establish clear use cases and a strategy to move quickly Almost everyone feels pressure to move quickly with AI, and that pressure mostly comes from the top. Interestingly, however, the motivation for the pressure varies. • Leaders in IT feel the pressure to be more competitive. • For leaders in HR, the pressure is about improving the employee experience. • Finance decision-makers say they’re being asked to address a skills gap. This clarifies that, along with speed, use-case alignment is a critical challenge. Regardless of use case, almost all respondents see the benefits of adopting AI and ML, and those benefits often fall into consistent benefits categories: • Competitiveness • Revenue • Customer and employee experiences “ If you don’t have an AI strategy, people will feel like this is not a company for the future. I would say that there’s no public company that is not preparing itself to be asked each quarter what its AI strategy is, even if it’s not its core business. Spiros Margaris Top Global Influencer in AI and Fintech • Risk, fraud and compliance • Working more efficiently •
These are common topics in any future-focused discussion. As Spiros Margaris, founder of Margaris Ventures and a top global influencer in AI and fintech, puts it, “If you don’t have an AI strategy, people will feel like this is not a company for the future. I would say that there’s no public company that is not preparing itself to be asked each quarter what its AI strategy is, even if it’s not its core business.” However, there are hints that some decision-makers don’t yet grasp how AI and ML maps to specific use cases and will coexist with their human talent. This brings us to the strategy portion of this discussion – concerns about how to best augment rather than replace human talent and keep the right stakeholders in the loop.
The human stays in the loop Respondents show shared concern for the AI and ML relationship to human talent. As mentioned above, decision-makers overwhelmingly believe it is important for a human to assist AI or ML when making significant decisions. So as organisations navigate this changing world of work, one of the key opportunities for AI and ML is to intelligently augment human experiences across finance, HR and IT. This frequently happens through intelligent automation that provides supporting information and recommendations while keeping humans in control of all decisions. In this way, AI and ML are accelerants: humans are freed to make faster, better decisions and focus on the strategic conversations required to evolve the organisation. AI and ML are accelerants: humans are freed to make faster, better decisions and focus on the strategic conversations required to evolve the organisation.
Ensuring responsible, transparent AI A related challenge is establishing and enforcing standards that can reliably govern innovation around AI and ML. A critical mass of respondents (39%) put potential bias among the top three risks for their organisation when implementing AI and ML, and we know this factor is on most leaders’ minds. Organisations understand this needs to be addressed. Only 29% of respondents are “very confident” that AI and ML are currently applied ethically to businesses, but 52% are “very confident” it will be applied ethically within 5 years. Several key ethical principles should underpin AI and ML, including building AI and ML models specifically to mitigate risks, and then providing transparency and auditability after they’re in use. Organisations will continue to be well served by making sure that no decision is fully controlled by AI and ML technology, and that people are the final decision-makers and are kept in the loop at the right places. Regardless of your vendor, challenge them to back up their ethics principles just as rigorously as they back up their ROI.
Challenge 3: overcome the skills gap The research surfaces some interesting tensions related to skills, implementation and the real or perceived attainability of AI and ML. For example, 80% of respondents say that AI and ML help employees at their company work more efficiently and make better decisions. However, 68% say that AI and ML are not accessible at their organisation. Meanwhile, although the pressure to adopt AI and ML typically comes from the top, almost three-quarters of respondents (74%) say that the general know-how around AI and ML is lacking among senior leaders at their organisation. There’s another wrinkle there, since most decision-makers (72%) think their organisation lacks the skills to fully implement AI and ML – but almost half of respondents (47%) say using AI and ML themselves to plug existing skills gaps is one of the three main reasons they are pressured to increase AI adoption or investments. Yet again, partnership is everything A familiar story is emerging as organisations are deploying AI and ML in their various forms: everything hinges on partnerships and the skills and experience they provide. Almost 9 in 10 (87%) of respondents’ organisations are cognisant of AI and ML when making technology purchasing decisions, with 4 in 10 (39%) deeming it an essential factor. But not everything is rosy among those who have already adopted AI and ML. Where AI is currently in use, almost one-quarter (22%) of respondents believe a lack of resources with needed skills is the main reason their organisation’s AI and ML use is falling short of expectations. All of this points to the need for businesses to lean on external partners to address the skills gap – specifically, by deploying the right applications that provide AI and ML as a native part of their core applications platform, making AI and ML a natural part of the user experience. While the average organisation has 57% of AI and ML use cases already embedded in vendorsupplied software, a crucial question remains: who is the right partner for us?
Conclusion Less artificial, more intelligent By now, few people doubt whether to embrace AI; they’re looking for guidance on how to do so. The decisions involved can be bewildering. In this context, it’s easy to forget that each of the three challenges explained above is actually a remarkable opportunity: • Conversations around AI and ML can propel organisations as they rethink their data foundations – improving every aspect of the business. • Getting use cases sorted out and affirming the central roles of your human talent is a necessary and timely strategic gain. • Addressing skills gaps allows you to face an uncertain future better prepared than you’ve ever been. We understand that pressure abounds when it comes to AI and ML. And yes, time is of the essence.
But nobody has to go it alone. The AI IQ research has shown us some of the most important barriers and patterns related to deploying AI within the enterprise. Fortunately, organisations are seeing rapid, tangible ROI from their AI and ML deployments – as long as they have the data foundation, strategic clarity and skills in place to deploy wisely. We wish you an intelligent and successful journey.
Regions: Americas 40% EMEA 45% APJ 15%
Function IT 44% Finance 33% HR 23%
Company Size 500-999 18%
1,000-2,999 22%
3,000-4,999 35%
5,000+ 25% Workday and AI: natural partners In the world of AI, a decade is a long time. Workday has pioneered AI and ML for that long, and the experience pays off. What makes us different? Data: Workday leverages the most reliable data on your people and finances, built on a uniform data model so the data that feeds your AI use cases is always up to date and reliable. With huge amounts of structured, accurate data to train the models, IT is free to focus on other key business transformation initiatives. And, since the quality and value of AI compounds with experience, this data-driven head start is unique and indispensable. Platform: Since AI and ML are built into the core Workday architecture rather than bolted on after the fact, we can scale solutions through all of our customers and deliver AI and ML freely across various use cases. In practice, this means increased agility, quick time to value and proven strategies that put our customers a step ahead. In addition, IT does not need to procure or manage a separate AI and ML stack or data integration. Trust: Because of our responsible approach, we transparently document each AI and ML model and keep people at the centre of everything by ensuring no decision is ever fully controlled by Workday AI and ML technology. Humans are kept in the loop at all the right places and are the final decision-makers
