Isn’t it a paradox to talk about implementing artificial intelligence technology (AI) with success by enhancing human-led factors? Yes, it may be it in a way. Behind all this, the fear of “being replaced by robots” is like a little thorn in many people’s minds, but we don’t need to worry, at least for now. AI needs us as much as we need it.
Over the last few years, companies in most industries have been recognising the importance of artificial intelligence developments and the impact they are having on business, and more broadly, on society. The urgency to catch up with the commonly named “AI revolution” has forced organisations to invest time and money as well as the best of their talent into AI related projects.
Let’s revisit some statistics. Findings from the “Use of Data Analytics and Artificial Intelligence 2021 Survey” that we published some weeks ago show that the acceptance of AI in Luxembourg companies has grown considerably. By now 52% of them are using AI tools and services, up from 24% in 2019. What’s more, there are another 20% planning on piloting it soon.
This is, by all means, positive news for the country. Luxembourg is moving from spreadsheets to tailored data solutions and the country seems to be well-positioned on a global level when it comes to the adoption of AI tools. However, it’s too early for complacency; the road leading to organisations that take full advantage of AI technologies is still long and, likely, zig-zaggy.
Even if Luxembourg organisations are adding advanced analytics tools to their day-to-day business and acquiring data expertise so they can optimise processes, gain customer insight and, ultimately foster innovation, there is a precept that applies to any transformation exercise on any level: it is challenging and requires nerves, patience and strategic planning.
Yes, when transformation is pursued not all that glitters is gold and that’s fully understandable. It is precisely the challenges that make any transformation trip unique and push organisations to find their own ways and differentiate from the others.
What then, are the human-led factors that are both the AI-enablers and challengers at the same time? Read on. In this article, we delve into some of the most remarkable findings of the Use of Data Analytics and Artificial Intelligence 2021 Survey.
Artificial Intelligence’s preferred menu: data
AI-based technology draws on data collected from different business fronts. For it to work, to give insights and enhance decision making, it needs data. Voilà, the reason why the survey we conducted focuses on both. Here is a useful chart to understand organisations’ main reasons for data collection.
As you can observe, just as in 2019, operational efficiency and client knowledge and experience remain the top reasons for data collection in 2021. However, the data collection for the purpose of leading innovations has more than doubled in 2021.
An interesting fact is that the more the use of AI technologies becomes an organisational priority, the more digital and/or data teams step in. Only a few years ago, piloting AI projects was rather a R&D and IT teams’ matter but the chart below shows that this fact is changing:
The two biggest hurdles when Luxembourg organisations deploy AI
Human-led factors or, put more accurately, factors where humans are pivotal, represent the biggest challenge in the deployment of AI.
1. Data and AI talent
Let’s tackle the first challenge, talent (or the lack of it). Actually, to a significant 58% of the survey respondents the lack of talent remains top in the ranking, up from 37% in 2019. That’s more than a 20% increase.
We understand there are two underlying reasons. One is more structural, namely, it’s linked to the lack of training or a scarce training offering in both market and Academia. As a result, there is a limited availability of graduates and experts in fields relevant to AI although it’s increasing. To this, one needs to consider that, in general, the knowledge of AI among executives still falls short.
The second reason is the increasing number of competencies that Data and AI professionals have to respond to. This has to do with the way AI solutions best work, fusing themselves with analytics, the Internet of Things (IoT) and other enterprise systems. This requires, clearly, specialised roles, resources and processes to keep all technologies integrated and properly running.
On the positive side of things, we notice there is an increased specialisation in the data domain, with growing roles in data science—a leap frog jump from 14% in 2019 to over 52% in 2021— data architecture, data management and data governance.
2. Bringing AI into operation
We cannot deny the fact that AI has matured in Luxembourg since 2019. While our study’s findings cannot tell, conclusively, if the COVID-19 pandemic is a detractor or a driver of this trend, the numbers indicate a move towards the operational use of AI.
And that’s precisely the second challenge that Luxembourg organisations face, putting AI into operation and embracing the right strategy. Only 35% of companies have a mature Data and AI strategy, a small increase compared to the 29% of them in 2019.
AI isn’t a technology that likes to remain single, it isn’t a solitary wolf, if you allow us that analogy. As we mentioned above, AI doesn’t do its best work when it’s isolated from other technologies, or when it’s siloed in a lone function or business line. For a full integration, it needs to work in tandem with broader automation initiatives, data analytics or both. Then, the need for strategising AI and defining a suitable operating model become even more necessary.
Here are some of the most conspicuous and relevant AI applications that organisations across all industries are currently using:
- Calculating and managing risk
- Detecting and preventing fraud and cybersecurity threats
- Improving AI ethics, explainability and bias detection
- Helping employees make better decisions by means of decision support systems
- Analysing scenarios using simulation modeling commonly powered by machine learning
- Automating routine tasks
- Content management using solutions for image and speech recognition and text generation and analysis.
Three tips to bridge the AI talent gap
It comes with little surprise that the lack of skilled people and the difficulty hiring the scarce talent topped the list of challenges when Luxembourg organisations deploy AI. And countless articles on the web, give account of this. Here are three tips that can help your organisation go about it.
1. Collaboration private sector – academia
While higher education enrollment in AI-relevant fields like computer science has risen rapidly in recent years, a growing student demand isn’t yet covered. Universities lack expert staff in the field too. And that has a cascading effect that ultimately impacts the world of work. The still scarce talent around the fields of data science, Analytics and AI could be further strengthened if companies and academia work hand in hand.
Have you thought about reaching out to higher education institutions in your country or region?
Here is an interesting observation from the European Digital SME Alliance: “Europe is a global leader in the research of (ethical) Artificial Intelligence (AI), but it lags behind the USA and China in its industrial applications” and acknowledges the need for the development of collaborative intelligence systems and connecting research with industrial applications.
2. Create new roles for new collaborations
In our survey, 45% of respondents consider their company to be mature in terms of data architecture, and almost 38% have hired a data architect compared to 29% in 2019. That’s a positive sign. Nevertheless, AI deployment also calls for integrating AI architects.
They may be traditional solution architects with solid data science expertise, or data scientists with a background in software engineering. Either way, they are often the ones best placed to drive the efficient implementation of AI use cases, helping to best allocate scarce resources.
Once a potential use case has successfully passed through the proof-of-concept and pilot-and-proof stages, it’s time to make it an actual and valuable AI solution: an asset that scales, is fault-tolerant, runs on your chosen platform, and can meet deployment needs. For this step, two additional roles come into play.
The first role is a machine learning engineer, with both data science and software engineering skills. The second is MLOps specialist, to manage post-deployment model performance. Together, these two roles supervise and integrate data, AI models, and supporting software throughout the AI life cycle.
These specialists also help drive the scientific, experimental mindset that AI needs: one in which hypotheses are continually challenged and models are continually improved.
3. Upskill but go beyond upskilling too
Unless you’re running an AI-only startup (and maybe even then), your workforce needs it. But the old kind of upskilling—offering learning opportunities focused on a siloed technology—is not enough to get your employees or your company ready for AI at scale.
True upskilling requires more than offering training courses. Organisations also need to give immediate opportunities and incentives for people to apply what they’ve learnt, so that knowledge turns into real-world skills that improve performance. Hackathons involving interdisciplinary teams, for instance, are very helpful to start putting AI skills into practice.
Organisations also need cross-skilling: giving specialists in one area (such as data science) enough basic skills in another (such as the business) so they can speak each other’s language. Such cross-skilling is critical not just for collaborating on AI-related challenges, but also for deciding which problems AI can solve.
While the World Economic Forum (WEF) has predicted that AI will lead to long-term job growth, these jobs will be different from the ones that have existed in the past. Business leaders need to reevaluate exactly what they’ll need from this future workforce.
Five tips to bringing AI into operation
Because of its newness, most organisations have been implementing AI technologies in pilot mode, experimenting or launching products in beta. To put it bluntly, operationalising AI is moving from that pilot stage to a business-scale deployment so as to solve real business needs.
That can be troublesome or even intimidating as not many organisations or professionals can claim extensive expertise in the matter. We’re all learning how to do it.
Here are five tips to operationalising AI that we trust are helpful for any organisation:
1. Choose the right operating model
Align AI operations with those for automation, data and analytics. If you already have centralised administration for these other technologies, then you also create an AI center of excellence or appoint an AI leader.
If an existing data analytics or automation team is mature enough, encourage them to “add on” AI to it.
In case your organisation is largely federated, choose to delegate AI strategy and governance to each line of business. The critical factor is choosing a model that will work well with the automation, analytics and IT teams that AI will both support and depend on.
2. Embed AI into your overall IT stack
For instance, incorporate AI models that are responsible for automation or key decisions or integrate trained AI models into production applications to scale up use.
Ideally, this embedding of AI into the IT stack should also support a common AI services layer for any application to integrate with AI models.
3. Choose the right technology — and architecture as well
There isn’t a standard interface yet to integrate technology tools for AI.
To choose the right one—whether you build or buy it, go to the cloud or keep it all on premise — focus on these two imperatives: integration and data.
Some technology platforms, for example, ingest data from internal and third-party sources that range from web services to PDFs, standardise that data and help verify its accuracy and regulatory compliance. This can’t entirely eliminate the need for your experts to label data for AI, but it can help. In any case, there is also a growing group of data labeling companies at your disposal.
4. Develop machine learning operations
The key to making AI part of daily operations is a new capability, MLOps (or ML Ops) that deploy, maintain and monitor machine learning models in production in a reliable and efficient manner.
The use of MLOps requires combined expertise in data science, software engineering and IT operations. For an effective MLOps function, most companies will need to hire and upskill talent.
5. Make your data trustworthy
To make AI operational at scale, it needs data which is not just accurate, but standardised, labeled, complete, free (as much as possible) of bias, compliant with regulations and secure. Only then can you trust your data—and the results of AI models based on it.
What we think
Luxembourg is an exciting place for AI. On the one hand, as our study reveals, investments in the field are increasing and there is a growing start-up culture. On the other, and very importantly, industries that are key to Luxembourg strongly believe in the benefits that AI can bring, and are implementing it more and more. While challenges remain, our respondents expressed remarkable optimism about the technology for both the society as whole and the future of their business.