Jan 8 — 2021

Trends in Data Science & AI that will shape corporate strategy and business plans

data science trends 2021


In response to the most atypical year that many of us may have probably lived, leading companies have wisely reconsidered where to place their money. Inevitably, the pandemic has noticeably sped up the adoption of Artificial Intelligence (AI) and has also motivated business leaders to accelerate innovation in the pursuit of new routes to generate revenue and with the aim of outrunning the competition.


Companies and clients entering a digital transformation and wanting to become data-driven can be overwhelmed by the sheer amount of technological solutions, tools and providers. Keeping up to date with the trends in the discipline may help. So without further ado, let’s take a look at what to watch for in the year ahead with regard to practices, talent, culture, methodologies, and ethics involving business strategy and organisational development.


From Data to AI literacy

The structural foundations of a company are its people, so ensuring that your workforce is ready to embrace change, because they understand what change means, is still the first step to take in the AI journey. Data literacy was a trend in late 2019 and throughout 2020, now together with AI literacy both will continue to be a trend for the next two years.

If you think about it, Computer literacy is now a commodity that allows us to engage with society, i.e. finding a job, ordering food, etc, in ways previously unimaginable. Similarly, AI literacy is becoming increasingly important as AI systems become more integrated into our daily lives and even into our personal devices. Learning how to interact with AI-based systems that are fuelled with data provides more options in the consumption and use of technology and will pave the way for successful AI adoption in the corporate world.

In order to have AI literacy we must first have data literacy. To understand what an AI algorithm does, you must firstly know which data you have at hand, understand its meaning and where it’s sourced and how it is produced, and only then can you know how to extract its value using AI. Therefore, investments in educational workshops, seminars and similar exercises for preparing all business units to understand how Data Science and AI will impact their lives and work is instrumental, especially because of the huge amount of preconceived notions that are floating about out there.

From abstract to actionable. CDO´s and human-centered methodologies.

AI has been on everyone’s minds as the next major step for humanity, but companies that have previously struggled to find a measurable return on their investment will now be taking a more pragmatic approach , making special adjustments right from the start. These adjustments may be structuring and reshaping the whole organisation so that AI is welcomed and supported appropriately and then correctly embedded across all business units.  One of the advantages of practical AI is that it can achieve ROI in real time, so this could be the year many organisations see their AI efforts begin to really pay off.

So far, statistics have shown that the vast majority of data projects never get completed or just fail to deliver, so the use of human-centered and Design Thinking approaches are relevant when trying to put strategy and ideas into action and to really know where to start and create impact.

Data scientists and engineers that have been working in silos will now be placed transversely across the whole organisation. We will see an empowerment of the Chief Data Officer role and its vertical to support initiatives that affect all organisational layers and this will ensure that any Machine Learning or Robotic Process Automation projects are aligned with the overall business strategy, creating a quick quantitative and qualitative improvement of the existing  operations. In turn, data professionals will need to quickly adapt by developing their soft skills, such as communication and business acumen, otherwise there is a good chance the clash between data professionals and business executives will continue, ultimately resulting in AI investments not paying off.

Growing lack of specialised talent

It is now difficult for a company to attract talent in this field, the demand for data professionals and AI specialists vastly exceeds the academic supply and the majority of companies currently lack the technical talent required to build scalable AI solutions. As a result salaries are going up which in turn leads to the majority of companies not being able to afford to build an internal data science team.

On the other hand, online learning portals have been providing courses and certifications that allow professionals to get up to speed, but those courses alone don’t teach everything that you need for the job; they are only a complement to other forms of training and hands-on projects. Therefore, senior specialists will still be a scarce resource and the current situation will not improve, but get worse.

One of the viable solutions to overcome this hurdle may be to provide access to self-service platforms i.e. automated machine learning tools as a way to optimise all processes that currently require highly specialised roles. This takes us to the next trend.

Self-service Data Science and AI

Taking into account the rising demand for data professionals, organisations who are unable to hire are facing the risk of being left behind.  As a consequence, a growing number of companies are turning to no-code or AutoML platforms that would help to assist throughout the complete data science workflow, from the raw dataset preparation to the deployment of the machine learning model. The underlying objective of these “self-service business models” is to harness the commercial opportunity that the growing lack of talent scenario presents.

With the rise of no-code or low-code AI platforms, we could start to wonder: Will the job of a data scientist, data engineer or data analyst disappear or will it just evolve? In my view, although a growing number of tools rightfully promise to make the field of data science more accessible, the average Joe will not be able to make the most out of these tools and projects will be very technically limited. There are solutions providing users with attractive user interfaces and a lot of prebuilt components that can ease Machine Learning developments, but I would dare to say that we are still 4 to 5 years away from a massive democratisation of the Data Science practice ( BI took more than 15 years as a reference) I am, however,  pretty convinced that in 2021 we will see many more self-service solutions both offered and implemented.

Automated Data Preprocessing

Inherently related to the previous trend, many Data Scientists have historically agreed that one of the most tedious and complex steps is preparing data sets to be analysed or used in order to develop, train and validate models.

Feature engineering for a Machine Learning algorithm can be more entertaining, Dimensionality Reduction in the form of Principal Component Analysis can be challenging but transforming and cleansing data sets is overwhelming and is certainly time consuming. New Python libraries and packages are emerging where the preprocessing step is automated, saving up to 80% of the time that is currently spent on early project stages. The trade-off, or even drawback, may be that data scientists would be unaware of how the resulting data sets’ features were transformed and some specific pieces of knowledge could be lost along the way.

Nevertheless, even if the preprocessing of data is automated, some data engineering tasks will still be performed manually such as moving data from different silos to unified data warehouses. This task may be the most time consuming part of the process; and it will be very hard to automate because it is case-specific.

IT ( CIOs) vouching hard for AI

In 2021 we would expect organisations to start to see the benefits of executing their AI and ML models at a global scale, not only getting them into production for some “local” or specific use cases, but also pushing them horizontally. IT will not and cannot continue to be a “bureaucracy and technical requirements gateway” for AI projects and in 2021 IT will have to keep evolving to be an innovation hub and a “visionary instrument” for businesses. CIOs around the globe will push for AI to be embedded across the whole organisation and the spin-off of the CDO office from the traditional CIO vertical will now speak volumes about a companies´ digital maturity and about the progress of their digital transformation journey.

Explainable and Ethical Data Science and AI

The more data science and analytics are central to the business, the more data we retrieve and merge, the higher the risk of violating customer privacy. Back in 2018 the big shift was GDPR, followed by browsers taking a stand on data privacy and now Google is planning to deprecate third-party cookies by 2022. In 2021 the ethics and operational standards behind analytical and predictive models will come into focus so that any AI mechanisms are clear from biases.

On one hand, algorithm fairness and on the other, the transparency and quality of the data sets that are used to train and validate these algorithms are two of the issues in the spotlight while companies will not be able to afford “black boxes” anymore. Leaders will be proactively managing data privacy, security and ethical use of data and analytics. It’s not only the right thing to do and an increasing legal requirement, but an essential practice to gain trust and credibility both when used inhouse to make decisions, and when used to outsmart competitors.

Augmented Intelligence going mainstream

The term data-driven has been in the mouths of many, but in reality only a small percentage have put it into practice. The maturity of the data technology and the expertise of data professionals will now enable the decision making process, at any level in the company, to be almost fully automated and data-driven. Note the word “almost”, as output from models can complement human thinking, but not completely overrule it as analytics are not perfect either i.e. variables could be missing or data may be biased from the source.

Augmented Intelligence, also known as Machine Augmented Intelligence, Intelligence Amplification and Cognitive Augmentation, has referred, since 1950, to the effective use of IT mechanisms for augmenting human intelligence. Corporations will now be building up their Augmented Intelligence capabilities where human thinking, emotions and subjectivity are all combined and strengthened with AI´s ability to process huge amounts of data, allowing these Corporations to make informed decisions and to plan and forecast accurately.

Again, this does not mean that the AI algorithm will dictate and tell C-level executives how to run their business, but will certainly provide them with the best guidance available by providing possible outputs with the data that is fed to it. 

Affordable modelling of unstructured datasets

Natural Language Processing, Computer Vision and other forms of unstructured data processing are being improved day by day. In addition to the effort of hundreds and maybe thousands of experts  in refining these AI models,  the increase in remote working will drive the greater adoption of technologies that embed NLP, Automated Speech Recognition (ASR) and other Capabilities of the like. Computing platforms have made it possible with tools like Google Natural Language API for every company to be able to use a Deep Learning NLP without needing to train the model locally. This affordable modelling could also coin the name of AI as a Service soon. The advancements in this field will allow small and medium businesses to process data in unstructured formats because of the accessibility to validated algorithms paired with affordable cloud processing power.

Data Storytelling and Artistic Data Viz go mainstream hand-in-hand

Any form of advanced analytics, whether it is descriptive, predictive or prescriptive does not make a lasting impact and cannot reach its full potential if insights are not communicated properly. Representing data in appealing visual ways while surrounding the numerical findings with the proper narrative and storytelling elements is now the recipe for success in the data science realm. The algorithm selection and the data set where the model was trained is undoubtedly critical, but presenting the findings and conclusions of “why something happened”, “how it could have happened” and “what could we have done about it” have never been as important as today, due to the complexity of the analytics beneath the surface.


Companies and organisations rely on data to drive their innovation agenda. However, business leaders still face significant challenges to make the most out of their investment in an immature data-driven culture.  Data as an asset, Data Science as a tool and Artificial Intelligence as a discipline will encompass the next revolution for humans and we are lucky to be present.

Recapping, we should expect 10 main trends in Data Science, Data Analytics and the Artificial Intelligence space in 2021:

  1. Data Science and AI literacy will continue to be a trend because humans remain at the centre.
  2. Artificial Intelligence moves from abstract to actionable and the CDO role will gain importance.
  3. The lack of specialised talent will not cease to grow.
  4. Self-service solutions. Many autoML solutions will thrive during the next few years, empowering non-technical users to be rookie data scientists. The self-service option will contribute to widespread adoption, but AI and Data Science consultants will still be critical to drive these initiatives both on a strategic and hands-on level.
  5. Automated Preprocessing of data could soon be feasible, allowing Data Scientists and Data engineers to focus on what really adds value to the business.
  6. IT is pushing for AI harder than ever before.
  7. Explainable, transparent and ethical data management will be on top of all agendas. The value derived from a predictive analytics project will not justify the means to an end.
  8. Augmented Intelligence will allow companies to outrun competition.
  9. Affordable modelling of unstructured datasets will result in a massive adoption of cutting-edge AI solutions.
  10. Data Storytelling and Dataviz will not be the icing on the cake, but the key ingredient in the data science recipe

The next couple of years will surely show a shift in AI from being an emerging technology to a widespread adoption.