2021 AI Trends & Predictions
TL;DR 2021 will be the year of widespread business adoption of AI. This time SMEs with progressive CEOs will reap most of the benefits.
My 5 AI Trends & Predictions for 2021:
AI will augment traditional systems such as ERP, decision support/BI or even call centre management systems.
Cloud platforms will dominate opensource frameworks. The market share of 100 % proprietary AI solutions will shrink.
AI solutions will differentiate by providing industry-specific business content, pre-trained models & exclusive data access.
High fail rates of AI transformation programs will continue to frustrate the corporate C-Suite.
Lower cost and oCDOs and oCIOs will play a key role in making AI accessible to SMEs.
1. AI will augment traditional systems such as ERP, decision support or even call centre management systems.
Specialized Artificial Intelligence systems & tools go beyond just Machine Learning. They consist of a varied set of building blocks. They help to communicate (Natural Language Processing), to understand images & video (Machine Vision), to execute processes (Robotic Process Automation) and to make decisions (Predictive & Prescriptive Analytics). Each of those building blocks requires different specialised skills, technologies and frameworks.
Before 2020 we have seen much effort going into integrating these blocks in ever more complex frameworks. These efforts continue, but in 2020 I have observed much more focus going into evolving specialised tools that are good at one job only.
In 2021 we will see this triggering an avalanche of AI-augmented systems. We will experience ERPs, Internet of Things (IoT), Communication Systems, and even Blockchain solutions that are augmented by AI capabilities.
One of my focus areas will be to use AI to help our clients to automate business to consumer communication and request fulfilment. Having seen this allowing one AI-augmented agent doing the work of > 100 regular agents has been a very inspiring experience for me.
2. Cloud platforms will dominate opensource frameworks. The market share of 100 % proprietary AI solutions will shrink.
The big cloud infrastructure providers such as AWS, Google, Microsoft, Oracle, IBM, etc. have heavily invested in providing specialised AI tools & frameworks. At the same time, we are seeing, even the heavily-regulated financial services industry massively speeding up their migration into the cloud. This trend is, of course, chiefly triggered by progressive regulators getting more open to cloud-based innovation.
Scalability, flexibility, lower investment cost, simplified data sharing and solid cybersecurity are strong pulls toward Platform as a Service (PaaS) solutions. Opensource frameworks that can not provide those benefits risk becoming academic niche projects.
As customers become captive of the cloud they chose, it will become more and more difficult for independent proprietary AI solution providers to position their products in 2021 and beyond.
3. AI solutions will differentiate by providing industry-specific business content, pre-trained models & exclusive data access.
I firmly believe that as leaders of AI businesses, we have to embrace this trend. Why should we compete with Amazon or Google for the best natural voice recognition engine? The next round in this contest is won by the best data and not the best algorithm. Echo speakers are collecting natural language data in 53.9 million households for Amazon every day. Replicating such a data collection engine is very capital intensive. The opportunity to differentiate for smaller players lies in making sense of the recognised input, combining it with industry-specific data sets and turning it into action.
At ReGov Technologies, our strategy has always been to re-use existing world-class components and develop the proprietary capability needed to achieve specific business outcomes for our clients.
I see similar strategies being adopted by AI startups and mid-size tech companies at scale in 2021. Smaller players will more and more differentiate by offering specialised business content, pre-trained data models and exclusive high-value data sets.
4. High fail rates of AI transformation programs will continue to frustrate the corporate C-Suite.
Research data suggests that in 2020 corporate executives entered the trough of disillusionment of the AI hype cycle.
According to the Harvard Business Review, 77% of executives report that the adoption of AI is a significant challenge, with 83% of respondents blaming people and process issues as an obstacle. PWC reported that only 4% of executives surveyed planned to deploy AI enterprise-wide in 2020 vs 20% in 2019.
This data is entirely consistent with anecdotal data obtained from my many conversations with AI practitioners and clients.
Several high-level studies refer to lack of C-level aspiration, incomplete visions, lack of organisational engagement and the inability of technical leaders to obtain sufficient buy-in and investment as critical reasons for past failures.
I think it is to easy to blame the classical "top reasons" that make transformation projects fail. I see two key reasons why big companies fail their AI transformation projects:
Lack of quality data. AI as any computer system follows the good old "garbage in, garbage out"-principle. I saw many companies start AI or analytics projects without first addressing the fact that they had no mature master data management practice in place. Equally that the data needed by the AI program was in different silos managed by, sometimes competing, organisations was ignored. External data acquisition also requires a solid, aligned strategy instead of departmentalised action.
It was a big AI transformation project. Unless the CEO makes it a personal priority to push the program through, a big program is destined to fail for all the reasons perfectly outlined by HRB & PWC. Small projects with empowered, dedicated and highly skilled teams have a much better chance to succeed and become the seed of a more significant iterative transformation.
Issue number two is easy enough to tackle. The predicament is that the small team will fail as well if issue number one is not addressed. And here comes the real problem: Data quality programs are some of the most challenging efforts a corporate manager can endeavour on. They require highly experienced (=in demand) cross-functional staff, take long, require significant operational process changes and are political minefields for the leaders in charge.
How to resolve this predicament goes beyond the scope of this article. I hope it illustrates, though, why it is difficult for several large organisations to adopt AI effectively and efficiently.
My prediction for 2021 is that only a few organisations will leave the trough of disillusionment and take decisive action to tackle the root of the issue. The success of those few will force their competitors to take action leading to potential mass adoption of AI and good data management practices within the next five years.
5. Lower cost and oCDOs and oCIOs will play a key role in making AI accessible to SMEs.
The AI adoption challenges faced by SMEs are very different from large corporates. They often lack the technical expertise and thought leadership to adopt AI successfully. In many cases, there is also no capital available to invest in expensive AI teams and experienced technical executives that could put them to use effectively.
A trend I observed closely over the last two years is the emergence of outsourced Chief Data and Chief Information Officers. Experienced consultants are offering their guidance for decades already, of course. What is different about an oCxO is that he/she takes accountability for executing on the strategy as well. Further, the most effective oCxOs come with their own existing technical teams. It is obvious how this can speed up moving from strategy to implementation. Another advantage, executives often overlook, is access to better talent. The best AI experts want to work in teams where they are surrounded by peers to learn from. They also prefer to work with a leader that understands the work they do at least on a fundamental level.
The key concern about outsourcing AI and other technical work in this way, I hear most of the time, is "lack of trust". Indeed trust is critical. Clients and friends that used this model successfully aligned goals by using revenue sharing or equity participation to align incentives. This focuses your oCxO on optimising his team's actions to grow your business. The ability of SMEs to act decisively, having fewer legacy data structures, combined with improved access to world-class AI talent, makes me predict that in 2021 they will be among the big beneficiaries of AI adoption. If... their CEO's take action... About the author
Simon Ulrich has been in the IT industry for over 18 years, with his primary focus on business intelligence, data science, artificial intelligence and enterprise blockchain. He was the Global Head of Business Intelligence, Data Science & Master Data Management of Roche from 2012 till 2018. Simon lead & built big corporate IT teams in 8 different countries on three continents. He learned a lot from starting his own HealthTech / InsureTech business and is now the CEO of ReGov Technologies, an innovator in Artificial Intelligence and Enterprise Blockchain.