Gone are the days when organizations could only afford an experimental approach to artificial intelligence and broadly understood analysis. Now it's time to accelerate the application of AI.
Businesses can no longer assume that managers will experiment and see what will come out of it. The time for experimenting with AI is over, mainly because artificial intelligence is already a foundation of substantial competitive advantage for companies—those who are still experimenting need to catch up quickly.
The experiments have already spawned implementations with tangible returns on investment. This situation has another consequence. It is no longer feasible to connect the existing systems in the organization, glue them together and build an intelligent environment through integration. Mainly because not artificial intelligence native software is outdated and doesn't support the free flow of data between departments.
In the early days of AI in business, the business benefits of this technology were not obvious, so organizations hired data analysts to investigate what was possible. At the same time, without focusing on creating a stable environment for developing artificial intelligence that could operate reliably 24 hours a day. It was more a play, experiment than an adequately planned roadmap.
Assuming that we are doing something experimental, companies did not focus on building a system that would be scalable. Data analysts and software engineers often invested in local solutions installed on servers that a small group of colleagues knew about. Technology architects used various tools for this exercise without considering how the software would assist the organization in the future.
Thus, being in the experimental mode, engineers and analysts left aside many tasks supporting the scaling of solutions, such as building a critical infrastructure on which all AI models can be reliably developed and efficiently run and distributed within the organization. Artificial intelligence adoption was slow, ineffective, and considered a go-to Mars kind of thing. Accelerating the application of AI was not the main thing yet.
Nowadays, market forces and consumer expectations leave no room for such ineffectiveness. Organizations that recognize the value of AI have quickly shifted the course from researching what technology can do to using it on a large scale and maximizing the value. The tech giants using this technology continue to make changes and gain market share in traditional industries. Artificial intelligence adoption became the mainstream, followed by substantial capital investments. At the exact time, consumer expectations for personalized, seamless experiences continue to rise.
Fortunately, as artificial intelligence adoption accelerated, a business world created processes and standards to ensure AI success on a large scale. Specialized roles such as data engineer and machine learning engineer have emerged, offering the skills necessary to achieve that scale.
A rapidly evolving set of technologies and services has enabled teams to move from a manual and development-focused approach to a more automated, modular, and tailored AI lifecycle, from managing incoming data to monitoring and repairing running applications. As a result, value is delivered to the end customer faster.
A good example of the application of AI is the GPT-3 engine offered by the OpenAI company founded by Elon Musk. The GPT-3 can write and summarize texts at an insane pace. Copy.ai uses this ready-made element as part of its platform for marketers. More than 300,000 marketers from companies such as eBay, Nestle, and Ogilvy are using the platform to develop text materials for their marketing and PR activities.
Statista prepared a study in which analysts tested whether consumers know and understand where artificial intelligence adoption happens. Researchers compared the results extracted from the group of consumers with the results from a group of businesspeople. It turns out that consumers understand in which cases artificial intelligence is used, but also do not differ from business people in this understanding.
Figure 1: Do people understand AI application, source: Statista
Organizations should invest in many types of reusable assets and components. One example is creating ready-to-use "products" that standardize a specific set of data (for example, combining all customer data to create a 360-degree customer image), using common standards, built-in security and surveillance, and self-service capabilities.
Weights and Biases is an example of a company that allows for a significant acceleration of the transition process from the experimental phase to market implementation. The company's solution allows speedy testing of artificial intelligence models, their versioning, and version tracking.
If your organization has customer data, Weights and Biases will allow you to quickly build AI models and significantly increase the speed of application of AI. These, in turn, will let you predict behavior and examine the customer churn and at what price level the customer will consider swapping you for the competition offering.
Thanks to this approach, teams can use data much faster and easier in many current and future use cases, which is especially important during the scaling application of AI in a specific domain.
A good example is Alteryx—another company that helps to accelerate artificial intelligence adoption. Alteryx enables analysts to prepare, aggregate, and analyze data faster without using classical programming. With the so-called Low-code (no need to write classic lines of code), data and software engineers can quickly build predictive models. A step-by-step or fully automated process allows you to create appropriately trained algorithms that are ready for implementation and fully scalable.
Organizations often invest a lot of time and money in developing AI solutions only to find out that a company stops using nearly 80 percent of them because they no longer provide value. Using ready-to-go solutions that provide scalable tools in the service model means that only 30% of AI models are wasted. How is this possible? Specialized platforms deliver faster and more effective conclusions from the processed data. Thanks to that, the process of designing AI-powered digital products is prompter and more cost-effective.
Implementing an agile approach requires significant cultural changes to loosen the firmly held belief that engineers should only develop software internally. Only then is the environment safe, and the control is tight.
The change happens in moving away from experimenting internally to creating products that leverage ready-to-go solutions and open source technologies.
Creating new capabilities will significantly transform the way analysts, software engineers, and data technologists work as they move from on-demand development to a more commoditized and standardized delivery and functionality process. When engineers bet on commoditization, artificial intelligence application accelerates.
But what should company directors and managers do to accelerate the adoption? Here are the three elements I recommend considering:
Building a culture of efficiency
Every digital transformation is based on specific foundations set by the company's management. Suppose managers want to speed up artificial intelligence applications and adoption. In that case, it is essential to communicate that product development does not have to be based solely on internal resources and systems. Using ready-to-go, third-party, scalable components and mingling them with internal systems should be allowed. Executives should simplify purchase and monitoring procedures and ensure access to payment for services is easy and free of unnecessary formalities.
Managers should make it clear that AI systems are as important to companies as ERP systems and that they should run 24/7 and generate business value daily. While setting a vision is crucial, it pays to communicate precisely how to conduct the execution process. Artificial intelligence adoption doesn't happen out of the blue. It requires guidelines and organizational efficiency.
What is worth communicating in an organization that wants and accelerate artificial intelligence adoption:
It may take 12 to 24 months for these goals to be fully achieved, but using an agile approach has a real chance of reducing the total to two quarters.
Building smooth cooperation between business and IT to accelerate the application of AI
One of the essential elements that influence the acceleration of AI projects is the alignment of the goals of business leaders with the goals of artificial intelligence teams and IT teams. Ideally, most of the goals of AI and data teams should serve those of business leaders. Conversely, business leaders should determine what value they expect from AI and how that value will monetize in the marketplace.
At one of my clients, I found a situation where the goals of the IT department and business units were not so much divergent as not integrated. The IT department pursued its operation strategy, and at the same time, the business units focused on their northern stars, which drove them towards the achievement of operational goals.
As a result, over 80% of 50 business units did not integrate their expectations with the IT department's business and technology roadmap. As a result, this led to the creation of over 200 projects. Business leaders did not leverage the potential offered by the technologies already used by the IT department. In this situation, the BITA model comes in handy.
Figure 2: Business and IT alignment model, Henderson, and Venkatraman (ResearchGate).
Business Alignment with IT is a strategy that prioritizes integrating IT operations and business objectives to lower costs, improve flexibility and increase ROI. In our case, it helps to increase the artificial intelligence adoption level and accelerate the application of artificial intelligence.
An important measure is the level of collaboration around strategic technology investments to deliver tools, technologies, and platforms that optimize workflows in tech projects.
Due to the rapid pace of technological change, IT often finds it challenging to balance the demand for new tools and technologies with concerns that short-term fixes increase the cost of the technology in the long run.
A good example is an old question. Should I improve the "old" CRM system or buy a new one, available in the Software As A Service model.
It is an excellent practice to develop criteria that will allow managers to build a technology development map to reduce complexity when this decision is on the table. So the decision is not whether to develop the old or buy new, but is to answer the question with whom to build a partnership so that the development is the fastest and the least complex possible.
It's a good practice to encourage AI leaders to build strong relationships with their IT counterparts, vendors, and field project leaders.
Investments in talent
The roles of data analysts and technical engineers are changing. Previously, their work was mainly focused on low-level coding. Nowadays, they need to assemble models from modular components and create products ready for production and scaling. The application of artificial intelligence requires a different approach.
There are also newer roles needed in AI teams. One is a machine learning engineer prepared to transform AI models into enterprise-class production systems that operate reliably at scale. Business leaders should communicate that change within the organization and coordinate the talent development map with HR managers.
Leave your name and e-mail below, to get my two ebooks completely for free: