Navigating the AI transformation

The AI Momentum Assessment — A tool with the five critical factors to get your AI initiative moving

The Dirty Little Secret of AI

The numbers aren’t pretty. 85%. That’s how many of the AI projects fail in practice (Gartner, 2018). At this run rate, it will be hard to realize the enormous potential of AI for businesses.

Recently, we stumbled across a tragically beautiful expression: “Prototyping Hell”. That’s what one manager called the situation at his company. They had built 300 AI prototypes and none of them were live. 300! Whoever budgeted that has some nerve.

What’s going wrong? Why is AI so prone to getting stuck? And is there any hope in the near future? Many opt to wait and see and let the others do their job. That might save a lot of money and nerves. And yet, when it comes to AI, experts and strategists warn: every organization has to go through its learning curves. And those who take the lead are likely to stay in the lead. First-movers will win even more than with other tech trends. Wait-and-see can help or can let you fall behind.

The one main reason for AI problems

When intelligent people fail at the endeavor, it’s bold to reduce it to one cause. With AI, many things must come together for organizations to benefit (more below). And yet, there is a common thread across projects that is evident: naïve expectation and misjudgment of requirements.

“When adopting new technologies,” says innovation expert Geoff Moore, “almost all forces inside and outside the organization work against you.” Moore has been a legend in Silicon Valley and beyond since his best-selling Crossing the Chasm (1991) and Inside the Tornado (1995). His thesis in Zone to Win: only with a clear plan, smart resources and dedication can you introduce new ways of working into established companies. Transformation Zone is his word for it. “It is a huge change management problem,” Moore says. Why?

Companies win through good processes and established management systems. Markets and tech constantly evolves and introduces requirements and changes into their processes. Focusing becomes a challenge. Often the priority list is overcrowded. And now comes AI.

AI introduces a new way of working. New complexity in the tech landscape and data requirements. And different ways to decide and new skill sets. Thus, AI pushes on an already full priority list. If AI doesn’t get enough attention and management support it won’t make it. Many things need to come together for AI to deliver value.

As AI pioneer Andrew Ng says, “Many underestimate how big change management is when introducing this revolutionary technology into the enterprise.” Change in the organization is the big hurdle. How do you tackle that?

Focusing on the bottleneck

The AI cake is baked from the ingredients of data, technology, organization, and psychology. We have developed a tool to break this down and show you the next steps to an AI-powered enterprise. Introducing new approaches is never easy with systems that have been optimized over the years. But that’s exactly the challenge — introducing new ways of working that empower AI in the enterprise. We reduce the complexity to what’s your next step in the next 3 months.

Here is our AI Momentum Assessment. It builds on two ideas. First, it describes your current maturity level for dealing with AI — and what drives that maturity. Second, it spells out your biggest limiting factors. The bottlenecks for AI in your situation.

Let’s look at this in practice. An e-commerce company started its AI journey years ago. Targeted merchandise offering and improved marketing are the business drivers. The company built some pilots and put them in production. In April 2021, we analyzed the setting with the AI Momentum Assessment:

In maturity, the company is on level two (a user). First systems are in production and create value. But the solutions are isolated and new applications require enormous effort. The data landscape is non-consistent and handling is complex. Experts are scarce and located on a few teams. Skills in development, decision-making, and operations are spread thin. But the company could make significant progress in the next 6 months. Let’s look at the details:

In the analysis, culture and the big picture emerge as bottlenecks. The culture is process-oriented and standardized. AI system developers struggle to get support for exploration. Customer feedback is missing throughout the process. This leads to a lot of wasted time, unconvincing solutions, and frustration. The lack of momentum cripples the ability to deliver value with AI.

The big picture suffers from clarity and stability. Management embraces AI but priorities change every 6 months. This undermines the maturation of approaches, building out of infrastructure, and frustrates learning curves. New projects start again with unsolved fundamentals.

With these two priorities, the company has homework cut out. Stabilizing the big picture and priorities will strengthen the AI push on many levels. Shifting the culture to agility and early customer feedback gives the AI solutions more impact. These two issues are not easy, but they are clear. Progress can be made to unlock AI for the company.

What makes AI Mature

In the AI Momentum Assessment, we assess the company’s maturity in working with AI. Three underlying shifts drive the company towards an AI-powered enterprise. Then, AI is no longer an idea or an initiative — AI takes over central value creation in the company and shapes the future.

The three shifts to maturity are: moving AI systems to production, becoming strategic about data, and building skills across the company. Let’s look at those in detail.

The first shift is to move from prototypes to production systems. Many companies start by building first prototypes in select areas. In practice, 90% of the prototypes work but do not make it into production. Production is much more complex. The quality of automation, monitoring, infrastructure, and robustness of pipelines must be higher. The shift into production systems matures the company to running an AI-powered business.

The handling of data makes up the second shift. Many companies have collected data selectively and unstructured. AI is driven by data. Moving from data silos to strategic data acquisition takes the enterprise to a new level. The company analyses its business processes and customer interactions to capture data. It builds systems on how to stream and integrate data. The company tracks its data potential more decisively. This must go hand in hand with ensuring legal requirements and governance.

The third shift: building up AI expertise across the company. Early on, a few experts and teams hold the knowledge and expertise to design and build AI systems. Skills and knowledge across the enterprise become important. Many need to understand how to identify AI use cases and what goes into AI products. Business users need to understand, contribute to and evaluate AI developments. Training and experiences empower employees and transforms the company into an AI-powered enterprise.

How does this look in practice? We look at an e-commerce company in its AI journey. The company started with a central data science team. They worked on initial prototypes to gain experience and show the potential of AI. The transition of the prototypes into production was difficult. Development teams and Data Scientists were in different teams with different agendas. The data landscape was scattered and each new project needed another data connection.

The first step merged the two roles into one team. Data scientists and developers worked closely together to bring AI solutions into production and develop them further. Next, the focus went to the data infrastructure. The company initiated a project to collect data in a data lake and make them easily available. Data quality was initially subordinate and then addressed later. This saved time in the project and avoided data preparation for non-important data. The team then built a central tracking architecture for customer click data. This is a big shift in the second dimension — become strategic in data acquisition. The company started to understand data as capital and proactively build up data capture.

Third comes the shift to empowerment. The company held workshops for developers, product owners, and upper management. The aim was to better understand what AI is and how to apply AI to business problems. The project team identified best practices from the pioneering teams. It developed standard tooling for teams to get a quick start on AI projects. Finally, top management created an owner to coordinate AI developments across the company.

These three shifts enabled the company to put AI at the center of value creation and revenue growth. Company culture plays a role, as does the technology environment and time. Companies starting out today already have access to a variety of AI tools from cloud providers. Every company must find its unique path. But these three shifts help drive more value with AI in any company.

What limits AI in practice

This is a recurring pattern: a company hires a data scientist to work on promising problems. He dives in, gets data, and talks to stakeholders. The first results look promising, only for the work to stall. A few months later he leaves the company for a better perspective. All the time and money invested shows little results. Why is that?

A Data Scientist can’t do much on his own. He needs relevant data of good quality. He needs time to explore, run trials and find patterns. He needs to fit AI systems to the rest of the technical infrastructure. If none of this is in place, frustration quickly grows. The Data Scientist deals with much other than data analysis. Hiring a Data Scientist is only part of an AI drive. This frustration — lost Data Scientists — is well described. Culture and setup are real factors for AI to show its power.

The complexity of AI and the changes needed in the company can be unnerving. So what is the way forward? AI adoption and scale need a prioritized plan to win. Since every company is different, we need a good perspective on how to approach this. The AI Maturity Assessment shows you five factors that need priority over the next months. This focuses your approach to drive AI maturity at your company.

The 5 factors for driving AI at your company are: Skills, Organization, Infrastructure, Change and Culture, and Big Picture. These factors consist partly of soft factors that are often overlooked by experts and technicians. They interact with the hard facts that make up the technology and data landscape. The interaction of the five factors is similar to a pipe — the diameter of the smallest factor limits the flow. The bottleneck limits the potential of AI in the company. This is what the factors mean in detail:

Skills: AI is a new technology and brings an added technical complexity. Data is dynamic and interacts with systems. AI changes internal processes and strategic direction. If skills are missing in the company, AI cannot be properly deployed and evolved. This affects all areas of a company: data scientists and engineering, software developers, product managers, business decision-makers, and the board of directors.

Organization: Another limiting factor is the setup of teams and the relevant experts. Organizations are focused on efficiency and work in focused teams. AI approaches often require data and expertise from different departments. Does an organization know what it needs for AI teams? Can it set up teams quickly and effectively?

Infrastructure: AI requires mastering technical dependencies and effective integration of data sources. In the prototype phase, a few open source tools might be sufficient. For production systems and strategic data acquisition, a central infrastructure (like a data platform) needs to be built out. Often legacy systems are an obstacle. Companies need decisions that enable AI for the long term.

Culture: Perhaps the biggest blind spot is the disregard of culture. Organizations are oriented by Tayloristic efficiency focus. Mistakes are bad and often there is a latent (or overt) culture of fear. AI projects need much more exploration than traditional software projects. This must be reflected in that culture that can deal with uncertainty and time for learning. Otherwise, you will not have the freedom to develop and execute AI projects.

Big picture: The last limiting factor we see is the big picture. AI needs a longer-term focus. Learning curves need to be taken, culture needs to evolve, groundwork needs to be done. Without a big picture, there will be a lack of stability to allow these parallel developments to mature. Also, product development in most areas needs to contribute to data collection. This makes AI possible in the first place.

These five factors interact. The challenge in AI projects lies in evolving these factors in your company. Change is always complex and multi-layered. It takes a transformation of your organization to become AI-powered. Ignoring this fact does not help. We provide language and focus with the AI Momentum Assessment for your journey. This lays out clear and logical steps for your way forward.

This is the way forward

Technical innovation needs supporting management practice. The steam engine needed Frederick Taylor’s management ideas to help entrepreneurs like Henry Ford achieve global success. The software revolution needed agile practices to deliver new products faster and closer to customers. And AI thrives on the right form of management to unleash its market power.

Here we describe the steps to move forward. The AI Momentum Assessment breaks down complex issues into practical focus areas. This makes empowerment for AI feasible. AI means change. Every business can become stronger and more profitable with AI. The best is yet to come. Are you clear on the steps you should take?

Mikio Braun, Jesse Anderson, and Marlin Watling work with companies to make AI more useful for businesses.

Dr. Mikio Braun comes to projects with 15 years as a postdoc in Machine Learning (Uni Bonn, Fraunhofer, and TU Berlin) and management responsibility in AI practice (Zalando, Get Your Guide).

Jesse Anderson developed cloud solutions in the data space for over a decade (Intuit, Cloudera). Now, he mentors companies in their data strategy and organizational design. He is the author of Data Teams.

Marlin Watling worked 15 years as an HR manager at well-known companies (SAP, Roche, etc) and was responsible for over 200 change projects. Now, he helps with AI transformation.

Together they combine the technical as well as the social view on change and are convinced that change for AI is within reach for all companies.

Interested in the AI Momentum Assessment for your business? Talk to us…