In this new series, Jan Bosch discusses the typical challenges and the typical evolution of companies, product development processes and products when adopting AI.
With new breakthrough models being announced all the time and hundreds of billions invested, the hype around artificial intelligence seems to keep going much longer than any other major technology I’ve experienced in my lifetime. The promise (and to some extent the threat) of AI is that it will provide new levels of automation and productivity improvements that go far beyond the traditional, algorithmic approaches.
That will, of course, mean that many jobs will disappear or employ much fewer people. Many are concerned about the implications for the current working population and their future. Although I can see that there will be significant changes in the short term, I’m much less concerned about the long-term impact of AI, as humanity has always found new ways to employ people in jobs that can’t be automated or that we don’t want to automate for ethical or social reasons.
If we look back to all earlier technology shifts or revolutions, including the adoption of agriculture, the enlightenment, the industrial revolution, the computer revolution and so on, I think it’s fair to say that each of them resulted in humanity being much better off than before. It’s very easy to romanticize the ‘good old days,’ but especially before the 20th century, many lived in abject poverty, died young from diseases that don’t exist anymore or can be easily treated and lacked personal freedoms that we today consider normal and part of our human rights.
For all the promise of AI, though, I experience quite a bit of fear, uncertainty and doubt in the people and companies I work with. Statements such as “we should adopt AI” and “we have to adopt AI” are typical, rather than embracing the promise of the technology. And even the people open to adopting AI in their company often struggle with operationalizing this ambition and executing on it.
In this series, we’re going to deep-dive into the notion of an AI-driven company. Over the last months, together with a colleague, I’ve conducted an interview-based study on AI adoption in European companies. We’ve talked to tens of people from more than a dozen companies, and some patterns are starting to emerge that I believe might be interesting to share.
Specifically, in this series, we’ll explore four areas. First, we discuss the challenges that companies typically experience as they’re adopting AI and specifically agentic AI in their business. These include challenges such as identifying clear use cases, organizing the data necessary for agents, agreeing on clear, quantitative outcomes, establishing the suitable level of human oversight and control, integrating agents in existing workflows, ensuring compliance with regulation, technical challenges and, not the least, organizational and cultural resistance in the organization.
In the second part, we present a five-step maturity model for how companies typically evolve in their journey of adopting AI. We see companies start with the ‘playtime’ phase and bit-by-bit integrate AI in process steps, reinvent business processes from an AI-first perspective, use agents to connect business processes that earlier weren’t connected for optimization purposes and, finally, move to an AI-driven company where every aspect is designed in an AI-first fashion.
In the third part, we discuss the product development process and how it evolves as we adopt agentic AI. We see that companies start with the AI assistant where individual developers use agents to automate or support some of their tasks. The next steps are concerned with AI agents increasingly taking over many tasks until all we provide is the intent of the system and the rest is generated by a team of agents.
The fourth part is concerned with the product itself, where we describe our findings in a similar, evolving maturity style. We see that companies start with static ML models, then add dynamic models with automated retraining, move to multiple models used in sequence or in parallel, adopt reinforcement learning where the system experiments with its own behavior and, finally, arrive at AI-first products where the system is continuously monitoring and generating itself at run-time based on its performance.
For all the promise of AI and specifically agentic AI, companies and individuals experience fear, uncertainty and doubt. Consequently, the adoption of AI is much less aggressive than what most companies would benefit from. In this series, we discuss the typical challenges and the typical evolution of companies, product development processes and products. To end with a quote by Ginni Rometty, former CEO of IBM: “Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we’ll augment our intelligence.” Technology is in service of humanity and AI isn’t different in any way.
