Maria Basso is the Head of AI Applications and Impact at the World Economic Forum’s Centre for AI Excellence, where she leads global initiatives that harness AI to transform industries and society. With a background spanning the United Nations, McKinsey & Company, and the University of California, Berkeley, she brings deep expertise in sustainable development, innovation, and digital transformation. Maria holds a Masters in Mechanical Engineering and Innovation in Energy Systems from the Politecnico di Torino and the University of Illinois. She spoke to The Innovator about the Forum’s recent whitepaper “AI-First Operating System: A Blueprint for Operating and Business Model Innovation” and what the Forum is trying to achieve with its MINDS (Meaningful, Intelligent, Novel, Deployable Solutions) Programme, which identifies and showcases companies that have deployed impactful AI solutions.
Q: What are some of the key takeaways from the white paper?
MB: We published this white paper because we see that AI adoption keeps increasing but some organizations are moving extremely fast and more effectively than others, and so we wanted to study and explore what those organizations are doing. Adopting AI is no longer a differentiator nor is the number of tokens that each employee uses, which is sometimes viewed as a key metric. The real differentiator is how organizations are redesigning themselves around intelligence. Most of the value of AI is achieved when organizations rethink how decisions are made, how works gets done, and how they scale the technology. This is really complicated because it often means changing reporting lines and oversight, and therefore it’s a massive change in many organizations. And that’s why it doesn’t get touched. It’s much easier to just say, “Oh, I let my employees adopt and use more AI tools.” But what is required goes far beyond adopting AI tools.
Q: What does it entail?
MB: We see three types of organizations evolving: the AI-enabled, which is where most organizations are today; AI-first, and then the AI-native organizations. At AI-enabled organizations, if AI tools disappeared tomorrow, workflows would continue largely unchanged. AI-first companies are enterprises that systematically redesign workflows, roles and decision rights so AI becomes the strategic lever for creating and delivering value at industrial scale. If AI tools were removed, the organization’s workflows and infrastructure would collapse. AI-native companies are new enterprises created with AI as a core production capability, with products, services and competitive advantage fundamentally dependent on AI. If AI were removed, their value proposition would not exist.
Today, most organizations are AI-enabled, so leveraging a lot of AI tools, but not really rethinking how they operate. This is why despite all the billions that are being invested globally on AI, only 20% of companies say the technology is having a transformative impact. Our white paper notes that 84% of companies have not redesigned jobs around AI. So, the challenge is not really a technology challenge, it’s the organization’s ability to adapt and transform. Another interesting insight is the fact that the organizations that are transforming are treating intelligence as a factor of production, which means they actually allocate it deliberately across decisions, across business outcomes, across workflows in a very similar way to how previous generations allocated capital and labor. This is especially important now that we see the costs of AI rising. Some tech companies are limiting the use of AI because they are realizing that it is important to use it deliberately, as if it were capital. They understand that they need to use the technology to reshape and rethink workflows and business operations and aim for impact. This links to the MINDS Programme. We created this program because we need to put a spotlight on examples of companies that are achieving impact. That’s where AI needs to go, not everywhere, not randomly. Companies need to start measuring outcomes and productivity and think about which metrics make sense so that they can track progress on becoming AI-first. Organizations need to ask themselves, ‘Are we using AI as an add-on, as a tool, and if we remove it, would nothing much change? Or are we taking a step back with our CEO and with our leadership and rethink the operating model with AI?’ AI-first organizations are not looking at the operating model and the workflows and saying ‘This is our way of working and this is where we can put AI.’ It’s the opposite. AI-first organizations forget about their current operating models and workflows, and they say ‘Ok, this is what AI, if we started today, could do for our system’ and then they place all the workflows and the relevant operations accordingly. These AI-first organizations are built around what we call the intelligence engine.
Q: What is an intelligence engine?
MB: It is an AI system based on all the data that the company has. It learns from every interaction, improving in every cycle, compounding the knowledge of the organization over time. Every customer interaction, every internal operational workflow, every decision taken from any level compounds and generates new insights, new data, new signals that can strengthen future performance. It’s really a new way of looking at competitive advantage. It’s this engine that the more data it has, the more you leverage it for the core of your operating system and the more it actually learns and improves. So, the intelligence engine needs to be at the core, and then we identified four other key building blocks.
Q: What are the other key building blocks?
MB: One is an adaptive technology stack. It’s important that companies build modular technologies: architectures that allow capabilities to scale while being flexible as new innovations emerge into the market. In this way, organizations can upgrade their models, upgrade their tools, upgrade their infrastructure without disrupting their workflows. It is important not to be dependent on a single model or vendor because no one knows how the technology is going to evolve in the future.
The third block is what we call intelligence allocation, which shapes how operations are designed. Workflows stop being processes to execute and become systems that learn. The intelligence needs to be treated as a strategic resource and leadership needs to deliberately decide where to invest across workflows. The objective is not just to automate everything, but to redesign how work is performed and how decisions are made to continuously support these intelligent engines.
The fourth block is human-AI collaboration. When it comes to human-AI collaboration, what is important is having cross-functional teams organized around specific workflows that normally combine technical and operational domain expertise, and operate with AI. What’s interesting is seeing how the org structure changes. Normally in AI-first organizations, the teams are much leaner. They are directly connecting to the intelligence engine, and the structures are more flat to enable more rapid decision making and more direct accountability for outcomes. It will be quite fascinating to keep exploring how this will evolve. Agility is one of the things that keeps emerging as being very top of mind for all the AI-first companies. That is why they normally have a centralized AI governance structure throughout the organization, but they decentralize all the business ownership. This allows all the business units to come up with new ideas to innovate while maintaining common standards, common infrastructure, and a common guardrail. The fifth building block is new value creation. What are the new business models that AI enables? We have started to see some of them through our MINDS program.
Q: Can you share some examples?
MB: Two companies in the MINDS’ Programme come to mind. Quipu, which is based in Colombia, is using AI to help informal workers and microbusinesses in Latin America access credit when they lack traditional credit histories or formal financial records. The platform uses alternative data, such as financial SMS phone notifications and business photos, to generate explainable credit scores and recommendations. It has reportedly processed more than 200,000 applicants and supported around 49,000 loans, with an 80% renewal rate and a decrease in clients’ use of informal lenders from 38.6% to 28.5%.
China’s Black Lake Technologies, a developer of AI-agent manufacturing software, links demand, design and production in real time, supporting new business models such as smaller production runs and faster delivery, and ensures that custom workflows can be deployed in days rather than months.
Q: What do you see as some of the key challenges around value creation?
MB: I would say customer retention is the defining challenge. AI reduces the switching costs and accelerates imitation. Novelty attracts users but what remains key is that even more than before only sustained value creation can drive long-term adoption. This really needs to be put at the core on this new value creation agenda. The most successful organizations are designing customer experiences that continuously strengthen the intelligence engine. In this way, through all the ongoing engagement and feedback and capturing everything, they can get all the insights to drive continuous value creation for their customers. That’s the new IP. It makes it more difficult to be replaced by other companies, but organizations now also need to consider the impact of AI agents and their influence on purchasing decisions. Discovery is completely changing and that is going to be a big challenge. It’s important that organizations stay on top of rethinking not only what they sell, but how customers discover and compare and commit to their offerings.
Q: In the white paper you stress that the building blocks are not a roadmap for transformation. Instead, you suggest leaders use them as a lens: to see what is now possible and to design accordingly. What can they learn from companies in the MINDS Programme?
MB: A number of interesting things. We announced cohort three in June. We had applications from over 30 countries in 23 industry sectors. 60% of applicants were small and mid-sized organizations, which demonstrates that impactful AI adoption is not limited to large enterprise. Operations and production were the most common area for AI deployment because it’s the easiest first step. That is followed by engineering and R&D. R&D took much more of a central role in this cohort. It is fascinating to see how fast AI progress is moving there. Something interesting that we found is that the single biggest driver of these AI initiatives was not internal but external in the form of customer demand. This shows that AI is really being pulled by the market. Data and talent are, not surprisingly, among the top constraints. But in terms of interesting insights, I would say a few overarching points we saw in all the MINDS applications. One, is related to workforce. Our data is limited, but we saw that organizations building those best-in-class AI applications are not the ones that are shrinking their workforces. These organizations are the ones that are investing in people first. To get to those best-in-class applications, they invested massively in training and in development of talent. We also got insights into how tasks are affected. AI performs some tasks but reallocation of the human head count outweighed the reduction in jobs in the MINDS organizations. And, even if work was delegated to AI, more than 90% of the MINDS organizations had a very thorough way to stay in control. The vast majority of organizations were keeping humans-in-the-loop for final decisions, citing accountability, legal liability, and ethical judgment, as the reasons. This is very interesting because if you go outside of the pool of MINDS companies, the question of who is in charge of the final decision is still quite messy, and there is not proper governance. It’s not well defined. Then, from a model perspective, we saw that instead of using large models across the board, MINDS companies are using a mix of smaller models trained on industrial, physical, scientific, or domain data. On agentic AI, agents are for sure emerging in impactful applications, but MINDS organizations are matching the autonomy to value, so putting agents where they really earn their place versus where they don’t and pairing higher autonomy with stronger human-in-the-loop oversight. And then we saw a lot of AI stepping into the physical world at MINDS companies, from autonomous driving to advanced adaptive manufacturing to the next generation of robotics.
Q: What would you like The Innovator’s readers to take away from this interview?
MB: Technology will continue to evolve at an extraordinary pace, and we’re continuously learning what good adoption looks like. If we continue to pair innovation with thoughtful governance and a willingness to rethink old assumptions, we have an opportunity to build organizations that are more resilient, more adaptive and ultimately better equipped for the future.
