Building businesses in the age of AI
For many companies, and especially companies for whom software is essential for value creation, the rapid rise of AI presents both an opportunity and a challenge.
This is true even for AI-native startups. In times of disruptive change, the initial instinctive reaction is often to fixate on 'how does this change how I create value?'. But, rather than trying to predict where AI will impact and to what extent (a major challenge for even the most informed experts), a potentially more productive question to ask is ‘what remains constant during such change, and will it enable me to create even more value?’
AI is the next step in a long evolution of software: from mainframes, to client-server, to cloud and mobile, and now to solutions that build on top of that infrastructure and data to unlock the development and deployment of AI at scale. But unlike the steps that preceded it, this one feels less like an evolution and more like a leap. And yet, even in rapidly changing and ultimately uncertain landscapes, certain convictions drawn from first principles thinking about what to build and how to build hold firm.
AI represents a profound opportunity for challenger companies whether a new breed of AI-native startups, or companies built before the current AI wave that run on modern, agile tech stacks and are led by visionary, customer-centric leaders who know how to shape innovation into scalable and sustainable business. It is precisely these teams that tend to deliver particularly well against long-term value creation, in both what they build and how they build and operate it.
Certain drivers of value creation in business model design have proven remarkably durable, transcending successive paradigm shifts in software. These drivers reflect the underlying reasons why customers choose a solution, stay with it, and expand their use of it, regardless of which technology layer changes beneath them. The primary value creation drivers (often referred to as competitive moats) are:
- Ownership of complex workflows
- Ownership of proprietary data
- The trust a solution earns with its users and other stakeholders
- Ecosystems that are built around a software and its resulting network effects
- Controlling distribution
Companies that build their business models around these moats have historically been better positioned to harness technology shifts and achieve sustainable success. It is plausible that the same logic applies now, making these drivers a useful compass for business model design in the age of AI. The following section revisits these competitive moats and provides an example of each from our portfolio of companies.
1. Ownership in complex workflows
Software that orchestrates complex, multi-stakeholder, multi-system workflows is structurally harder to replace. When a single workflow touches multiple departments, systems, and operators, customers require high reliability and accuracy. That kind of workflow knowledge, embedded in a platform that hundreds or thousands of operators rely on every day, compounds over time and becomes increasingly valuable as the platform grows. Complex workflows are costly to learn and even costlier to replace and organisations and their people rarely want to start over.
Kinnevik Portfolio Insight - Mews
Mews provides the operating system for c.15,000 customers across more than 85 countries. The platform sits at the centre of hotel operations, managing room availability, rate management, billing, housekeeping coordination, guest profiles, audit trails, and integrations with hundreds of third-party systems. What makes Mews structurally hard to displace is not any single one of these functions, but the connective tissue between them. A change to a room rate flows through to bookings, billing, housekeeping schedules, and guest communications in real time, and a hotel running on Mews has wired its day-to-day operations around that integration. With AI being layered on top of these core workflows, Mews is increasingly positioned as the place where those workflows are automatically executed.
2. Ownership of proprietary data
Companies that accumulate proprietary data such as expert-labelled outcomes and real world feedback loops build an advantage that compounds over time and is hard to replicate from the outside. What makes this especially powerful is knowing not just that a decision was made, but what actually happened as a result. That kind of outcome data is rare, and a new entrant simply cannot conjure it.
Kinnevik Portfolio Insight - Enveda
Enveda is an AI-native biotech company using machine learning to discover new medicines from the chemistry of natural products. Enveda is not developing a single drug, it is building a platform. This platform combines mass spectrometry data with machine learning to map the chemistry of tens of thousands of natural sources, building a proprietary metabolomics database that we believe is among the most valuable in biotech for evaluating AI-generated molecular candidates. Every experiment the company runs feeds back into the platform, which means each new program makes the underlying models more capable, and makes each new drug candidate cheaper to develop and hopefully also more likely to succeed. Six years in, Enveda has shown it can identify new drug candidates c.4x faster and at 1/10th of the cost of traditional pharma, and already has three drug candidates in clinical trials. The combination of irreplaceable training data, a continuous experimental feedback loop, and early clinical validation creates a moat that is very difficult to replicate.
3. Trust a solution earns from its users and other stakeholders
In complex industries, especially in mission critical and highly regulated ones such as healthcare, compliance and certifications are key to obtain customer trust and are among the hardest things for new entrants to replicate. This trust, from customers and regulators, requires years of real-world operation.
Kinnevik Portfolio Insight - Oviva
Oviva is a leading European virtual chronic care platform, delivering clinically validated programmes for obesity, hypertension, and related conditions through public healthcare systems. In AI-enabled healthcare, trust and regulation are critical, requiring rigorous clinical evidence, real-world outcomes, data privacy standards, and integrated clinical oversight. Oviva has spent years building this position, with over 90 peer-reviewed studies and large-scale real-world evidence demonstrating best-in-class outcomes, including 15% weight loss over 12 months and a 35% reduction in sick days within six months of treatment. With over one million patients treated to date, the platform continues to strengthen its evidence base while leveraging AI-driven treatment personalisation to further improve outcomes at scale. This has created a strong regulatory and trust moat, particularly valuable in Europe’s fragmented healthcare landscape.
4. Ecosystems that are built around a software and its resulting network effects
Digital platforms tend to become more valuable as more participants join, whether those participants are users, developers, or in future even bots (so called AI agents). True network effects happen when more usage attracts more activity and contribution, and more contributions make the platform more useful for users. Ecosystem depth reinforces this. Once a customer has wired its adjacent systems and workflows around the platform, the platform becomes the connective tissue across the entire stack.
Kinnevik Portfolio Insight - Tandem Health
Tandem Health is a European medical AI co-pilot used by over 5,000 healthcare organisations across Europe. The platform integrates with more than 100 electronic health record systems, enabling one-click transfer of clinical notes, codes, and documents directly into each organisation's medical record. Each new EHR integration makes Tandem accessible to a wider base of clinicians, and each new clinician adopting the platform increases the institutional context Tandem accumulates about how that organisation documents and delivers care. The system improves with every interaction through a continuous learning loop by which real-world clinical errors are captured and fed back into the model, and the platform adapts to each clinician's documentation preferences and specialty, becoming more accurate and efficient with use. This creates a compounding dynamic where the product gets smarter with every clinician interaction.
5. Controlling distribution
Companies with established customer bases, embedded distribution, and the scale to launch additional products faster than new single-product challengers can have a structural advantage that new entrants must spend time building first. A good product with great distribution may well beat a great product with weak distribution, and as AI lowers the cost of building new products, distribution becomes more valuable as bottlenecks may shift from product development to customer acquisition.
Kinnevik Portfolio Insight - Cedar
Cedar is the leading healthcare financial engagement platform in the United States, handling the full patient billing and payment journey across health systems and provider organisations. To date, the platform has supported more than 50 million patients, processed over $10 billion in payments, and accumulated more than 1.5 billion patient interactions across its provider network. From this base, Cedar has expanded beyond billing into adjacent products including coverage discovery, AI-powered patient support through its Kora voice agent, and most recently Cedar Intelligence, an AI decision engine that personalises patient financial journeys across more than 80 behavioural attributes. Each new product is distributed through the same provider relationships Cedar has already built, and each new patient interaction strengthens the data foundation that powers the next capability. This combination of a blue chip health system client base, a growing patient data corpus, and disciplined expansion into adjacent financial workflows is a clear example of scale-driven compounding in product innovation.
In concluding on the above, it is clear that successful challenger companies share some common DNA and mindsets. Firstly, they have a strong and unrelenting customer focus and, secondly, they strive constantly for operational improvement and advantage. Such companies will typically embrace AI precisely because they have more to gain than to lose.
1. Customer focus
In challenger companies, customer centricity tends to be deeply embedded in the culture and felt across the entire organisation. AI sharpens that edge. Those who truly understand their customers and embrace a 'working backwards from the customer' mindset are now able to ship new features faster, personalise experiences at scale, and close the gap between what a customer wants and what the software delivers, more efficiently and effectively than ever before.
Kinnevik Portfolio Insight - Spring Health
Spring Health is the US' leading provider of virtual behavioural health care, serving more than 10 million members across over 4,500 employer clients. The company has accelerated its product roadmap in ways that would not have been possible without the underlying AI and data infrastructure. Recent launches include a continuous care product that provides AI-first interaction and care navigation between sessions, Vera-MH, an open-source evaluation framework that is creating the standard of care for conversational AI in mental health, and AI-powered clinical tools such as notetaking, in-session copilot insights, and panel management. Each new product is built on the same proprietary dataset of millions of care journeys, which means the cost and time to launch the next capability falls with every interaction the platform processes. Spring Health showcases how a strong data foundation and strong AI adoption translates directly into faster product development and broader clinical value.
2. Operational improvement
Challenger companies refuse to accept that the current way of doing things is good enough. They constantly question, challenge, and look for better ways to operate. Teams that were already pushing for efficiency gains will find AI to be a genuine multiplier for operational improvement: lower cost-to-serve, faster internal workflows, and human intelligence freed up for the work that benefits the most from it.
Kinnevik Portfolio Insight - Perk
Perk is an end-to-end business travel and spend management platform, processing billions of dollars in annual travel spend across more than 9,000 customer companies globally. One of the clearest examples in our portfolio of AI as an efficiency lever is Perk's customer support operation. In the past three years, AI-driven automation of customer support has driven a c.70% reduction in cost-to-serve, resulting in a gross margin expansion from c.40% to c.75%, while customer satisfaction scores have improved from 86% to 91% over the same period. Currently, AI handles routine modifications, confirmations, and simple queries autonomously, while anything complex, high-stakes, or ambiguous is escalated to a human agent with full context so the customer does not need to repeat anything. Perk translates AI efficiencies directly into fixed cost savings, which in turn creates the financial optionality to continue investing for customer innovation.
To benefit from the latest wave of technological change, companies will need to discern the signals in the noise and set clear and achievable priorities accordingly. Not all competitive moats may have moved as drastically as one may think, but the ways of broadening and deepening them have. The degree to which AI defines, shapes or enables business models varies - understanding that, moving quickly, and keeping the customer front of mind will define success. Now is the time to build the foundations that will continue to drive customer adoption, satisfaction and trust over time.
