BUILD Conference 2025: AI directions for the next decade
On 4 November 2025, the BUILD conference, sponsored by Snowflake, was held under the title "A luminous conversation: the AI blueprint for the next decade", with Andrew Ng, Sridhar Ramaswamy and Swami Sivasubramanian. The speakers addressed corporate AI adoption, the role of the API in business models, the rise of agentic AI and the shift toward consumption-based pricing.
4 min read
_Source: https://www.snowflake.com/en/build/_
On 4 November 2025, we had the opportunity to attend the BUILD conference, sponsored by Snowflake and titled "A luminous conversation: the AI blueprint for the next decade".
Speakers Andrew Ng (founder of DeepLearning.AI), Sridhar Ramaswamy (CEO of Snowflake) and Swami Sivasubramanian (vice-president in charge of agentic AI at Amazon Web Services) shared their views on upcoming trends in corporate AI adoption, the change this technology could bring to organisational models, and how industry professionals will need to adapt to these transformations.
A sector moving at high speed
They began by discussing the current AI landscape and the rapid pace of development across the industry, with large companies presenting new capabilities every quarter.
Despite the growing specialisation of models (coding assistants, image generators, and so on), they stressed that ChatGPT should remain the main reference point for a large share of users, not only because of its technical performance, but also because of the strong association of its brand with the very concept of artificial intelligence.
The API as the cornerstone of AI business models
Turning to what makes an AI-related business model successful, the speakers agreed that the most effective model to date is the one that offers the best API, that is, the one that lets companies reduce their costs (notably in product development) while improving their margins and generating new revenue. They nonetheless insisted that a technically superior model creates no value without a solid commercial base.
Swami Sivasubramanian summed this up in one sentence: "The best model is your business model." In other words, an excellent architecture counts for little without a product or service able to actually attract and retain customers.
The rise of agentic AI and the shift to consumption-based pricing
The conference also underlined the growing importance of robust business models as the industry moves toward more agentic AI systems.
These systems, which operate with less direct human intervention, could lead to lower implementation costs. This evolution will likely favour a shift from subscription pricing (per user) to more consumption-based pricing, closer to cloud models, where users pay for the work actually done rather than for a licence.
On this point, the speakers were optimistic about the rise of agentic AI, noting that these tools were born from developers' own need for more efficient workflows.
The main challenge, however, lies in managing compute costs. Each advance in AI models tends to increase demand for processing power. To address this, companies such as OpenAI are developing intelligent routing systems that assign each task to the most suitable model, thereby reducing costs and optimising performance.
Product-market fit: the absolute priority
The consensus was that the future of AI remains open, but that the main objective must remain the search for genuine product-market fit. Companies adopting or developing AI must prioritise user value over novelty, because customers will only adopt these products if they deliver tangible benefits that outweigh their costs. Understanding the consumer therefore remains essential.
What this means for developers
For developers, the speakers recalled that, even if everyone should learn to code, demand for fundamental computer-science knowledge has not diminished. There will always be a need for people who master compilers, memory optimisation and algorithmic efficiency. AI has automated a large share of high-level programming, but a solid grasp of the basics, together with an aptitude for mathematics and science, remains indispensable.
And for business teams?
From a commercial standpoint, although models are becoming valuable tools for evaluating and designing products, the human ability to interpret markets - notably the emotional and contextual factors that influence decisions - will remain crucial. Professionals will likely spend less time coding and more time solving concrete problems and collaborating closely with users.
Conclusion
In conclusion, the BUILD conference offered a valuable perspective on the evolving AI landscape at the intersection of technology, commercial strategy and human creativity. The speakers clearly stated that success over the next decade will depend not only on building advanced models, but also on designing resilient business models that deliver real value to customers.
As AI becomes more agentic, professionals will need to focus on responsible implementation, cost optimisation and an understanding of user needs. Ultimately, the organisations that thrive will be those that combine innovation, deep market knowledge and human sensitivity.