The main problem with artificial intelligence lies not in the model itself, but in the data infrastructure

Many companies face the challenge of implementing enterprise artificial intelligence (AI): they hire highly intelligent models capable of analyzing vast amounts of information and solving complex problems, yet these models often lack access to customer data, understanding of business processes, or the ability to act within company systems.

Despite pilot projects being conducted and budgets being approved, many executives wonder why powerful AI has not yet transformed workflows. The answer is that intelligence becomes a utility, while context, trust, and execution capability remain critically important.

BlueSky, as a partner of Anthropic, sees the solution in combining Anthropic's Claude intelligence with existing capabilities of Salesforce and Amazon Web Services (AWS). The company believes that this way organizations can move beyond experiments and start applying intelligence in practice.

The Power of Three Components

The first wave of enterprise AI focused on dialogue, where users asked questions and the AI provided answers. While useful, this did not always lead to transformative changes.

The next stage of development will be defined by the ability to execute. For example, when reviewing a loan application from a small or medium-sized enterprise (SME), a powerful AI model can analyze financial statements in minutes. However, if this model cannot access customer records, extract documents, trigger workflows, or interact with approval processes, the process will still take weeks.

The main problem for many organizations is not the intelligence itself, but connecting that intelligence to the systems where the work is actually performed. This is where the importance of the 'Power of Three' concept comes into play, represented by the analogy of 'brain, hands, and spine':

• Brain — this is Anthropic, whose Claude model provides the necessary analytical ability, judgment, and reasoning to solve complex business problems.
• Hands — this is Salesforce, because intelligence without context is merely an opinion. Salesforce provides customer data, workflows, and business processes, allowing the AI to understand what is happening and take meaningful action.
• Spine — this is AWS, as enterprise AI requires a secure, scalable, and fault-tolerant environment to operate. AWS provides the infrastructure, management, and foundation necessary for large-scale AI deployment.

Each of these capabilities is strong on its own, but their combined use yields a much greater effect. They answer three key questions that every executive should ask about AI: Is it intelligent enough? Does it understand our business? Can we trust it to work at scale?

The Fourth Element

Nevertheless, there is a fourth component. The brain, hands, and spine need a nervous system to connect them, and this is the role BlueSky plays. The partnership with Anthropic for BlueSky is not the beginning of their journey in AI, but rather the next stage of years of work helping organizations modernize operations, improve customer experience, and unlock value through Salesforce, AWS, and data-driven transformation. Their task is to connect intelligence, enterprise context, and execution in ways that lead to measurable business results.

It is important to note that this does not mean replacing existing technology investments. On the contrary, organizations should view their Salesforce and AWS environments as the fundamental foundation that makes enterprise AI effective.

Opportunities Beyond Automation

In the long term, the possibilities are far broader than just automation. The world is moving toward a place where organizations will increasingly work side-by-side with digital workers, or what some call 'silicon employees.' These systems will not just answer questions; they will help analyze information, coordinate tasks, prepare recommendations, and advance work across the organization.

This marks a profound shift. AI is not replacing one human skill; instead, it is beginning to augment a very wide range of human capabilities. The leadership challenge will be mastering the ability to utilize intelligence that operates thousands of times faster than any single manager or team.

Organizations that succeed in the AI era will start not with a model, but with the work that needs to be done and the desired outcomes. They will rethink processes around intelligence, realize that execution is more important than conversation, and recognize that scaling AI is ultimately an organizational, not just a technological, challenge.

The winners of the AI era will not be those organizations that have access to the most models, as access to intelligence is becoming universal. Instead, the builders will win: organizations that connect intelligence to their data, systems, and people, transform ideas into actions, and turn intelligence into measurable business results. Any business can now utilize this smartest new employee. The question is whether you are ready to give it a desk.

In recent years, improving AI performance has depended on the ability to write higher-quality prompts. However, this paradigm is rapidly changing as AI systems begin to reason, make decisions, and perform actions within entire workflows autonomously.

At the TechSparks 2026 event, YourStory's flagship event, Dr. Vignesh Subrahmaniam, an AI lead scientist at Intuit, held a discussion with Aditya Charlie, Editor-in-Chief of YourStory and The CapTable. The discussion, titled 'Beyond the Prompt: Building AI That Thinks, Acts, and Learns,' focused on what developers need to rethink as intelligence becomes increasingly autonomous.

The session began with questions about Subrahmaniam's typical workday and a request to simplify fintech and AI concepts for the audience. He used a chess analogy, asking: if an algorithm is not taught the rules of chess but shown a million completed games with winners, can it deduce both the rules and the most probable winning strategies?

Intuit's team applies a similar approach to support small businesses, as the company serves both consumers and accountants. Subrahmaniam noted: 'We study the game of business by analyzing data, real actions, and the results that enterprises undertake.' The goal is to understand the factors that contribute to business success in customer acquisition, rapid growth, and improved cash flow.

The conversation then shifted to the paradigm shift for developers: moving from years of studying how to write smarter prompts to designing agents given a goal, which they determine the steps to achieve themselves. Subrahmaniam compared prompt engineering to the long history of programming languages, starting from binary code and assembly up to Fortran, C, and Python. Large language models have allowed anyone to program in simple English, but this method has its limitations.

Researchers quickly realized that intelligence should not reside only within the model. The model must act as an orchestrator, breaking down a complex task into a sequence of actions and tool calls. Real intelligence moves into the called tools; and these tools must also be designed differently: 'not so that a human, or a software developer, or anyone else can use them, but in such a way that the LLM orchestrator can actually use them.'

This requires creating highly precise tools with extremely deterministic outcomes. For developers, mastery shifts from clever phrasing to reliable components that the model can call repeatedly without surprises.

Regarding sources of new data, Subrahmaniam openly stated that training on text data is largely solved. The more complex task is training on multimodal sources and real-world actions. He cited the example of how a young child primarily learns through what they see, hear, and observe, rather than reading, asserting that incorporating such real-world signals into AI systems will help models move beyond text generation and start suggesting actions. He called this the next frontier.

When considering regulated industries, he took the audience back 18 years, to one of the first machine learning labs in Bangalore where he worked on logistic regression and multilayer perceptrons. Back then, the biggest problem was not the algorithm itself. He recalled: 'Actually, it was the quality of the data.'

This problem has not changed. Whether a team is training simple regression or fine-tuning a transformer, the principle of 'garbage in, garbage out' remains valid. According to him, companies capable of collecting, aggregating, and storing the highest quality data will gain an advantage.

During a short Q&A session, one participant asked how restrictions on storing financial data affect Intuit's operations. Subrahmaniam named consent as the foundation of any training system and referenced the Framingham study, where residents of Massachusetts were allowed to track their health over time, which shaped the modern understanding of cardiovascular disease. He added that companies are also obligated to explain the benefits of data usage to the general public.

As the conversation concluded, the overarching idea was that as AI transitions from reacting to acting, foundations become more important than interfaces. Prompts will remain a starting point, but the real work now lies in creating robust tools that agents can call, supplying systems with high-quality data, and obtaining the consent that allows them to learn. It is here, beyond the prompt, that the next wave of AI will be won for developers.

In December 2025, the game Clair Obscur: Expedition 33 received two negative awards at the Indie Game Awards due to the use of Artificial Intelligence (AI)-generated textures in the final version. In contrast, corporations like Nvidia employ AI to design their own GPUs, which are used in various games, without facing calls for boycotts.

This scenario raises the question of the gaming community's prejudice against AI. The distinction lies between what is considered acceptable and what is not, requiring analysis not only of data but also of examples where the technology was well-received and of internal practices within the hardware industry.

A survey conducted by GDC 2026, based on responses from over 2,300 industry professionals, assessed opinions on the use of generative AI. The results indicated a worse scenario compared to the previous year, especially regarding the use of generative AI.

Among the interviewed professionals, 52% believe that generative AI harms the industry, a significant increase from the 30% recorded the previous year and 18% two years prior.

The numbers are even more critical when looking at the gaming public. A survey conducted by Quantic Foundry, which collected data from 1,799 players, revealed that about 85% of them expressed an unfavorable opinion towards the use of generative AI, with approximately 63% holding a very negative view.

The Quantic Foundry study points out that the gaming public's motivation is a relevant factor. They demonstrated greater negativity when AI is applied to the creation of art, music, sound effects, narrative, and dialogues.

For comparison, the use of AI to adjust difficulty was viewed negatively by 27% of the public, while its use in visual arts reached 83% aversion. This public resistance was also high in other creative areas, with 77% against music and sound effects, and 70% against narrative and dialogues. This suggests that public objection tends to decrease as the use of AI moves away from strictly creative processes.

An example of AI application that found some public acceptance is DLSS and related technologies. These systems use AI to increase frame rate (FPS), reduce latency, and enhance visual quality. Users welcomed these features due to the remarkable quality and performance achieved.

However, problems began to arise during the evolution of DLSS. Features initially focused on performance turned into dependencies for game development, causing new titles to require these tools, often presenting performance flaws even on powerful computers.

The situation worsened with the presentation of DLSS 5. This tool, which incorporates real-time neural rendering to reconstruct pixels with photorealistic lighting and materials, showed difficulties in maintaining visual coherence between scenes, resulting in the distortion of the original work and artificial and generic characteristics. Such problems were received extremely negatively by the gaming community.

Companies like Nvidia apply AI in their chip designs, covering everything from component layout to final project validation. A notable result was the development of a 64-bit adder—a circuit responsible for summing binary values—that occupies about 25% less chip area when created using PrefixRL, as reported by Jeff Dean of Google and Bill Dally of Nvidia.

At the 2026 DAC (Automation and Design Conference), companies such as Synopsys, Cadence, and Siemens displayed AI-assisted automation projects. Synopsys' long-duration design verification agent demonstrated a process 50 times faster, according to the company itself. Simultaneously, Siemens' project, in collaboration with Nvidia, presented agents capable of characterizing libraries more than ten times faster.

Although these are only two examples, they share one common point: the partial replacement of human work by AI did not generate revolt, protests, or boycotts from the public.