Swanand Pagnis
Swanand Pagnis

Global enterprise investment in generative AI has reached striking levels, while questions around returns appear to be gaining attention. According to EY's AI Investment Trends report for the first half of 2026, global AI venture capital investment reached approximately $430 billion during the first six months of the year, exceeding the $254 billion invested throughout 2025. The scale of spending may be intensifying pressure on boards and executives to demonstrate meaningful business value, particularly as some AI pilots struggle to translate technological potential into visible commercial outcomes.

The financial pressure can become especially visible when enterprises evaluate standalone AI applications. Licensing commitments, implementation expenses, and employee training can turn an initial experiment into a substantial investment before meaningful usage has been established. Some organizations may subsequently find themselves with software that sees limited adoption, creating a form of "shelfware" that adds cost without becoming part of ordinary workflows. As AI spending expands, questions around how enterprises test, control, and justify those investments may become increasingly important.

The workforce presents another part of the challenge. Complex platforms can introduce learning requirements that favor employees already comfortable experimenting with emerging technology, while other workers may have fewer opportunities to develop comparable capabilities. PwC's 2026 Global Workforce Hopes and Fears Survey, covering nearly 50,000 workers across 48 countries and regions, found that 64% reported using AI at work during the previous year, while access to learning and development resources stood at 51%, down from 59% the previous year. The report also identified an "engine room" group representing 56% of workers whose skills were less scarce and who were less advanced on the AI learning curve.

Technology history provides another perspective on the moment. Earlier transitions, including the movement from physical processes to digital tools and eventually to online services, often involved stages of adaptation before more advanced capabilities became commonplace. Workplace software first translated familiar tasks into digital formats, while industries such as entertainment and music progressively moved from physical products to digital distribution and streaming. These examples suggest that major technological shifts can gain traction when new capabilities build on familiar ways of working, creating a foundation for increasingly sophisticated forms of technology.

Those pressures help explain the thinking behind Swanand Pagnis, co-founder and CEO of Superlinear Insights, an enterprise AI engineering and advisory firm. His experience across cloud, data, and AI-led transformation exposed him to organizations attempting to reconcile emerging technology with established operating environments. "A client was considering making a substantial commitment simply to try an AI tool, and that made me question the way enterprise AI was being introduced," he shares. "It reinforced my belief that the starting point should be the workflow itself, understanding how people already work and finding where AI can make that work more effective."

That experience contributed to a philosophy that places existing operations at the starting point for AI integration. Superlinear Insights works on embedding AI capabilities into environments employees already use, including Microsoft Office, internal databases, and collaboration platforms. The significance extends beyond reducing training requirements. Established tools contain institutional knowledge, permissions, processes, and patterns of collaboration. Integrating AI within those environments can give organizations room to experiment while retaining much of the context employees already understand.

Pagnis states, "I describe the principle as bringing 'AI to where work happens.'" he notes that for complex enterprises, that can involve systems supporting multiple stages of work, including research, drafting, review, approvals, document management, and internal communication. This perspective informs work across agentic workflows, private and air-gapped AI, workspace intelligence, custom AI systems, multimodal applications, and real-time decision engines. The focus is on connecting AI capabilities to the sequence of activities through which work actually gets completed.

That distinction becomes particularly relevant as enterprises move beyond simple chat-based use cases. A document may pass between employees, receive comments from a manager, incorporate information from internal databases, and emerge from a meeting with new decisions that need to be recorded. AI systems designed around those connected activities can potentially contribute across the workflow instead of assisting with an isolated task. For Pagnis, this is part of the progression from providing employees with AI capabilities toward building systems that can eventually handle increasingly complex sequences of work.

The implementation layer is another significant part of Superlinear's role. The company uses forward-deployed "utilization experts" who work alongside enterprise teams to examine workflows, identify potential applications, and help employees incorporate AI into their responsibilities. This creates a practical link between technical development and the operational knowledge held by employees. It also acknowledges that organizations can contain dozens of departments, each with distinct processes, systems, and requirements.

That human involvement becomes particularly relevant to Pagnis' view of AI adoption as a progression toward greater autonomy. He sees value in giving people powerful capabilities within familiar environments before asking organizations to make larger changes to how work is performed. In journalism, compliance, and other high-stakes fields, AI can assist with information-heavy activities such as research, aggregation, and document preparation, while human professionals retain responsibility for judgment and consequential decisions.

The broader implication is a view of enterprise AI that places implementation alongside technological capability. The scale of investment may provide organizations with access to increasingly sophisticated models, yet the practical value of those systems can depend on how they connect with workflows, employees, data, governance, and costs.

As the technology develops, AI may increasingly take responsibility for routine and multi-step operational work, giving people greater capacity for creativity, critical judgment, problem-solving, and service. The next phase of enterprise adoption could therefore be defined as much by how AI fits into human work as by how capable the underlying models become.