Building AI-First Companies and the Agentic Evolution
The Decade-Long Perspective: The Evolution of Iris.ai
The journey of Iris.ai, founded at Singularity University (GSP 15), provides a blueprint for the “long game” in artificial intelligence. Long before the current hype cycle, the company set out to solve the fundamental problem of information overload in scientific research. Their decade-long evolution illustrates a critical transition: building for a technical future that had not yet arrived, only to find the market finally catching up during the “ChatGPT moment.”
For the modern executive, the Iris.ai story is a masterclass in navigating pivots to find high-value applications for deep-tech systems.
The Commercial Evolution of an AI Pioneer
| Target Market | The Offering | Outcome / Lesson Learned |
|---|---|---|
| Entrepreneurs | AI-driven research tools for solving “big hairy” global problems. | Market Misalignment: Entrepreneurs lacked the significant time required to digest research and, crucially, lacked the capital to fund the service. |
| University Libraries | Research access and discovery tools. | Scaling Bottleneck: While they captured the entire Finnish market, the segment lacked the scale and resources to support exponential AI growth. |
| Corporate R&D | Rigorous, systematic tools for knowledge management and discovery. | Structural Resistance: Strategic AI funding often bypassed R&D departments, which remained hesitant to adopt complex, rigorous tools despite technical efficacy. |
| General Enterprise | Data unification and agentic knowledge work automation. | Product-Market Fit: Floodgates opened ~1.5 years ago. High demand for moving scattered, expert-level data into production-ready, actionable systems. |
Bridging the Enterprise “Valley of Death”
In the current landscape, the “Valley of Death” is the graveyard of Proof of Concepts (PoCs). While many organizations can make AI work on a curated set of 100 documents, moving to a production environment of 10,000 or 100,000 documents remains the single greatest bottleneck to ROI.
The Two Gaps Halting Progress
- The Data Gap: There is a pervasive myth of “clean data.” In reality, internal expert data is “entirely scattered,” messy, and unorganized. Scaling beyond a pilot reveals that data readiness is an infrastructural requirement, not a secondary concern.
- The Human Gap: A structural tension exists between C-suite mandates and knowledge worker anxiety. When leadership pushes for “automation,” workers fear displacement. Success requires reframing AI not as a replacement for the worker, but as a replacement for the worker’s most “boring” tasks.
Strategic Mandate: The Interpretation Layer To cross the Valley of Death, enterprises must move beyond simple storage. Success requires a Data Unification and Interpretation Layer—a specialized context layer that transforms raw, scattered data into a structured format that agentic systems can actually act upon.
The Shift to Agentic AI: Beyond the Chatbot
We are moving past the era of the “chat function” into the era of the Agentic System. While standard LLMs provide information, agentic AI executes tasks by interacting with browsers, internet resources, and internal computer systems.
- From Chat to Co-Thinking: The transition is best exemplified by the shift from basic Claude to Claude Co-work. This is no longer a back-and-forth dialogue; it is a partner that can open browser tabs, execute searches to fill information gaps, and assist in multi-step task execution like slide creation.
- The Deceptive Nature of the “Agentic Moment”: Unlike the ChatGPT moment, which felt like a revelation, the agentic shift is deceptive because it fulfills our current, latent expectations. We are rapidly approaching a point where we will simply forget that chatbots once couldn’t open a browser or interact with a desktop. This “normalization” of exponential change is a sign of its deep structural integration.
Strategic Frameworks for AI-First Leadership
“Buying a license is not a strategy.” The default move—purchasing enterprise-wide licenses for tools like Microsoft Copilot—is often a “cop-out” that avoids the hard work of identifying where AI will create the most transformative value.
The C-Suite Strategy Checklist
- In which specific business unit will this investment generate unique competitive advantage?
- Is this a transformative vision or merely “housekeeping”?
- Does our data infrastructure support this tool at scale, or are we funding another failed PoC?
- Critical: Is this investment driven by a long-term vision, or is it a default purchase from an existing service provider?
The “Outside-In” Model for Legacy Pivot
A 78-year-old, 4th-generation Nordic retail giant recently demonstrated how “old-guard” companies can lead. Their strategy development involved:
- External Pressure Testing: Bringing in diverse AI experts for five intensive, two-day sessions over a full year to forecast a 10-year horizon.
- Internal Talent Discovery: This process uncovered “hidden” internal experts who were already practicing AI but lacked a formal framework or leadership support.
- The Sparring Partner Role: Leadership utilized external experts as “sparring partners”—a safe space to test radical ideas and identify opportunities, such as cloud migration strategies, that internal silos often obscure.
Cultivating an AI-First Culture
Cultural shifts cannot be mandated; they must be engineered through engagement.
The “Excel Spreadsheet Method” for Buy-In
To eliminate fear, ask knowledge workers to break their day into a step-by-step spreadsheet and highlight the “boring” tasks. By focusing automation exclusively on these undesirable elements, you transform AI from a “job taker” into a “drudgery eliminator.”
Operational Mandates: The Iris.ai Blueprint
To become an AI-native organization, leaders should adopt these specific internal policies:
- The Experimentation Mandate: All staff (regardless of role) must spend 3–4 hours per week experimenting with AI tools. The only required output is a weekly session to share what failed and what worked.
- The 65% Benchmark: Iris.ai demonstrated the power of this mandate by increasing its AI-generated code from 30–35% to 65% in a single year.
- The Architectural Hiring Shift: Prioritize “ridiculous enthusiasm” for AI applicability. Hiring should move away from “coders who code” toward individuals who understand overall architecture, orchestration, and quality assurance.
- Performance Integration: AI collaboration must be a formal metric in performance reviews, evaluating how effectively an employee augments their role with technology.
The Future of Work and the “Human Premium”
The future organization will likely bifurcate into two distinct commercial models: the “One Person/One Thousand Agents” high-efficiency model and the “Human-Centric” premium model.
“People for people and automation for everything else.”
This aphorism should guide all strategic deployment. AI remains insufficient—and often inappropriate—in scenarios requiring:
- Empathy and Accountability: Situations requiring personal understanding, love, care, or life-altering sensitivity.
- The Anti-Slop Filter: As the web becomes flooded with “AI slop” (low-quality automated content), brand value will increasingly be tied to the “Human-AI joined effort” that maintains quality and nuance.
- Complex Context: High-stakes customer service where a human must take personal responsibility for a non-linear problem.
Societal Shift: We must prepare for a future where “jobs” are fluid and project-based rather than lifelong. This decoupling of labor from survival will eventually necessitate government-led Universal Basic Income (UBI) to ensure societal stability as automation scales.
Deployment and Ethics
Recommended Strategic Tech Stack
- Claude / Claude Co-work (Anthropic): The current gold standard for “co-thinking” and agentic task execution.
- Llama (Meta): The preferred choice for open-source, on-premise deployment where data security and model modification are paramount.
- Gemini / Tactic: Highly effective for automated meeting transcription and synthesizing strategic notes.
Ethical Selection as Business Intelligence
Strategic tool selection must look beyond performance to the provider’s ethical stance. There is a clear distinction in the market today:
- Anthropic is the leading example of a provider setting clear ethical boundaries.
- The Warning: Leadership should be wary of providers who cooperate in the development of autonomous weapons or government surveillance/population spying. Selecting a provider like Anthropic isn’t just an ethical choice; it is a risk-mitigation strategy for companies that value long-term credibility and human-centric values.
Conclusion: The Long-Term Vision and Strategic Choices
The workforce of the future will be “fluid,” moving away from fixed 40-year careers toward project-based relationships. Enterprises must decide now which path they will take: the High-Efficiency Model (one person managing 1,000 agents) or the Premium Model (where AI handles logistics, but the value is human-centric).
Leader’s Directive
Strategic leadership now requires answering three high-stakes questions:
- Sovereign Jurisdiction: Will you align with US-based providers who are “in bed” with governments to build autonomous weapons and surveillance systems? Or will you prioritize the EU/GDPR model, favoring ethics, compliance, and rigorous data boundaries?
- The Social Safety Net: As work becomes fluid, how will your organization advocate for or adapt to Universal Basic Income (UBI) to ensure a stable consumer base and social fabric?
- Ethical Resource Management: How will you lead in an era where financial resources are increasingly concentrated among a few model providers?
The transition to an AI-first enterprise is not a technical upgrade; it is a fundamental reorganization of human contribution and value.
