From AI Pilot to Enterprise Scale: 4 Key Insights from the 2026 Human AI Co-Innovation Summit in Paris

How can large companies scale isolated AI pilots to cross-enterprise production?

What’s creating durable moats for AI startups and scaleups?

How is AI changing the enterprise software market and tech stack?

What roles do security, governance, and data quality play in AI purchasing and adoption?

These are a few of the pressing questions that drove the dialogue at the recent 2026 Human AI (HAI) Co-Innovation Summit. Co-hosted by the Ethical AI Governance Group (EAIGG), Forgepoint Capital, BGV, White & Case, and Eurazeo, the invite-only event gathered over 100 global enterprise leaders, AI-native founders, and investors at White & Case’s Vendôme headquarters in Paris for a series of closed-door conversations and panel discussions.

The summit marked the European expansion of EAIGG’s Human AI Co-Innovation Consortium and built upon EAIGG and BGV’s expanded AI-Native Startup Playbook, a practical guide on scaling AI-first companies with insights from 150+ enterprises, startups, and investors.

Here are four key insights we heard from enterprise leaders, AI-native builders, and technology investors scaling AI.

1) Agentic AI is reshaping the enterprise software stack and redefining competitive moats

AI is shifting the foundations of enterprise software. Software has evolved from a tool to archive data, facilitate engagement, and enable other key business functions to an autonomous driver of outcomes. We saw the first glimpse of this change with AI copilots, which operate as an assistant and productivity booster. Autonomous AI agents are taking it a step further, acting as an operational partner that executes multi-step tasks delegated by humans.

As a result, the most important success metrics for AI implementation are operational. AI technologies must demonstrate the ability to operate within business-critical enterprise systems, safely and securely initiating actions and orchestrations across tools without disrupting core systems.

For AI startups and scaleups, it’s critical to design and embed solutions that own the execution layer. Innovators that focus on vertical workflows and specialized domains- from streamlining data infrastructure, to automating niche cybersecurity and ESG compliance functions, to driving model efficiencies and automations- and generate business-critical proprietary data will solve operational pain points for enterprise buyers.

2) Execution challenges hinder enterprise AI adoption

Most large companies run dozens of AI pilots, yet few make it to enterprise-wide production. Early insights from EAIGG’s Corporate Survey Questionnaire show that 44% of enterprise AI adopters report slow, inconsistent AI solution scaling and just 33% report repeatable deployment patterns.

Model capabilities and performance rarely block adoption. More often, it’s legacy integration, change management, and user adoption challenges that derail AI initiatives. Much like the startups building AI technologies, enterprises need to be pragmatic and focus on execution and outcomes to move from pilot to production at scale.

The people, process, and technology framework remains more relevant than ever. Enterprises require AI solutions delivered with audit logs, professional services, training, and change management support. Builders and technology partners need a co-creation mindset to continuously adapt to complex legacy environments.

3) Strong AI governance is paramount

AI buying decisions are increasingly shaped by security, trust, risk, and compliance concerns. Regulations like the EU AI Act bring AI risks to the board level, elevating governance expectations at the point of purchase and throughout production.

Enterprises require highly governed knowledge layers to successfully deploy autonomous AI agents. Data provenance, privacy, quality, and ownership are key. However, data governance is a persistent challenge: according to EAIGG’s survey, 66% of organizations report battling fragmented data sources with limited governance.

AI systems are only as effective and secure as the data and context they are grounded in. AI-native startups need to be able to show how they use and secure enterprise data. They must incorporate or, at minimum, have a roadmap to enterprise-grade data frameworks like SOC2, ISO standards, and SLAs.

AI systems also must pass enterprise security reviews to detect and manage risks from data leaks, hallucinations, and unsafe or unapproved actions- particularly in the case of autonomous AI agents operating in high-risk contexts. Startups need to embed monitoring and evaluation controls into their products to address this reality. On the buyer side, enterprises should treat AI more like regulated infrastructure- not experimental tech- to align governance, risk, compliance, legal, and technology priorities.

4) As AI changes revenue models, investors and enterprise buyers expect strong fundamentals and real-world impact

Seat-based pricing is being challenged as AI features expand, driving a shift toward consumption- and outcome-based business models that link spending to value. However, without proper management, AI infrastructure and compute costs can rapidly erode margins.

Operational efficiency is foundational and unit economics matter. Gone are the days of hypothetical outcomes, vague promises, endless technology spending, and growth at all costs. Investors and enterprise buyers are looking for strong ROI and defensible revenue models when evaluating AI-native startups. Startups innovating AI solutions need to deliver impact in real-world workloads while demonstrating controlled inference costs and sustaining healthy gross margins.

Conclusion

As investors at the intersection of cyber, AI, and infrastructure software, our thesis at Forgepoint is clear: the AI experimentation phase is over. Co-innovation is essential as we navigate the shift from AI pilots to enterprise adoption. The success stories of the next era in software and AI will be the enterprises, startups, and investors who relentlessly focus on operationalizing and governing AI.

With appreciation

We would like to extend gratitude to our event partners at EAIGG, BGV, Eurazeo, and White & Case, whose collaborative insights and ongoing thought leadership informed the perspectives shared in this blog post.

Thank you to the startups and scaleups who showcased their innovative technologies: Maisa, Multiverse Computing, Qevlar AI, Fiddler, Dataiku, Neo4j, Greenly, and Asenion.

Finally, thank you to the speakers, panelists, and guests who shared their insights, expertise, and participation. Here’s to the continued work of co-innovation.

• Emmanuel Benhamou, Head of Platform, BGV and Managing Director, EAIGG
• Anik Bose, General Partner, BGV and Executive Director, EAIGG
• Neil Costigan, Chief Architect Applied AI, LexisNexis Risk Solutions
• Hala Fadel, Managing Director, Eurazeo
• Anna Felländer, Co-founder and Head of EU Operations, Asenion
• Hakim Jakhjoukh, GTM Lead, Qevlar AI
• Giovanni Leoni, Global Performance Management, Novo Nordisk
• Kurt Muehmel, Head of AI Strategy, Dataiku
• Sam Mugel, Co-founder and CTO, Multiverse Computing
• Alexis Normand, Co-founder and CEO, Greenly
• Isabel Parker, Chief Innovation Officer, White & Case
• Fahad Rizqi, President GTM, Fiddler AI
• Nicolas Rouyer, Senior Manager Solutions Engineering, South EMEA, Neo4j
• Antonio Barranco Sanchez, VP AI Startups & Venture Ecosystem, Santander
• James Summer, Managing Director, JP Morgan
• Xavier Vasques, Vice President and CTO, IBM Technology France and Director of R&D, IBM France
• David Villalon, Co-founder and CEO, Maisa

With Damien Henault and the Forgepoint Capital team.