From Writing Code to Building Software: i2 AI Summit

How AI transforms software delivery

Our team (Remus, Razvan, Sorin, Florin, and Voicu) spent three days in Cambridge (January 20-22, 2026) at the i2 AI Summit, working alongside client teams to rethink how AI transforms software delivery. What started as presentations evolved into hands-on collaboration, and by day three, the shift in thinking was unmistakable: we’re no longer just writing code, we’re building software with a product mindset.

i2 AI Summit Yonder

Day 1: Setting the foundation

Our CTO, Remus Pereni, kicked things off with AI in Modernizations, showing how AI can unlock value in legacy systems without risky rewrites. This framed the entire summit: the real opportunity isn’t Greenfield projects; it’s making existing systems smarter.

Razvan Dragomir ended the presentation sessions with Playbooks in AI DLC, introducing structured approaches to AI-augmented development. The message was clear: “vibe coding” doesn’t scale. You need playbooks, governance, and repeatable patterns.

Day 2: Hands-on discovery

Day two shifted from presentation to facilitation. Eight client teams worked through workshops designed to answer one question: How do we put AI to work and get the most out of it?

Our role was to guide, challenge assumptions, and help teams move from abstract AI interest to concrete use cases. The energy was high, teams were connecting AI capabilities to real business problems, not just experimenting. For example, they removed legacy components in the core of Analyst’s Notebook and rewrote them with modern frameworks and languages, using AI to enhance ingestion of unstructured data into Analyst’s Notebook for further analysis.

Day 3: Synthesis and showcase

Adam opened the final day by synthesizing the experience. Voicu Moldovan, our digital and AI transformation expert, followed with Yonder’s recommendations and key takeaways, and then each team presented its work.

The standout themes:

BMAD (agent-based development with full governance) resonated as a framework for production-grade AI work: not just prototypes, but systems with observability and SLOs.
SpecKit showed how spec-driven workflows turn AI from a productivity toy into a delivery multiplier.
Human-in-the-loop wasn’t debated – it was assumed. Teams understood that AI accelerates work, but critical gates need human judgment.

The Core Insight: Product Mindset Over Code Mindset

The summit’s plot became obvious by day three: we’re transitioning from writing code to building software.

Adam Leach, VP R&D, said, “At i2, we’ve realised that unlocking AI’s full potential requires more than bolting on a few point solutions; it calls for a fundamental, system-wide change in how we build software. We’re moving beyond quick AI fixes to an AI-first approach embedded in every stage of development, which means changing our culture and adopting a product mindset at every step.”

Writing code is tactical. Building software is strategic. It requires thinking about:

Specifications before implementation
Governance and quality gates
Measurable outcomes (velocity, defect rates, time-to-value)
Systems that evolve, not just scripts that run

AI doesn’t replace thinking; it amplifies it. But only if you adopt the discipline of a product builder, not just a code writer.

What This Means Going Forward

For teams serious about AI-augmented delivery:

Invest in playbooks, not just tools. AI coding assistants are table stakes. Structured workflows (specs, evals, HITL gates) are the differentiator.
Measure what matters. Track velocity, focus factor, and estimation accuracy, not just lines of code generated.

The Cambridge summit proved that organizations ready to make this shift don’t need more AI hype. They need frameworks, discipline, and a product mindset.

By Voicu Moldovan,
Digital & AI Transformation

Join our AI-First Hackathon in Iasi and Cluj-Napoca

Are you interested in finding out more about AI-augmented software delivery and how you, as a software company or enterprise, can apply this in your company? Then read more about our Yonder Hackathon here.

At this 2-day hackathon, we aim to have AI write all the code, making for an energetic and thrilling experience. Reach out to your Yonder representative if your company is interested in joining.