How Giselle is Shaping a New Era of AI-Powered Engineering

PUBLISHED FEBRUARY 02, 2025

Takafumi Endo, CEO

I’ve spent the past few years keeping a close eye on how engineering teams build and deliver software. Lately, it’s become clear that we’re standing at the threshold of a profound shift in the coding landscape. Just last quarter, for example, Google announced that one-fourth of its new code is now AI-generated—a disclosure that has sparked debate about the future of development roles. Meanwhile, AI-focused startups like Codeium (which recently secured $150 million in its Series C) and Poolside (raising an impressive $500 million) are racing ahead with advanced AI coding assistants. Even established IDE players like JetBrains have jumped in, rolling out their own AI agent integrations.

A recent McKinsey report suggests these tools might boost developer productivity by as much as 45%. In my assessment, I find that estimate conservative. From our experience at Giselle, the real impact feels significantly higher, especially in agile environments like startups or newly formed divisions, where fewer legacy constraints allow for deeper AI integration. Given Google's current pace of AI adoption, I am convinced that within five to ten years, AI will generate 75% of newly written code across the industry. At Giselle, our team has tested a range of AI solutions—ChatGPT, GitHub Copilot, Devin, Cursor, and more—and we see firsthand how these “helpful coding assistants” are rapidly evolving into primary drivers of software development and maintenance.

Beyond Coding to Full Business Operations

What’s even more striking is that AI-powered automation extends far beyond writing code. We’re already seeing signs that business operations, research, documentation, and project management are next in line for major disruption. Within a decade, we might find that three-quarters of everyday work—not just coding or product development, but tasks ranging from drafting internal memos to conducting market research and compiling legal documentation—will be either AI-augmented or entirely AI-driven. This might sound like a sci-fi plot twist, but the reality is that AI model performance is improving at breakneck speed, and the broader ecosystem, from cloud infrastructure to enterprise AI tooling, is expanding just as quickly.

All this brings a pressing question: Which organizations will thrive when mid-level work can be automated so thoroughly by AI? Now that AI-generated outputs have reached a 'good enough' baseline, the real edge will belong to teams that push far beyond “good enough.” In practical terms, that means:

1. Building unique AI engineering capabilities—custom modeling, specialized hardware, and sophisticated prompt engineering.
2. Establishing efficient but high-fidelity processes that integrate AI for rapid testing and iteration, while retaining targeted human oversight.

Three Imperatives for Future-Proof Organizations

To stay competitive in an AI-driven world, organizations must rethink how they structure teams, manage knowledge, and govern operations. Below are three imperatives that help build future-proof systems, ensuring businesses leverage AI fully while preserving essential human oversight, creativity, and ethics.

1. A High-Velocity, Full-Stack Team of Specialists

• Deep Expertise + Speed: Whether it’s legal, financial, or data science knowledge, your experts must work hand in hand, leveraging AI to accelerate insights rather than displace them.
• Leaner Teams, Faster Feedback: Because AI handles so many routine tasks, teams can remain lean. But it’s not enough to have isolated “gurus”; everyone should be skilled at quickly testing hypotheses and iterating within AI-driven workflows.

2. AI-Friendly Knowledge Assets and Prompt Management

• Capturing Tacit Knowledge: Don’t just store logs or specs. High-functioning teams record the thought processes behind how their experts solve problems—turning those into systematic prompts or “knowledge flows.”
• Prompt Libraries as a Competitive Edge: When your organization’s best thinking is distilled into prompts, AI can rapidly generate high-quality outputs reflecting unique domain expertise. This is a powerful new form of “intellectual property.”

3. A Flat Communication Layer and Integrated AI Governance

• It Goes Beyond Code: As AI coding tools become standard, we’ll still need human engineers for validation, but a broader AI governance layer (versioning prompts, tracking multi-agent interactions) becomes essential.
• Shared Language Across Teams: AI increasingly bridges the gap between technical and non-technical stakeholders, letting product features emerge from AI-mediated discussions that drive product development. Non-coders can set requirements without writing a single line of code.

Startups are especially poised to embrace these strategies quickly. With smaller teams and fewer entrenched processes, they can bridge domain silos and weave AI into each stage of product development.

Giselle’s Knowledge-Driven AI Workflow

We’ve outlined three imperatives for any organization aiming to stay future-proof in an AI-driven world: building high-velocity teams, creating AI-friendly knowledge assets, and establishing integrated governance. At Giselle, we tackle these head-on with a knowledge-first approach to AI. Our vision revolves around an Agentic Workflow Builder that moves beyond conventional AI prompts—capturing and structuring human reasoning as a dynamic “knowledge flow.” Concretely, we offer:

• AI Workflow Orchestration with Node-Based Design: Our platform lets anyone configure and deploy powerful AI agents—akin to “virtual team members”—through an intuitive, node-based interface. This lowers barriers for domain experts, enabling them to collaborate directly in AI-centric workflows.
• GitHub-First Integration and Beyond: We’ve prioritized GitHub integration, where code generation, issue tracking, and pull request reviews can be largely automated. Looking ahead, we see product creation transcending programming languages altogether, focusing instead on real-time knowledge exchange.
• Knowledge-Centric Prompt Evolution: At Giselle, “prompts” aren’t just text inputs; they’re structured representations of expert thinking—covering problem-solving processes, data sources, assumptions, and validation methods. By formalizing and sharing these knowledge flows, teams can refine AI interactions and scale expertise across the company.

Currently, our focus is on helping product development teams—streamlining code reviews, documentation, and even marketing research. However, we envision a near-future where lawyers, accountants, authors, and artists collaborate alongside developers on one platform. In this scenario, AI handles routine or algorithmic tasks, freeing humans to concentrate on creativity, conceptual thinking, and strategy. Ultimately, Giselle is about transforming AI into a medium for capturing, refining, and sharing structured intelligence at scale—satisfying each of the three imperatives needed to thrive in tomorrow’s AI-driven landscape.

Building a Responsible AI Platform

One area we mustn’t overlook is the societal and ethical impact of a world where AI composes the bulk of our professional outputs. As thrilling as this prospect is, it raises serious considerations:

• Quality Assurance and Liability: When AI takes on high-stakes tasks, accountability becomes critical. If AI-generated results carry errors or bias, who’s held responsible?
• Talent Development: Junior coding tasks may vanish, which means the next wave of engineers must adapt quickly—shifting to AI oversight, data governance, and architecture. Likewise, experts in other fields need a baseline of “AI fluency” to collaborate effectively.
• Inclusive Design: Properly done, AI bridges gaps, letting more people contribute across disciplines. Mishandled, it risks becoming the domain of a select few specialists.

Because AI has become a socially responsible business theme, the need for transparency and manageability is paramount. At Giselle, we aim to offer a platform where diverse professionals—from legal experts to engineers—can easily share knowledge, monitor agent status, and coordinate execution prompts. By giving everyone a clear window into how these AI systems operate, we believe we can foster both trust and accountability across all stakeholder groups.

We’re also working to implement responsible AI deployment from the ground up: transparent versioning, ethical oversight, clear logging, and strong feedback loops. We want AI to amplify human collaboration, which will only happen if we manage these tools responsibly and shape their evolution collectively.

Giselle’s Role in the 75% AI-Generated Era

The explosion of AI coding agents and automated workflows signals an entirely new era. Very soon, Google’s 25% AI-generated code milestone may feel antiquated, replaced by an industry standard closer to 75%. And this shift extends well beyond software engineering—business ops, research, and creative output will also undergo a rapid AI-driven redefinition.

Organizations that can skillfully integrate these AI capabilities and surpass the “good enough” threshold stand to benefit the most. By investing in specialized AI engineering, developing robust prompt and knowledge libraries, and fostering more agile internal structures, teams can stay at the forefront of innovation.

At Giselle, our commitment is to enable precisely that future. We’re building a platform where people and AI agents work side by side—cutting out drudgery, accelerating iteration, and catalyzing new forms of human creativity. Keep an eye on our progress as we refine our product, add deeper integrations, and team up with innovators across industries to illustrate the enormous potential of AI-first, agentic workflows.

The destination seems clear: AI will power the majority of development and operational tasks. The bigger question is, how do we channel that power in a way that elevates everyone’s capacity for original thinking?

I believe the best days are still ahead. Let’s welcome them—together.