Why enterprises hesitate on agentic AI in development

In recent months, agentic artificial intelligence has become one of the most watched topics in enterprise technology. Tools like Anthropic Claude Code, OpenAI Codex, or Google Gemini can write code faster than an experienced developer. The problem arises when these tools are used in large organizations, where it's not just about speed, but also cost predictability, compliance with internal standards, and application lifespan.

Large language models charge for every token — a unit of text they process. When generating entire blocks of code, these costs quickly add up, especially when a team iterates and has AI rewrite the same parts over and over. Additionally, the output from generative models is not always consistent: two identical prompts may produce different architectural approaches, leading to so-called design drift — a gradual divergence between the original intent and the final implementation.

These pain points are addressed by the American company WaveMaker, headquartered in Plano, Texas. On February 19, 2026, it announced the launch of its agentic application generation system, which is built on the principle of "architecture first."

How the two-phase code generation works

The foundation of the new platform is the so-called Two-Pass Coding System. As Vikram Srivats, Head of Product Experience at WaveMaker, explained in an interview with SiliconANGLE: "We don't go directly from intent to code. We have a so-called two-pass model."

In the first phase, the developer uploads a Figma design — the most widely used user interface design tool — and supplements it with a natural language description of what the application should do. Based on these inputs, WaveMaker generates a technology-independent WaveMaker markup. This is a lightweight intermediate layer that describes the application structure without binding it to a specific programming language or framework.

The developer or architect reviews, modifies, and approves this markup. Only then does the system launch a deterministic engine that converts the markup into full production code. Because the LLM only works with this lighter intermediate layer, token consumption remains significantly lower than with direct source code generation. "Token consumption is very, very low," Srivats stated.

Design and development in one environment

WaveMaker aims to break down the traditional barrier between designers and developers. The platform functions as a hybrid integrated development environment (IDE) that combines a visual canvas (WYSIWYG editor), agentic prompts, and a classic code editor. Developers can switch between three modes: agentic, visual, and editor. This allows them to inspect and modify everything AI generates without losing control over the architecture.

The platform supports modern technology stacks including TypeScript, React, React Native, Angular, Spring Boot, and Bootstrap. Emphasis is placed on cloud-native deployment and open standards, meaning the resulting applications are not locked into a proprietary ecosystem.

Who is the platform for and what are the costs

WaveMaker primarily targets enterprise development teams, not individuals or hobby projects. "We didn't build this for individual developers. We built it for teams with diverse skills," Srivats emphasized. This means the platform works best in organizations where designers, developers, and product managers collaborate.

While the company has not disclosed exact pricing, available information indicates that WaveMaker offers trial access via wavemaker.ai and focuses on the B2B segment with a predictable cost model. The main promise: guaranteed architecture, precision, and quality at low and predictable AI costs. This is crucial for corporate CFOs, because token costs for competing tools can be opaque and quickly escalate.

For Czech and European companies, this approach is particularly interesting in the context of the EU AI Act and requirements for transparency in algorithmic decision-making. The deterministic engine and human control over the markup intermediate layer correspond to the "human-in-the-loop" principle, which European regulation supports.

Who is already using WaveMaker

Among the first publicly announced partners and customers are significant technology and industrial companies. Nokia uses WaveMaker within its Network Monetization Platform. "This step is strong proof of our shared vision: enabling AI-native software development and building better applications faster for the AI era," stated Mikko Jarva, Head of Portfolio and Architecture in the relevant Nokia unit.

Blue Yonder, a company specializing in supply chain, is deploying WaveMaker for the extensibility of its solutions. "This launch is part of our ongoing efforts to accelerate AI capabilities for customers," said Nunzio Esposito, Chief Design Officer at the company.

Additionally, WaveMaker announced a strategic intent to collaborate with Accenture, which should help growth-oriented organizations scale the agentic AI platform across industries.

How WaveMaker stands against the competition

The market for AI software development tools is extremely competitive. Beyond the aforementioned Claude Code, Codex, and Gemini, there are also specialized platforms such as GitHub Copilot, Replit Agent, or Vercel v0. WaveMaker differs from them in several key aspects:

• Architecture first: Markup with verified architectural guardrails is created first, then code. This prevents chaotic codebase growth.
• Deterministic outputs: The second phase of generation is not probabilistic — the same markup always produces the same code, which is critical for enterprise.
• Cost efficiency: Lower token consumption means significantly lower operating costs compared to direct LLM code generation.
• Application lifespan: WaveMaker emphasizes so-called long-lived applications — applications that are developed and maintained for years, not one-off prototypes.

At the same time, the platform is model-agnostic — it can work with various large language model providers, so companies are not tied to a single vendor.

Availability for the Czech and European market

WaveMaker is a global platform headquartered in the USA with clients worldwide. For Czech companies and development teams, it is available via the website wavemaker.ai, where you can register for product testing. While official Czech localization is not listed as a primary feature, the platform supports application development in any language and its technology stack is standard, which facilitates integration into existing Czech development teams.

Given that this is a solution for regulated industries (telecommunications, logistics, financial services), WaveMaker will likely need to address full compliance with the EU AI Act and European data processing requirements in the future. For now, however, it can serve as an interesting alternative for companies seeking a balance between AI speed and control over the final product.