Reshaping the economics of software: Building a future-ready core with Al/works™

📝 摘要

Thoughtworks 的 AI/works™ 智能体开发平台正从根本上重塑软件交付的经济学。Constellation Research 的报告指出，该平台通过在数字价值链的每个环节注入自动化，使开发者速度从"传说"变为现实——单个开发者产能可提升 5-10 倍。文章核心论点是：仅仅在现有 IDE 上"外挂" AI 功能是不够的，企业需要构建 AI 原生核心，才能彻底改变软件开发的成本和速度。

核心概念： 规格驱动开发（Spec-Centric Development）是 AI/works™ 平台的核心方法论——AI 先将遗留代码逆向还原为可理解的规格说明，再经行业最佳实践丰富，最后从规格自动生成高质量代码，从而桥接业务意图与技术执行。

关键技术栈 / 关键要点：

Code-to-Spec — AI 解析、解构和合成遗留代码库，将其提升回人类可理解的"规格层面"
Spec Enrichment — 规格说明可融入行业最佳实践和垂直领域知识
Spec-to-Code — 平台以空前的速度生成符合规格的高质量代码
多智能体编排 — 最佳实践建议优先选择确定性编排平台（multi-agent orchestration + harness engineering），补充而非替代现有工具
自适应治理 — 安全至关重要，但不应让僵化监督阻碍进展；拥抱随使用量扩展的护栏，而非一刀切的"否决"
现代化经济学翻转 — AI 加速大幅压缩过渡窗口，打破维护双系统的财务僵局

关键结论 / 实践建议： C-Suite 应聚焦三点——(1) 优先选择确定性编排平台；(2) 先构建再学习再治理（Build, learn, then govern）；(3) 以业务成果和复利价值评估技术，而非仅看成本。重建核心不只是更快编码，而是构建更快、更智能、更经济的数字化转型引擎。

For the modern enterprise, experimenting with AI is over. Now, the focus has shifted from "what if" to "how fast," and as digital value chains evolve, CXOs are seeking to fundamentally reshape the economics of software development rather than just looking for increased productivity.

A report by Constellation Research on Thoughtworks’ AI/works™ platform highlights how the fabled goal of developer velocity has become a reality by injecting automation into every link of the digital value chain. With the AI/works™ Agentic Development Platform, teams can instantly decode legacy code, bridge the gap between business requirements and execution, and accelerate the entire build cycle.

But truly changing the economics of software delivery takes more than just adding AI plug-ins to old workflows. It requires rebuilding an AI-ready core that ensures security and scalability.

Moving beyond “bolted on” AI

Many organizations started to incorporate AI by adding features to traditional integrated development environments (IDEs). However, as the Constellation Research report notes, these "bolted-on" capabilities often lack a deep understanding of enterprise demands.

To change the cost and speed of development, enterprises are now moving toward AI-native platforms that don't just suggest the next line of code, but provide a framework for the entire software development lifecycle (SDLC). In this new model, the human role shifts from writing code to editing and validating it.

The power of spec-centric development

One of the most significant aspects of modern AI platforms is the return to specifications (specs). While code is often understood only by developers, specs can be read and updated by business leaders across the organization.|

By prioritizing spec-driven development, companies can bridge the gap between business intent and technical execution, across:

Code-to-spec: AI can now parse, deconstruct and synthesize legacy code bases, elevating them back to a "spec level” that humans can understand.
Spec enrichment: These specs can then be enriched with industry best practices and vertical-specific content.
Spec-to-code: Finally, the platform generates high-quality, spec-conforming code at previously unseen speeds.

This "spec-centricity" ensures that the software actually does what the business requires, which Constellation calls "the ultimate prize of software development".

Reshaping the economics

This economic shift in IT services is revolutionary. By leveraging AI-native platforms and a composable architecture of continuously updated context and components, an individual developer can multiply their capacity by 5x to 10x.

This unprecedented velocity reshapes enterprise strategy in two ways:

Flipping modernization economics: AI acceleration drastically compresses the transition window, breaking the financial gridlock of maintaining dual systems. Organizations move from static codebases to a stream of continuously updated, modernized code.
Evolving maintenance: As autonomous agents handle end-to-end tasks, the traditional maintenance cycle is replaced by software that constantly adapts to evolving user needs, UI choices and cyber defenses.

The bottom line is that optimizing the balance between human supervisors and AI platforms permanently drives down the cost of building and running digital assets.

Guidance for the C-suite

To rebuild for this new reality, executive leadership should consider three best practices:

Prioritize deterministic, orchestrated platforms: Look for foundational platforms/frameworks, like multi-agent orchestration and harness engineering, that complement and scale your existing tools — rather than replacing them with unproven tech.
Build, learn, then govern: While security is paramount, don't let rigid oversight stall progress. Instead, embrace adaptive governance — using guardrails that scale with usage rather than a blunt "no" up front.
Focus on value, not just cost: Evaluate technology by its business outcomes and compounding value, not just cost. By focusing on reliability, fast modernization, human-augmented AI and deterministic outputs, you can avoid technical debt and build a predictable, future-proof engine for growth. Rebuilding the core isn't just about coding faster. It's about building a faster, smarter and more cost-effective engine for digital transformation.

The Constellation Research Pulse Report: Thoughtworks jolts enterprise AppDev into the AI era

Read the full report to discover how Al/works™ enables enterprises to achieve higher developer velocity through a spec-centric approach. Discover critical 2026 trends and get a deeper insight into Thoughtworks' unique ability to modernize legacy code into actionable specifications and high-quality automation.

Disclaimer: The statements and opinions expressed in this article are those of the author(s) and do not necessarily reflect the positions of Thoughtworks.