From energy grids to data centers, trillions of dollars are being committed globally to new infrastructure and large-scale capital projects . As these projects grow in size and technical complexity, even minor inefficiencies can significantly drive-up costs.
And yet, despite decades of experience, improved tools, and well-established engineering standards, major capital projects continue to suffer from cost overruns and schedule delays. Nearly 40% of large CapEx projects exceed budgets, often by wide margins. These failures are frequently attributed to execution challenges, supply chain disruptions, or late-stage scope changes. In reality, the root cause often sits much earlier, embedded in the design itself.
In large-scale capital projects, most value is determined long before the first shovel hits the ground. Early design decisions define how materials are specified, how equipment is sized, and how tolerances are set. Once these choices are made, cost is effectively locked in, with limited opportunity to unwind them later.
Among the most persistent, and least visible sources of value leakage at this stage are overengineering and gold plating. Individually, these design choices often appear reasonable or even prudent. Collectively, they can add millions, or billions in cost without delivering material improvements in safety, reliability, or performance.
Historically, identifying overengineering early enough to correct it has been difficult. Design reviews depend on expert judgment applied across thousands of drawings, specifications, and standards. Even experienced teams can only scrutinize so much before timelines compress and designs advance. But that constraint is beginning to change.
New applications of generative AI, including BCG’s Design Challenger, are making it possible to challenge designs at a level of scale and consistency previously out of reach.
The Hidden Costs of Early Design Decisions
A well-known rule of thumb in value engineering holds that up to 80% of a project’s lifecycle cost is determined during the design phase, while only 20% can be influenced through later process efficiency improvements. Once core design parameters are set, even modest adjustments become expensive and difficult to implement.
Traditional value-engineering methods were built with this in mind. Techniques such as functional analysis, design-to-cost, and lifecycle costing were all developed to challenge unnecessary complexity and ensure designs deliver required functions at the lowest total cost of ownership.
In practice, however, applying these approaches consistently across large, complex projects remains a challenge due to:
- Design complexity: Modern assets can involve thousands of drawings, standards, and interfaces.
- Fragmented accountability: No single function is typically responsible for challenging overdesign end-to-end.
- Cumulative safety margins: Engineers, suppliers, and operators often add conservatism independently, leading to unintended compounding.
- Time pressure: As schedules tighten, reviews tend to prioritize compliance and delivery speed over fundamental design choices.
The result is a familiar pattern across capital-intensive industries: designs that technically meet requirements, but at a cost far higher than necessary.
The Mechanics of Overengineering
Overengineering rarely stems from a single poor decision. More often, it develops incrementally, through a series of well-intentioned choices that go unchallenged.
It tends to manifest in recurring ways, including:
- Overspecification: Components or structures designed for loads far exceeding actual requirements.
- Safety-factor inflation: Multiple stakeholders independently adding conservatism (e.g., a structural safety factor of 1.5, layered with supplier and operational margins), compounding to more than twice the required capacity.
- Tolerance tightening: Manufacturing tolerances that exceed what is functionally necessary.
- Material grade escalation: Use of premium materials where lower-cost alternatives would meet all standards.
- Redundancy beyond reliability needs: Designing for extreme reliability levels (e.g., 99.999%) where lower levels are acceptable (e.g., 95%).
While each of these choices may be defensible in isolation, together they create systemic cost inflation without delivering proportional value.
The challenge is not identifying a single flawed decision. It is building the capability to detect these patterns across thousands of design elements early enough to meaningfully intervene.
How GenAI Changes the Equation
The constraint in value engineering has not been the methodology; it has been capacity. Classical value-engineering techniques are sound but labor-intensive. They rely on deep expert knowledge and manual analysis, which limits how broadly and consistently they can be applied.
GenAI
changes the equation by enabling engineering teams to:
- Process large volumes of unstructured data, including PDFs and drawings
- Cross-reference multiple standards at the same time
- Identify patterns across portfolios of designs, not just individual components
- Apply consistent, explainable challenge without fatigue
GenAI makes it possible to examine far more of the design surface area, earlier and more consistently than manual review alone. Assumptions that might previously have gone untested can be challenged, unlocking value that would otherwise remain hidden.
Design Challenger: A New Tool for Engineering Cost Discipline
To put that capability into practice, BCG X developed Design Challenger, a GenAI-powered value-engineering tool that systematically identifies and helps reduce overengineering in complex engineering designs.
By combining classical value-engineering principles with modern generative AI, Design Challenger gives engineering teams a scalable, objective mechanism to test and refine critical design decisions while leverage is greatest.
The tool applies GenAI to compare what a design provides against what is actually required based on standards, functional requirements, and operating conditions. Rather than replacing engineers, it augments engineering review by surfacing potential overdesign risks that warrant closer examination, a job no-one has typically done historically.
To do this, Design Challenger ingests a broad range of inputs, including:
- Engineering drawings and revisions
- Company and industry standards
- Safety and regulatory requirements
- Design calculations and specifications
- Bills of materials and cost information
Using computer vision, large language models, and domain-specific logic, Design Challenger analyzes designs in minutes and flags areas where capacity, thickness, tolerances, or materials may exceed requirements.
Importantly, the output is not a black-box verdict. Engineers receive:
- A ranked summary of overengineering risks, highlighting where potential value is highest
- Clear explanations of why a feature may be overdesigned, including references to applicable standards
- Preliminary recommendations to guide engineering review and redesign
In this way, Design Challenger strengthens engineering rigor rather than undermining it, accelerating judgment while preserving accountability.
Integrating Design Challenger into the Engineering Workflow
Design Challenger is most effective when embedded within a broader value-engineering workflow. It supports multiple steps across the design value chain, including:
- Defining required functions based on the business brief
- Analyzing functionality to identify unnecessary or missing requirements
- Developing designs aligned to those requirements
- Comparing designs against functional needs and standards using GenAI
- Flagging mismatches between function and design (gold plating or non-compliance)
- Iterating designs to reduce cost while maintaining safety and performance
This approach ensures value engineering is not a one-off exercise or late-stage intervention. It becomes a continuous, data-driven discipline embedded in the design process itself.
A Real-World Case: GenAI Value Engineering in Mining
The impact of applying GenAI to value engineering can be seen in BCG X’s recent work with a global mining company. The organization was facing escalating design costs across major capital projects, with billions of dollars in contestable design spend. While overdesign was widely suspected, no single function owned the mandate to identify or challenge it at scale. Traditional reviews struggled to keep pace with the volume and complexity of designs in flight.
Within three months, Design Challenger was deployed and tailored to the company’s standards and asset types. The tool was used to analyze live designs, comparing drawings directly against internal and external standards.
During the pilot phase:
- Multiple overdesigned drawings were identified, representing millions of dollars in potential savings
- Non-compliances missed by prior expert reviews were detected
- Engineers adopted the tool with less than one hour of training
- End-to-end analysis took approximately two minutes per design
Beyond immediate savings, the organization gained something more enduring: a repeatable way to address overengineering across projects, rather than relying on individual vigilance.
Extending GenAI Across the Design Lifecycle
The mining case illustrates what becomes possible when design assumptions are challenged consistently rather than episodically. By applying GenAI to engineering drawings with Design Challenger, teams can embed structured cost scrutiny directly into the design process.
Building on early client results, BCG X is planning several enhancements to extend the application of Design Challenger, including:
- A standards knowledge base with a RAG-enabled chatbot for engineers and suppliers
- Automated identification of “most standard” designs to build preferred design libraries
- Supplier-facing chatbots to accelerate technical queries during design
- Comparison of as-built scans and videos against design models to identify issues before commissioning
Together, these capabilities integrate disciplined value engineering into everyday workflows where the cost of missed opportunities is highest.
Protecting Value in Early Design Decisions
Overengineering is a systemic problem, not an individual failing. It is the natural result of complexity, layered conservatism, and fragmented accountability. And because most lifecycle costs are determined during design, the biggest value opportunities lie upstream, in the choices that define specifications, margins, and performance requirements.
What has changed is the ability to address this in a disciplined way. GenAI makes it possible to challenge designs at scale without slowing delivery. Design Challenger shows how combining deep engineering expertise with modern AI can bring sharper scrutiny to critical design decisions, while they can still be changed.
As capital costs rise and technical demands intensify, protecting value at the design stage has become a strategic advantage. The organizations that embed it into their design process will be the ones that deliver capital projects with tighter cost control, fewer late-stage corrections, and far less excess built in from day one.
