Crossroads? Complex Integration vs. Simple Tools (+AI) in Engineering Data Management

The engineering, procurement, and construction (EPC) industry stands at a crossroads. After decades of many EPC's pursuing the dream of fully integrated 3D models and unified data environments, we need to ask a hard question: Are we chasing the wrong solution?

The Current State of Integration

The EPC world has effectively split into two camps. The first serves major clients, oil and gas operators, large mining companies, who mandate 3D models and integrated datasets as contract requirements. These Owners and EPCs invest millions in platforms like Hexagon and Aveva, building complex data environments that promise seamless handoffs from engineering to construction to operations.

The second camp works with clients who simply don't care about these capabilities. They want drawings that work, specifications that are clear, and projects delivered on time and budget. Some forward-thinking EPCs use intelligent tools as differentiators, genuinely improving their deliverable production. But here's the uncomfortable truth: I haven't encountered a single EPC that executes integrated data and 3D models truly well. - I've been places...

That they keep trying speaks to remarkable persistence (or margins).

ISO standards and industry working groups have nudged the bar forward incrementally. Yet the fundamental challenge remains: EPCs must understand their business model, grasp customer requirements, and work efficiently to meet project needs—not chase technological sophistication for its own sake.

Why Integration Continues to Disappoint

The promise of monolithic engineering solutions sounds compelling. Single source of truth. Automated clash detection. Seamless, progressive data handoffs. Real-time coordination across disciplines. But reality tells a different story.

Every project becomes a snowflake requiring bespoke setup and administration. Even when clients don't explicitly request customization, EPCs struggle to rapidly deploy their own "standard builds." The complaints are universal: pipe specifications never quite match customer requirements, and even when they do, engineers still verify everything manually. Pipe support libraries, Standardized structural components? Keep dreaming.

This isn't a failure of engineering talent or effort, it's a fundamental mismatch between tool complexity and project reality. The big vendors haven't solved the core problem despite years of development and refinement.

The Lost Opportunity: Encoding Organizational Intelligence

Beyond the technical integration failures lies an even more significant missed opportunity: the major platforms have failed to help organizations encode their own intelligence.

Every successful EPC and owner organization has developed proprietary approaches, design standards refined through experience, vendor relationships built over decades, quality procedures that actually work for their business, troubleshooting knowledge accumulated through countless projects. This organizational intelligence is what differentiates a good engineering firm from a mediocre one.

Yet the "smart tools" remain stubbornly generic. They know how to route pipes and detect clashes, but they don't know:

• Your company's preferred equipment specifications for specific applications (or what you used last time in the exact same design)

• Why your team chose a particular design approach on a challenging project (holistic implementation of knowledge and lessons learned)

• Your quality review gates (which happen with exported deliverables - not in the tools themselves) and who needs to approve what

• Your hard-won lessons about what works in your operating environment

This is the real reason EPCs can't rapidly deploy standard builds. It's not just that pipe specs need customization, it's that the tools have no mechanism to capture and enforce company-specific intelligence. Every project starts from a generic baseline, then requires manual configuration to reflect how your organization actually works.

The vendors had a decade to solve this. They could have built platforms where organizational procedures, standards, and accumulated wisdom became embedded, searchable, and enforceable. Instead, they chased universal project isolated features (in EPC mode) while ignoring company-specific intelligence.

This missed opportunity is why the window has likely passed them by. In an AI-enabled world, organizations can now build lightweight systems that encode their specific intelligence without requiring monolithic platforms. The competitive advantage shifts from owning expensive tools to systematically capturing and reusing organizational knowledge.

In an AI-enabled world, organizations can now build lightweight systems that encode their specific intelligence without requiring monolithic platforms.

The AI Disruption

Enter artificial intelligence and machine learning. AI can now parse thousands of deliverables, learn patterns from decades of deliverables, and extract meaningful insights from unstructured engineering data. This capability should fundamentally shift our thinking.

Why maintain complex, integrated environments that require months of setup when AI can review simpler outputs more effectively than human checkers? The pendulum may be swinging back toward simpler tools augmented by intelligent review layers.

Oil and gas operators have made reasonable progress toward genuine integration. Mining tells a different story, I know of one major miner that has flip-flopped between platforms, each transition costing millions in licenses, training, and lost productivity. Some forward-thinking owners are exploring self-hosted engineering and design applications, betting that control and customization will deliver better results. The jury remains out.

Most owners struggle to realize value from their engineering data investments outside major capital projects. Those mega-models that cost millions to develop? They sit dormant in operations IT departments, waiting for projects significant enough to justify resurrecting them. The business case rarely closes.

The Digital Twin Graveyard

Digital twins provide perhaps the clearest evidence of this integration failure. Organizations invest millions building comprehensive engineering models during EPC execution, promising operational nirvana: predictive maintenance, real-time optimization, perfect as-built documentation. Then handover happens, and the twin flatlines.

The engineering phase produces models optimized for design coordination, not operational maintenance. Asset naming conventions don't match CMMS systems. Parameters captured during design aren't the ones maintenance teams actually need. As-built updates get rushed or skipped entirely. The data structure that made sense for engineering becomes incomprehensible to operations.

Most critically, there's no integrator bridging these worlds. Engineering consultants finish their scope at IFC. Operations teams inherit a sophisticated model they didn't spec, can't easily query, and don't trust. Within months, technicians revert to institutional knowledge and manual records because the digital twin doesn't answer their actual questions.

This isn't a technology problem, it's a handover and integration problem. The digital twin fails because nobody ensured the engineering data would serve operational needs from day one.

The Democratization Opportunity

Here's what the major EPCs have fundamentally failed to grasp: engineering data is a reusable asset, and we're entering the first era where that reuse can be truly democratized.

For decades, proprietary formats, monolithic systems, and walled gardens kept engineering knowledge locked inside expensive platforms accessible only to specialists. The promise was integration; the reality was vendor lock-in. Every project required the same pipe specs to be rebuilt, the same equipment libraries to be recreated, the same lessons to be relearned.

AI changes this equation completely. Machine learning can extract patterns from decades of deliverables regardless of source format. Natural language processing can make engineering knowledge searchable by anyone, not just CAD specialists. Automated validation can check designs against accumulated organizational wisdom without requiring every project to use identical tools.

This creates unprecedented wealth creation opportunities for both owners and engineering firms willing to think differently:

For owners: Your past capital projects contain millions of dollars in reusable engineering intelligence. Equipment selections, design decisions, lessons learned, as-built configurations, maintenance patterns. Right now, that knowledge sits fragmented across file servers, locked in proprietary formats, accessible only through tribal knowledge. Democratized data reuse means extracting that intelligence and making it queryable, comparable, and actionable for future projects—dramatically reducing engineering costs and improving outcomes.

For engineering firms: You've delivered hundreds of projects. The competitive advantage isn't in protecting that knowledge within proprietary systems—it's in being able to rapidly deploy proven solutions while customizing for unique requirements. Firms that build lightweight, reusable knowledge systems augmented by AI will deliver faster and better than competitors still fighting with monolithic platforms.

The Build vs. Buy Reality

When organizations ask about democratizing their engineering data, the instinct is to buy a solution. Another platform. Another enterprise system. Another promise of integration.

Here's the uncomfortable truth: you cannot buy operational intelligence.

The vendors can sell you tools, but they cannot sell you your organization's accumulated knowledge structured for your business model. They cannot sell you the context that makes a pipe spec decision meaningful. They cannot sell you the integration between how your engineers think and how your operators maintain assets.

More critically, the major platforms have completely missed the opportunity to encode company-specific intelligence. Every organization has developed standards, procedures, design philosophies, and hard-won lessons over decades of project delivery. These represent genuine competitive advantages—your particular approach to equipment selection, your quality workflows, your commissioning procedures, your maintenance strategies.

The so-called "smart tools" remain generic. They don't know your company's preferred vendors for specific applications. They don't enforce your design review gates. They don't capture why your team chose one approach over another on a challenging project three years ago. They can't guide a junior engineer through your organization's specific way of solving recurring problems.

This is why EPCs struggle to rapidly deploy their "standard builds"—the tools aren't smart about their business, just generically smart about engineering. And this is why the opportunity may have already passed the big vendors by. They've focused on universal capabilities rather than organizational intelligence platforms. They've built tools for everyone that encode wisdom for no one.

You must build this capability. Not necessarily build software from scratch—but build the strategy, the data architecture, the encoded procedures, and the organizational muscle to capture your company's specific intelligence in ways that create value for your business.

This means:

• Codifying your standards into queryable, enforceable formats (not just PDF procedure manuals)

• Capturing decision rationale so future teams understand not just what was chosen, but why

• Embedding your workflows into data structures so quality gates and review processes happen automatically

• Making your institutional knowledge searchable so the answer to "how do we typically handle this?" is instant, not dependent on finding the right veteran engineer

This is not a technology project. It's a capability-building, wealth generating exercise that requires understanding both your engineering processes and operational realities, then creating systems that amplify your specific organizational intelligence.

The Call to Action: Mobilize Now

The window for competitive advantage is open, but it won't stay open indefinitely. As AI tools mature, they'll become table stakes. The advantage goes to organizations that move now to:

Audit their engineering data assets: What knowledge do you actually own? Where is it? What format? What quality?

Define their reuse strategy: What decisions get repeated across projects? What knowledge would reduce costs if reusable? What questions do operations teams ask that engineering data should answer but doesn't?

Build lightweight architectures: Stop fighting to make monolithic platforms work. Design simple data structures that capture essential intelligence and can be queried by AI.

Create the integration layer: Bridge engineering and operations from day one. Design handover processes that ensure models serve operational needs, not just design coordination.

Develop organizational capability: Train teams to think about data as a reusable asset, not a project deliverable.

The organizations that crack this—that learn to democratize their engineering knowledge and reuse it systematically—will deliver capital projects 20-30% more efficiently than competitors still locked in proprietary systems. They'll build digital twins that operations actually use because they were designed for operational questions. They'll make better decisions faster because their accumulated wisdom is searchable and actionable.

The question isn't whether to pursue integrated engineering data. The question is whether to keep chasing expensive integration platforms that have consistently failed, or to build organizational capability for intelligent data reuse that creates genuine wealth.

What's your mobilization strategy? Are you building the capability to extract value from your engineering data, or are you waiting for vendors to solve problems they fundamentally cannot?

What's your experience with engineering data integration in EPC environments? Have you seen successful implementations, or do you share this skepticism about current approaches? How do you see AI changing the game over the next five years? Share your thoughts in the comments.

Thoughts by Peter Scaife 

2026-Jan-28

Revision 0

View our Project Systems Service Offering