Over the past year, I’ve noticed an interesting pattern in conversations with venture capital firms evaluating PostSig Investor Rights Intelligence. Almost every firm reaches the same conclusion at some point in the discussion.

“Couldn’t we just build something like this ourselves?”

A year ago that question would have been hypothetical. Today it’s completely reasonable.

The Golden Age of Building Software

The combination of Claude, ChatGPT, Cursor, Lovable, and other AI-native development tools has fundamentally changed how software gets built. What once required months of engineering effort can now be prototyped in a matter of days. A small team can create a polished portfolio dashboard, a KPI collection workflow, an internal reporting tool, or even an AI-powered portfolio management application with remarkable speed.

The quality of these first versions is often impressive. In many cases, they solve the immediate problem well enough that buying commercial software begins to feel unnecessary.

What’s interesting is that I found myself thinking exactly the same way.

Earlier this year, I started exploring whether we could build parts of our own internal operating stack using AI-native development tools. Like many founders, my instinct wasn’t to compare pricing or evaluate feature lists. Instead, I looked at them and thought, we can build this ourselves.

And we did.

Within a short period of time, we had assembled an internal CRM and GTM platform that covered much of what we needed. It tracked accounts, managed contacts, organized opportunities, automated repetitive tasks, and connected naturally to AI. Looking at the product, it was easy to feel that we’d made the right decision. The software worked, it fit our workflow, and we had complete control over the experience.

Then reality slowly caught up.

The first feature request was easy enough. Then another team wanted a different workflow. Sales asked for changes that marketing didn’t want. Customer Success needed additional data. Someone discovered edge cases around permissions. Reporting became more complicated. Integrations had to be maintained. Bugs appeared. Documentation was needed. Every improvement created two new ideas that hadn’t existed before.

Without realizing it, we had started building a software company inside our software company.

That experience changed how I think about AI-generated software.

AI has dramatically reduced the cost of creating software. It has not reduced the cost of owning software.

Those are two very different things.

The Reality:  Creating Software Is Not the Same as Operating Software

Creating version one has become almost trivial. Maintaining version fifty is where the real work begins. Every application accumulates technical debt, operational complexity, security requirements, user management, feature requests, and ongoing maintenance. None of those responsibilities disappear simply because AI wrote the initial code.

This realization has become increasingly relevant as we work with venture capital firms.

For some firms, building internally absolutely makes sense. A younger firm with a relatively small portfolio, straightforward ownership structures, limited reporting requirements, and no formal audit obligations can probably achieve excellent results with Claude, a modern development environment, and a focused internal project. AI has opened opportunities that simply didn’t exist two years ago, and many smaller funds should take advantage of them.

But complexity has a habit of arriving gradually.

Portfolio companies raise additional rounds. Ownership changes. Side letters introduce new governance requirements. Consent rights become more nuanced. Limited partners request increasingly sophisticated reporting. Auditors need supporting evidence. Team members join and leave. New funds are launched. Data begins flowing between multiple systems. What started as a useful internal application slowly becomes operational infrastructure that the entire firm depends on.

At that point, the conversation is no longer about building software. It’s about operating mission-critical systems.

The Difference Between a Tool and Infrastructure

This distinction is why we built PostSig IRI the way we did.

What we learned building PostSig is that the hardest part was never extracting information from documents. AI can increasingly do that. The hard part is creating a trusted intelligence layer where every answer is connected to evidence, permissions, and context.

What matters even more, however, is the foundation beneath those capabilities.

Our LineageAI engine continuously understands the relationships between governing documents, ownership structures, consent rights, reporting obligations, portfolio metrics, and the decisions made throughout the life of an investment. Every answer is traceable back to source documents. Every permission is governed. Every workflow builds upon the same trusted data model rather than another disconnected application.

As firms adopt more AI and see their portfolio evolve, that foundation becomes increasingly important.

Where AI Needs a Trusted Operating Layer

One of the biggest misconceptions in enterprise AI today is that organizations must choose between using general-purpose AI tools and buying enterprise software. I believe the opposite is true.

Investment teams should absolutely continue using Claude, ChatGPT, Cursor, and whatever generation of AI tools comes next. Those systems are becoming extraordinary interfaces for knowledge work, analysis, and productivity. The question isn’t whether AI should be part of the workflow. It already is.

The real question is where AI should retrieve information from, where sensitive data should live, and which system is responsible for governance.

That is precisely why we built PostSig MCP alongside IRI. Rather than replacing the AI tools firms already love, MCP allows those tools to interact with governed information while respecting permissions, document lineage, and organizational controls. AI remains the interface. PostSig becomes the trusted operational layer beneath it.

The Future Is Not Build vs. Buy

I suspect we’ll see this pattern emerge across many categories over the next few years. Every company will build more internal software than ever before because AI makes it economically feasible. At the same time, organizations will become increasingly selective about which systems they choose to own for the long term. Building an application and operating a platform are fundamentally different commitments.

The venture firms we work with are among the most technically sophisticated organizations in the world. If anyone can build an internal portfolio management platform, it’s them. Yet many ultimately decide not to. Not because they lack the capability, but because they’ve concluded that their competitive advantage lies in sourcing exceptional investments, supporting founders, and generating returns, not in maintaining another enterprise software platform.

That distinction is becoming one of the defining strategic decisions of the AI era.

The future isn’t a choice between building and buying.

It’s knowing which capabilities deserve your engineering talent and which are better delivered by platforms that evolve, scale, and improve alongside your business.

AI has made software creation dramatically easier.

It hasn’t made software ownership any simpler.