Where the Analyst Bottleneck Shifts

A piece circulating widely right now makes a striking argument: we're in the "this seems overblown" phase of an AI shift far larger than most people realize. The author of "Something Big is Happening", an AI founder, describes watching his own job transform in real time. The manual work is now handled autonomously, and the models doing it are improving faster than he can track.

I try not to index heavily on the AI hype and fear mongering we see online. But the magnitude of change didn't fully click for me until I incorporated AI into my own engineering work about a year ago. The job of software engineering is changed forever. His warning is simple: what happened to software engineers is coming for every knowledge worker, sooner than they expect. The author is right that AI is making its way into CRE, Finance, and Legal by eliminating an enormous amount of the manual, analytical work. But there's an additional layer to this story: it's not just about tasks being automated. It's about where the value shifts next, and whether teams have the infrastructure to capture it. The analyst role in CRE has always had two sides.

"Job one" All of the manual work required to produce that analysis: underwriting, deck creation, PDF-to-Excel conversions, rent roll analyses, market research, etc.

"Job two" Distilling market and deal intelligence into something leadership can act on: pipeline reviews, investor reports, deal activity summaries, administrative todo lists to close deals.

General AI is solving job one. The largest labs are post-training models on finance and real estate specific tasks, partnering with vertical AI companies, and hiring investment bankers and real estate professionals to build accurate benchmarks. Teams can now feed massive context of contracts, spreadsheets, and decks, into these models and offload a significant portion of the work. Manual, low-fidelity tasks like converting PDFs to Excel, abstracting leases, cleaning rent rolls, and drafting initial market research are already largely handled by general models. Many analysts have quietly built their own workflows around them.

At institutional firms we are seeing meaningful adoption of general chat tools to help analysts write memos, draft emails, conduct market research, and extract data from rent rolls. What we are not seeing yet is broad adoption of fully embedded AI systems. The hesitation often comes down to unclear AI strategy, security concerns, and uncertainty around ownership of outputs. Education gaps and the challenge of finding or building AI systems that match a team's specific process have slowed deeper integration.

Mid-market firms have moved faster. Early adopters are redesigning workflows entirely. Teams are using browser agents to generate leads directly into a CRM based on prompts like "prospect owners of self storage facilities in upstate New York that purchased in 2021." Asset managers are connecting accounting and asset management software into products like base44 to build live portfolio analytics. Some are forwarding offering memorandums and rent rolls directly into Manus AI and receiving preliminary underwriting in minutes. Others are running deep market research that combines new development pipelines, demographics, tax appraisals, and surrounding parcel data into a cohesive view of a submarket.

Four months ago we tried underwriting with an early Excel agent. It struggled. A few weeks ago we underwrote a property using Claude's Excel agent. The difference was material. It may not be perfect, but the rate of change is undeniable. The market has noticed. Large CRE firms have become more cautious about hiring, and some smaller firms have paused analyst hiring entirely. Macro uncertainty is part of the story, but the more structural explanation is that AI is absorbing the tasks that used to consume most of an analyst's day.

This does not mean analysts are going away. It means the bottleneck is shifting.

As job one gets automated (we are not 100% of the way there yet), job two becomes the work. The value is no longer in how many deals a team can underwrite or how many memos they can produce, it is in how well a team can organize, synthesize, and act on the market intelligence available to them. Teams that do this well can spend more time with brokers, capital partners, and owners in markets they previously had no bandwidth to explore. A small shop that was heads-down underwriting can spend more time raising capital. An emerging manager can develop a credible thesis in a new market without tripling their headcount.

What does this shift actually mean in practice?

The constraint is no longer analysis capacity. It is data coordination and truth management.

For senior leadership, that means having a clear and consistent view of deal progress and portfolio exposure. The old analyst role translated fragmented information across emails, models, broker updates, and internal notes. Now leadership needs a system that tracks deal history, surfaces risks, and reconciles competing versions of truth. Market numbers, underwritten assumptions, broker-reported figures, and survey data rarely align perfectly. An LLM can process them, but someone still has to define what is credible and decision-ready.

For asset managers, the same constraint shows up differently. It means aggregating and standardizing portfolio data across P&Ls and general ledgers in real time, benchmarking line items across properties, and continuously ingesting financials from disconnected systems that export reports in inconsistent structures. The analytical work is becoming automated. The orchestration, normalization, and storage of the data is not.

In both cases, the problem is not generating insight. It is maintaining a reliable source of truth that AI systems can operate on.

The marginal cost of analysis is falling.

The marginal cost of maintaining truth is not.