From AI Agents to Institutional Memory: What's Next for Real Estate

AI is already changing how real estate investment teams work. Accelerating research, data analysis from leases and OMs, supporting underwriting, and helping teams turn complex information into investment decisions.

But the next question is much bigger:

At the upcoming PERE London Forum, we’ll be exploring this question in our panel, “AI in Alternative Asset Management: Turning Investment Edge into Agentic Memory.”

For real estate, the opportunity goes well beyond automating individual tasks.

An investment decision rarely depends on one data point. It draws on years of experience: how a team thinks about rent growth, tenant risk, leverage, capex, liquidity, market cycles and downside scenarios. Much of that knowledge sits across old underwriting models, IC memos, deal histories and in the heads of experienced investors.

What happens when AI can retain that context?

Imagine evaluating a new acquisition and asking not only:

“What does the data tell us about this asset?”

but:

“How would our investment team underwrite this deal, based on what we have learned from similar investments?”

That shift — from AI as a copilot to AI as an institutional memory layer — could fundamentally change how real estate firms source, underwrite, manage and ultimately exit investments.

From underwriting to asset management

The real opportunity spans the entire investment lifecycle.

AI agents can help analyze thousands of pages of diligence, reconcile rent rolls and financials, identify lease and tenant risks, challenge underwriting assumptions and prepare investment committee materials.

But the more interesting application is what happens after the acquisition.

An intelligent system could continuously compare actual asset performance against the original investment thesis, identify emerging risks and draw on historical portfolio experience to help explain what is happening — and what the team should consider doing next.

The asset becomes part of a growing institutional memory.

The production question

Of course, none of this matters if the AI cannot be trusted.

For investment managers, “good enough” cannot mean that an answer sounds plausible. AI needs to demonstrate that it can retrieve the right information, perform accurate financial reasoning, distinguish facts from assumptions, cite its sources and apply the firm's own investment framework consistently.

This is where Evals become increasingly important.

At Terminal X, we believe investment firms need to evaluate AI against the questions and workflows that actually matter to their business — not generic benchmarks.

• Can the system find the right precedent deal?
• Can it identify the assumption that is driving downside risk?
• Can it explain why a previous investment was rejected?
• Can it reconcile conflicting information across a data room?
• And can it demonstrate its reasoning well enough for an investment professional to trust the output?

The bigger opportunity

Real estate has never suffered from a lack of information. The challenge has been turning institutional knowledge into institutional infrastructure.

At Terminal X, we’re building toward an agentic operating layer for investment managers — connecting proprietary data, investment workflows, AI agents and institutional memory so that every investment decision can make the next one smarter.

That’s the conversation we’re looking forward to having in London.

Panel: Turning Investment Edge into Agentic Memory - AI in Real Estate Investment Management

We’ll explore:

• Where AI is creating real value across the real estate investment lifecycle
• How AI can move from copilot to institutional memory
• What it takes to evaluate and deploy AI safely in investment workflows
• How domain-specific Evals can help firms move from AI experimentation to production

If your firm's investment edge lives in the experience of its people, how do you make sure your AI remembers it?