Over the past year, we’ve had a front-row seat as AI has transformed the way sponsor firms do business. Across our discussions with many of these firms, we’ve seen a pattern emerging in how these firms are actually using AI in their day-to-day work.
These sponsor firms have selected their preferred AI tools, written highly effective prompts, and shipped workflows. This might include a principal using it to draft an LP letter, a senior associate building a Custom Skill for the team’s investment memo template, or an ops lead running their own AI project for quarterly reporting prep. These small acts of building are happening every day at firms of all sizes.
But beneath those workflows lies the same friction: Their carefully selected tools and perfected prompts know nothing about the fund. They don’t know which LPs are in which deal, which capital calls landed last week, or which distributions are scheduled next month. The work happens in the AI tool with none of that data on board beyond what has been copied in.
The question across sponsor firms today isn’t whether to use AI. It’s how to close the gap between what the team is already building at the desktop and the data that should be flowing seamlessly through it.
The Big Firm Playbook
It’s worth looking at what the firms further along with AI are doing right now, not as a template but to clarify how scale changes the conversation:
- Goldman Sachs embedded Anthropic engineers inside their tech teams for about six months to build accounting, reconciliation, and compliance agents.
- BlackRock has more than 50 engineering teams contributing plugins to its Aladdin Copilot, orchestrated through open frameworks.
- Norges Bank Investment Management, with 700 staff managing $2.2 trillion, built a Claude-based agent that screens every new equity holding within 24 hours for forced labor, corruption, and fraud links.
A sponsor firm may not have the headcount, internal data platforms, model-evaluation infrastructure, and engineering teams required to build at that level. But we’ve seen across the sponsor segment that smaller firms are building, too:
- They’re crafting their own Skills.
- They’re writing prompts codifying how the firm thinks about deals.
- They’re setting up projects that hold institutional knowledge that a well-prompted AI can draw from.
While these sponsors may not have the same capabilities as the giants with their extensive resources, the bottom line is that they, too, are taking the initiative to author their own AI tooling.
The gap that matters here isn’t between large and small firms or even between firms that build and firms that don’t. It’s between AI tooling that knows the firm’s data and AI tooling that doesn’t.
How Are the Platforms Developing AI?
It’s logical to expect the sponsor’s investor management platform to solve this disconnect between the AI tool and the data. And in fact, every CRE software vendor in the category is developing AI. But how they are developing it is important.
Some platforms will ship features unilaterally with unpredictable release dates, prescribed workflows, and a broad footprint that covers the common use cases of their customers. Some sponsors will find those features useful while others will find that they can’t be customized to truly fit what they’re already building. For LP-facing data, these platforms will ask sponsors to extend a great deal of trust from the start, before working out the structure of that trust.
Other platforms will start with the connector, the read-only structure that earns the right to widen. They’ll publish the access sponsors need to build their own custom workflows on top. And they’ll form partnerships with AI-forward sponsors that need real-time data, write access, and tight platform integration to protect their LP relationships.
The first path produces useful features while the second provides a platform sponsors can build on. The question for sponsors is which platform they want behind their LP data.
As of right now, only two of the legacy investor management and real estate ERPs have shipped the native Model Context Protocol exposure capable of offering the second path. The decisions made over the next year will define what successful tools look like, and the sponsors who help shape those decisions will be rewarded with tools that fit the workflow that works best for them.
The Three-Layer Architecture
When looking for what separates the sponsor firms that are pulling ahead on AI from the rest, something stands out: a three-layer architecture. Each of these layers is critical because each reinforces the next.
The first layer is connection. The AI tool the team has chosen must be connected to the platform data running the firm. For most of the platforms shipping into the space today, that connection is read-only right now. And for LP-facing data, that makes sense. It lets the team use the AI they’ve standardized on with the data the platform already governs, without taking a position of trust that has yet to be earned.
The second layer is sponsor-built. The AI projects the team has already built get richer with that first layer—connection—in place. The team no longer needs to feed their LPs’ commitment history into the AI. It’s already visible. The work the team has been doing at the desktop deepens, with no AI engineer or new platform feature rollout needed.
The third layer is partnership. Some agents are too deep, too high-stakes, or coupled too tightly to platform data for a sponsor to build alone. In this layer, capital call generation, distribution notice review, and IC memo synthesis from underwriting data get built in collaboration. While the platform brings the integration depth, the sponsor brings their established workflow, guardrails, and human patterns required to earn and keep the trust of their LPs.
This three-layer architecture applies across sponsor sizes, but it’s most acutely felt in the underserved middle, with sponsors managing around $10 million to $500 million of investor equity.
These firms have AI fluency at the desktop and strong prompt writing within the team. The problem is that they’re limited by the data access gap between their AI tools and their platform. These firms are perfectly capable of building and poised to benefit from the three-layer pattern, but they’re operating in a category where that first crucial connector layer is, as of right now, largely absent.
The Three-Layer Architecture in Practice
Of course, the real story plays out in day-to-day sponsor firm activity. It’s in these everyday moments that it’s easiest to see the true impact of the three-layer architecture.
A sponsor principal has a morning call with one of her largest LPs. Twenty minutes beforehand, she runs the prompt her team built for investor prep. She’s been refining it for months. And this time, the prompt pulls directly from the LP’s actual commitment history, recent activity, and capital position. She’s ready for the call.
A senior associate has been maintaining a Custom Skill for quarterly LP reporting. He kicks off the reporting cycle, and the Skill pulls performance data, occupancy detail, and distribution history from the platform of record. Rather than spending the morning assembling inputs from a pasted PDF, the associate reviews the narrative and ships it.
A team drafts a capital call notice using a workflow they built around their preferred AI tool, and the workflow knows the exact commitment balance, prior calls, and LP-specific terms for every investor on the call. This is the kind of workflow that would eventually become an agent: a built-in partnership between the sponsor and platform, with the guardrails and approval flow defined by the sponsor to ensure they get it right and maintain LP trust.
One platform building toward this architecture is InvestNext. They offer a read-only Model Context Protocol connector today, and the agent is layer being shaped collaboratively with the Early Access cohort. The AI tool a sponsor connects today isn’t a bet on the AI tool they’ll use tomorrow, so InvestNext offers open architecture by design.
Looking Ahead
The sponsor firms we see pulling ahead on AI right now aren’t the ones with the biggest engineering budgets. They’re the ones connecting the AI tools their teams already use to the platform data that runs their firm, building their own custom agents at their own scale, and working with the platforms on the agents that require deep integration.
InvestNext is currently running a small early access cohort to build the agent layer of this architecture in partnership with the sponsors using it. Meanwhile, sponsor firms planning their architecture right now may want to consider how their own everyday scenarios would play out with a three-layer structure in place.
We’re at a turning point where the architecture decisions sponsors and platforms make this year will define what AI in CRE investor management will look like at sponsor scale for the rest of this decade.
