AI & MCP

ESPM

Compliance

GRESB

Partners

Book a Demo

AI & MCP

ESPM

Compliance

GRESB

Partners

Book a Demo

AI Strategies for Real Estate Data Management

As AI transforms industries, real estate owners, investors and service providers are racing to find the best ways to unlock new insights and drive value across their operations. The options are many, and each reflects a different philosophy about where data lives, who controls the AI and how far its output can be trusted. In a landscape this fast-moving, understanding those strategies matters as much as choosing a tool.

Below, we compare three common approaches on flexibility, control, security, accuracy and effort.

AI Strategies

#1: Point Solution w/ External AI

#2: Point Solutions with Integrated

#3: Point Solutions with MCP

Strategy #1: Point Solutions with External AI

How it works: Users export static data from their point solutions, upload it into an AI client such as Claude, ChatGPT or Copilot, and direct it to complete a task. For example, a user might upload three years of energy consumption data and ask the AI to flag data quality issues and performance anomalies, compare the building against an industry benchmark or draft a decarbonization plan.

Strengths: This is a highly flexible approach. Users can upload any number of files in any format (Excel, PDF, Word), ask any question, and decide exactly what data the AI sees and how much they rely on it. Combining files from different sources also supports multi-dimensional analysis: climate and physical risk data, energy and emissions data, and building audits can be merged into a forward-looking financial risk analysis or due diligence report. Paired with an IT-approved enterprise AI client, this approach also protects against data leakage and keeps users in control of token use and AI spend.

Weaknesses: The AI only knows what is in the uploaded files, so it must fill in the gaps. That is where errors, misinterpretation and hallucinations creep in.

• Missing reference data. AI must fill in gaps from what it finds online and, if several versions exist, decide which to use. The AI may mistakenly choose stale or incorrect reference data (e.g. benchmarks, emissions factors), resulting in incorrect and/or inconsistent analysis.

• Missing context. Complex tasks need information the files rarely contain. For example, asked to disaggregate Scope 1, 2 and 3 emissions, the AI may arbitrarily assign meters to space types and misreport the results.

• Stale data. As users upload new versions of files, the AI can mix contexts and base its insights on outdated or incorrect data.

• Source-specific rules. Every data source has its own conventions. ENERGY STAR Portfolio Manager, for example, separates in-metric meters, which count toward whole-building metrics, from out-of-metric submeters, which do not. The AI must be taught each rule for every source, or data is missed or double-counted.

Bottom line: Before any analysis can begin, the data still has to be collected, which requires additional solutions and a missed opportunity for maximizing efficiencies. This approach provides great flexibility but with the trade-off of a high risk of error and significant manual effort.

Strategy #2: Point Solutions with Integrated AI

How it works: Many software vendors now build AI directly into their platforms, as chat assistants, automated insights or AI-powered features that run on the vendor's own data and the vendor's chosen model.

Strengths: Because the AI works directly on the platform's structured data, it sees the full scope of that data and the context needed to interpret it, such as: which meters count toward which metrics, which emissions factors apply, how properties roll up into portfolios. The vendor has already encoded these rules, so the risk of hallucination is low, and there is nothing to export or upload.

Weaknesses: Users can only leverage AI in the context provided by each individual vendor, precluding integration across data sources.

• Limited flexibility. Users are confined to the features the vendor provides, and the AI only sees that one platform's data.

• Limited accessibility. Users can only access their sustainability data within the vendor solution rather than easily querying their data across all devices, including mobile.

• Limited control over tokens or cost. The vendor chooses the model, prompts and context. AI costs may be bundled into licenses or billed as usage-based add-ons, which makes spend hard to predict or manage across vendors.

• Fragmented security. Each platform with built-in AI sends data to its own AI sub-processors, under its own terms for retention and model training. Every one becomes another AI vendor for IT to vet, outside the enterprise AI client and governance the organization has already approved. As tools multiply, so does the exposure.

Bottom line: Reliable answers within a single platform, at the cost of flexibility, control and security oversight. However, if specific functions are not implemented well or if too many constraints are added, this strategy can be worse than the first.

Strategy #3: Point Solutions with an MCP Connector

How it works: The Model Context Protocol (MCP) is an open standard that lets an AI client query an external platform directly. Instead of exporting files or relying on a vendor's built-in AI, users connect their organization's approved AI client to any compatible point solution, and the AI looks up live data as their questions require.

Strengths: MCP combines the best of the first two strategies.

• The flexibility of external AI. Users stay in the AI client they already know, ask any question, and combine live platform data with their own files, other connectors and analysis.

• The accuracy of integrated data. The AI pulls current, structured data with its context fully intact. This avoids stale data and the need to train AI around the context of each data source.

• Rich supporting information. The MCP can connect to user and technical documentation as well as live, validated platform data. Users can ask the AI to explain what they see and get detailed answers specific to their own data, without any risk of misinterpretation.

• Enterprise security and control. Data is encrypted and stays within the organization's approved AI environment, under its existing governance, token limits and spend controls. IT teams can easily monitor and control access as needed.

• Minimal manual effort. No exports, uploads or reformatting.

• Ease of access. Query the MCP on desktop, mobile and other devices.

With the MCP approach, one AI client can connect to several solutions at once, allowing clients to integrate their preferred sustainability platform, climate risk provider, internal data warehouse and other data sources. The AI can securely query, combine and analyze live data from all of them, supporting complex analyses across data sources. With MCP, users can also build custom dashboards, reports and agents that refresh with live data and can be shared with other stakeholders.

Bottom line: The flexibility of external AI with the security, control, accuracy and memory of integrated solutions. Users have the freedom to refine queries, customize functions and their own quality thresholds.”

AI Strategies: Direct Comparison

For most organizations, the question is no longer whether to use AI, but how to give it data it can trust without giving up control. External AI offers freedom without reliable data; integrated AI offers reliable data without freedom. MCP connectors deliver both, inside the AI platform your organization already governs.

Dimension

#1: External AI

#2: Integrated AI

#3: MCP Connector

Flexibility

High

Low

High

Cost Control

High

Low

High

Security

Medium

Low

High

Data Scope

Low

High

Accessibility

Low

High

Error Risk

High

Medium

Low

Manual Effort

High

Low

Value to Organization

Limited

High

How Abisko does it

Abisko offers the industry’s most comprehensive MCP connector, giving clients access to all of their sustainability data directly within their enterprise AI platforms. Abisko’s MCP connector is:

• Secure. All data is encrypted and read-only.

• Private. Abisko never receives the conversation, the AI's memory or its prompts.

• Accessible. Users can query their data via desktop, mobile and other devices. Empowering property teams with data at their fingertips.

• Permission-bound. Users sign in via Abisko and only see the data they can already access.

• Admin-controlled. Users or organization administrators can disconnect it instantly.

• Extensible. New queries and analytics can easily be added based on client needs.

• Customizable. Users can create custom dashboards, report and agents with live data.

• Comprehensive. Users can query all of their sustainability data, including: portfolios, properties, meters, consumptions, emissions, certifications, measures, assessments, compliance requirements, and much more.

Setup takes about a minute. From there, users can ask for anything from a detailed capital plan to a simple list of expiring certifications across their portfolio.