The promise of AI in buildings is everywhere—predictive maintenance, energy optimization, seamless automation. But here’s the uncomfortable truth: None of it works if your data is trapped in silos, buried in PDFs, or formatted like a 1990s spreadsheet.
In this post, we discuss the real roadblocks to AI-ready buildings, from political data hoarding and vague asset definitions to lease agreements that defy analysis.
1. The Data Accessibility Challenge
A major hurdle in leveraging AI for smart buildings is accessibility. Many organizations struggle with siloed, unstructured, or poorly managed data. As Glenn pointed out, more than half the data needed for AI-driven operations often resides in other people’s systems, making integration a nightmare.
🔹 “The environmental health and safety people aren’t happily sharing their data with the HVAC people, and vice versa.” – Ken
The problem isn’t just technical—it’s political and organizational. Different departments guard their data, creating roadblocks for AI implementation.
2. The Real Estate Data Dilemma
Commercial real estate faces unique challenges:
- Lease agreements are often trapped in PDFs or scanned documents, making structured analysis difficult.
- Asset definitions vary wildly—ask 100 people what an “asset” is, and you’ll get 100 different answers.
- Property tax, insurance, and tenant data are often mismanaged, leading to inefficiencies.
Without clean, structured data, AI tools struggle to provide meaningful insights.
3. The Role of Ontologies and Standardization
To make data AI-ready, experts emphasized the need for:
✅ Standardized ontologies (shared definitions of building assets and systems).
✅ Common Data Environments (CDEs) for seamless data exchange.
✅ Contractual requirements that mandate structured data outputs from the design and construction phases.
🔹 “If the request for structured data is given from the beginning, it shouldn’t cost more—it just needs to be part of the process.” – David
The ISO 19650 BIM standards were highlighted as a potential framework for ensuring data readiness from the outset.
4. AI’s Role in Fixing Messy Data
While the ideal is structured, clean data, AI can help:
🤖 LLMs (Large Language Models) can extract insights from unstructured lease documents.
🛠️ AI-powered data cleansing can normalize disparate datasets.
🔍 Predictive analytics can identify gaps in building performance data.
🔹 “AI can answer questions no human or traditional system could—like calculating the exact weight of energy transfer in a chiller system.” – Mike
5. Who Owns the Data?
A recurring theme was data ownership:
- Building owners may not prioritize long-term data management if they plan to sell.
- Property managers focus on short-term operational needs.
- Tenants have their own data requirements.
🔹 “Everybody has their own ‘central’ data repository, but none of them talk to each other.” – Anto
The solution? Collaborative frameworks where stakeholders agree on data-sharing protocols.
Final Thought: Don’t Let Your AI Get Bored
🔹 “We don’t want bored AI—we need to give it meaningful data to work with.” – Keith
AI is only as good as the data it’s fed. If the smart buildings industry wants to harness AI’s full potential, data readiness must be a priority.
What’s Next?
The discussion continues next week, diving deeper into ontologies and semantic normalization. Stay tuned!
📌 Access the slides: MondayLive.org
What’s your biggest data challenge in smart buildings?