Trust in the Era of Agentic AI - How LOFTx.ai Secures Your Enterprise Data
Narasimhan Vallur | 2026-06-30
Trust in the Era of Agentic AI: How LOFTx.ai Secures Your Enterprise Data
Agentic AI is reshaping how real estate and multifamily operators work. It covers back-office workflows, revenue intelligence and 24/7 leasing assistants that answer prospects in their own language. To do that, AI agents read a great deal of your data: leases, ledgers, vacancy, and the names and phone numbers of the people who live in and inquire about your properties.
Handing that work to AI takes trust. Property management data includes resident leasing history, corporate accounting and live vacancy, and it needs real protection.
At LOFTx.ai, we treat security as part of the product, not an add-on. This post explains how we protect your data. It covers the security framework we build on, the Microsoft Azure infrastructure we run on, and how we keep your residents' personal information out of public AI models.
1. Built on the SOC 2 Trust Services Criteria
SOC 2 is the security framework from the American Institute of Certified Public Accountants (AICPA). Its Trust Services Criteria describe what a service provider should do to protect the customer data it handles. We design our security controls, operating procedures and data handling around those criteria. We hold ourselves to them as our own standard, and we review our controls against them as the platform grows.
The criteria cover five areas.
🛡️ Security
Protecting the platform against unauthorized access.
- Organization-level isolation: Every request is checked against the organization that made it. Your team sees your portfolio and nothing else. Another customer's users can never read your rows, and yours can never read theirs.
- Role and page-based access: Access is granted by role (property manager, admin, corporate accounting) and by the modules your organization has turned on. A user who does not have a module cannot reach its data through the app, the API or an AI assistant.
- Revocable access tokens: API and AI-assistant connections use tokens issued per user. If a token is revoked, or your organization turns API access off, cached answers stop being served to it as well.
- Secrets kept out of the browser: Integration credentials, such as your PMS, CRM and AI service keys, stay on our servers and are never sent to your team's browsers or the chat widget.
🔄 Processing Integrity
When AI works with your leasing workflows or financial data, the results have to be accurate and complete.
- Numbers from the source: Reports are built from your PMS and accounting data. Totals are checked against the statements they come from, and a figure that does not reconcile is flagged, not hidden.
- Clear labels: AI-generated drafts, estimates and recommendations are labeled as such. Asking rent is kept separate from effective rent, and potential is kept separate from actual.
- Read-only by default: Our AI tools read your data. Anything that would change a record goes through explicit, permission-checked actions.
🔒 Confidentiality
Keeping data that is sensitive or proprietary out of unauthorized hands.
- Your data stays yours: Your portfolio, financials and conversation history are kept separate from every other customer's.
- Need-to-know internally: Our team accesses customer data only to support you, and only as far as the task requires.
🌐 Availability
An AI leasing agent is only useful if it is online when prospects reach out.
- Cloud resilience: The platform runs on Microsoft Azure's managed infrastructure.
- Backups and recovery: Data is backed up so that service can be restored after an incident.
- Dedicated capacity for time-sensitive work: Messages to and from prospects run on their own processing capacity. A large background job can't hold up a reply.
👥 Privacy
How personal information, such as applicant and resident names, phone numbers and email addresses, is collected, used and kept.
- Data minimization: Each AI task receives only the data it needs. A leasing assistant answering a prospect sees that prospect's conversation and the property they asked about, not your rent roll.
- Purpose limitation: Personal information is used to run your leasing and operations, not to market to your residents or for anyone else's benefit.
- Retention: We keep data for as long as your service needs it, and handle deletion requests according to your instructions and applicable law.
2. Hosted on Microsoft Azure
LOFTx.ai runs on Microsoft Azure, which provides the physical and network security beneath our platform:
- US data residency: Your database, our application servers and the Azure OpenAI service our AI runs on are in US-based Azure data centers. Texts, calls and email go through specialist providers (sub-processors), described in our Privacy Policy, who may use the data only to deliver those services. A current list is available on request.
- Encryption in transit and at rest: Data moving between your browser, our services and your systems is encrypted in transit with TLS. Azure encrypts the data we store at rest.
- Managed, patched infrastructure: Azure's data centers, physical security and core infrastructure are run and maintained by Microsoft. Our team can focus on securing the application layer.
- Activity history: Changes to your configuration and AI-driven actions are logged with who made them and when. Your team can see what happened and why.
- Third-party integrations: Your PMS API keys and passwords are encrypted before they are stored, with encryption keys held in Azure Key Vault, on top of the database's own encryption at rest. Only our servers decrypt them, and only to sync your data. Your team can update or revoke them at any time.
3. Keeping Your Residents' Personal Data Out of Public AI Models
The question we hear most is simple: "If I use your AI, does my residents' information end up in ChatGPT, or training somebody's model?"
No. Here is how we make sure of it.
Enterprise AI inside Azure, not consumer chatbots
LOFTx.ai's AI runs on Azure OpenAI Service, Microsoft's enterprise AI service, under our Azure agreement. That is a different service from consumer chatbots:
- Not used for training: Under Microsoft's Azure OpenAI terms, the prompts we send and the responses we receive are not used to train or improve the underlying foundation models.
- Not shared: Your data is not available to other Azure customers or to the public versions of these models.
- Processed within Azure, in the US: Your requests are handled by Microsoft's enterprise service in US Azure regions, not by a consumer app where conversations may be kept and reused.
Only the data the task needs
We don't send your database to an AI model. Each request carries only what that task needs:
- A leasing reply includes the prospect's own messages and the listing they asked about.
- A market report includes the unit, its comparables and public market data.
- A financial summary uses figures we have already computed on our own servers. Large statements are summarized before any AI sees them, not passed over in full.
The less data that travels, the less there is to expose.
Your organization's walls apply to AI too
The AI works under the same rules as your users. Every AI request runs as a specific user in your organization, with that user's role and enabled modules. An assistant cannot reach a property, report or resident record that the person using it could not open in the app.
When your team connects its own AI assistant
LOFTx.ai offers a secure connector, using the Model Context Protocol (MCP), so your team can query portfolio data from an AI assistant your organization uses. That connection is in your control:
- It uses a personal access token that your admin can issue and revoke at any time. Organizations can switch API access off entirely.
- Data goes only to the assistant your organization has chosen, under your own account and that provider's terms with you.
- The connector returns only what that user is allowed to see.
No PII in places it doesn't belong
- Integration secrets and personal data are never sent to the chat widget on your website.
- Your residents' information is not used to build shared datasets, benchmarks or models for other customers.
Your Data Belongs to You
LOFTx.ai sits on top of your business systems. It draws out the insights your team needs to run leasing and operations, and it does that without giving up control of your data. Whether our AI agents are researching unit-level rents, pricing concessions or building financial dashboards for your accountants, your data stays yours.
As we keep building agentic AI for multifamily operations, our commitment to security and privacy doesn't change. With LOFTx.ai, you don't have to choose between efficiency and peace of mind.
Want to see how our secure AI agents can drive a 25% increase in your net rental income safely? Book a 30-minute demo with the LOFTx.ai team today.
A public website example and the information boundary
Redside Partners is a company website with LOFTx chat installed. It provides a useful public-facing reference when thinking about the distinction between property questions and private account information. The installation itself is not evidence of a security certification or the customer's internal controls; our security overview describes LOFTx's practices.