10 Questions Before Deploying Building AI

Title: 10 Essential Questions to Consider Before Deploying Autonomous AI in Your Building

Summary: 10 Essential Questions to Consider Before You Deploy Autonomous AI in Your Building Management

Full Content: 10 Essential Questions to Consider Before You Deploy Autonomous AI in Your Building Management

By James C. Waddell, President, Cognitive Corp

The building technology market has reached a pivotal moment. With VergeSense unveiling their "orchestrated intelligence" for workplace AI, Facilio releasing named AI agents for facility operations, and Willow claiming the title of "#1 Operational AI platform," the landscape is evolving rapidly. However, research from Nuvolo indicates that many organizations are not realizing significant value from their AI investments.

As countless vendors roll out AI agents, the critical question surfaces—not "can you deploy them?" but rather "should you deploy them?"

A checklist exists that is unfortunately overlooked by many: the lack of a structured governance framework is why most facility AI deployments end up amplifying risks rather than mitigating them. After examining eight leading autonomous building AI vendors, I've found that not a single one has a formal governance framework, mandates a rigorous testing process for their AI agents, or provides transparent decision trails sufficient for compliance audits.

Before you proceed with any contracts, it’s imperative to ask these 10 questions. If your vendor can't provide satisfactory answers, you might just be looking at a demo rather than a robust enterprise solution.

THE 10-QUESTION GOVERNANCE CHECKLIST

1. Can you show me the audit trail for any autonomous decision this system has made?

Not just a dashboard or log file—a clear, human-readable explanation of the rationale behind any specific autonomous decision. For example, if the system adjusted your HVAC setpoints at 2 AM last Tuesday, you should be able to access the reasoning within five minutes.

Why it matters: Compliance with Building Performance Standards (BPS), LL97 audits, and Environmental, Social, and Governance (ESG) reporting requires clear explainability. Simply stating, "the AI decided," is insufficient for regulatory scrutiny.

2. What testing protocols does an AI agent undergo prior to operational approval?

There should be a comprehensive testing protocol in place—not merely a Quality Assurance checklist, but a set of scenarios to identify potential failures. How does the AI respond under conflicting objectives or high-pressure situations with irreversible consequences?

Why it matters: If an agent cannot demonstrate safe behavior in critical situations, it shouldn’t have permission to control your building systems.

3. Who is held accountable if the AI makes a detrimental decision?

It is essential to delineate accountability within the governance structure. Who carries the responsibility, and what is the escalation path?

Why it matters: Accountability relies not on technology but on strategic organizational design and process.

4. Which decisions are made autonomously by the AI, and which need human oversight?

This should be laid out clearly in a documented matrix rather than mere default settings. High-impact decisions must require human approval, while routine operations can be automated.

Why it matters: While "fully autonomous" may appear efficient, it can lead to a lack of oversight on significant risk-generating decisions.

5. How does the system navigate conflicting objectives?

Conflicts are inevitable—such as balancing energy efficiency against tenant comfort. How does the AI manage these trade-offs? Does it have a protocol for escalation beyond its decision-making authority?

Why it matters: AI is programmed to optimize based on its instructions. Without a defined priority system, the algorithm may not align with your organizational goals.

6. Can you demonstrate that the AI’s decisions are free from bias?

The AI impacts various aspects, from space allocation to maintenance scheduling. Are decisions equitable in resource distribution? Are recommendations fair for all tenants?

Why it matters: Addressing algorithmic bias is crucial to prevent unfair impacts on tenants, employees, and the community. Bias, once established, is complex and costly to reverse.

7. What protocols exist when the AI encounters untrained situations?

Edge cases—like new equipment failures, extreme weather, or unexpected occupancy patterns—are unavoidable. Does the AI fail gracefully? Is there an escalation policy, or does it revert to safety protocols?

Why it matters: The most severe AI failures often stem from confidently incorrect actions—those that go unchecked because they seem correct at first.

8. How do you adapt the AI’s decisions to align with evolving policies?

Regulations change, BPS standards tighten, and tenant expectations shift. How does the AI accommodate these changes? Is there a governance structure that can be updated promptly without retraining the model?

Why it matters: Governance is not a one-time activity; it demands constant adaptation. If policy updates necessitate vendor involvement, the governance is effectively out of your control.

9. Can your team explain the system’s actions to non-technical stakeholders?

Communication is key. Can board members, REIT investors, tenant committees, or city council members grasp what the AI is doing and the rationale behind those actions? If explanation requires technical jargon, there’s a transparency issue.

Why it matters: Trust from stakeholders hinges on clarity—if your team can’t articulate the system’s functioning, trust erodes.

10. What contingency plans are in place for AI failures?

It's not a matter of if, but when the system may fail. Are there protocols for rollback, manual overrides, notifications, and business continuity strategies should an AI system fail? Have these plans been rigorously tested?

Why it matters: Building resilience involves preparing for failures. Failing to account for potential breakdowns can magnify the impact of failure, making recovery challenging.

The Pattern Nobody Wants to Admit

These ten questions reveal a sobering truth: many AI deployments in facilities prioritize technology without considering governance. Vendors promote capabilities while buyers deploy without questioning the governance surrounding those decisions.

This oversight is not due to vendor negligence; the market simply hasn’t required governance transparency. Buyers inquire about functionality but neglect to ask who is governing decision-making processes. Until this changes, every deployment of autonomous building AI is merely an ungoverned experiment within live environments.

Successful deployment of autonomous building AI hinges on robust governance, not just superior technology.

Technology devoid of governance becomes a liability, while governance without technology is merely policy. However, when technology meets governance, it breeds trustworthy autonomy—this is the only kind truly worth deploying at scale.

What Governed Autonomy Looks Like

At Cognitive Corp, we created the Building Constitution and CST-1 testing framework out of a commitment to ensuring that facility AI is trustworthy, transparent, and properly governed. In a landscape where AI can make thousands of daily decisions across portfolios, it is essential that every action handled by AI adheres to standards our teams, boards, and regulators can comprehend.

If you’re in the process of evaluating autonomous AI for your buildings, begin with these ten essential questions. If your current vendor can address all of them satisfactorily, you are on solid ground. If not, it’s time to pivot and prioritize governance in your deployment strategy.

Downloadable PDF

For a comprehensive guide, download our PDF for easy reference. This resource will help you lead conversations around governance in autonomous AI and ensure that every aspect of your deployment is covered.

Schedule your 15-minute Governance Gap Assessment:

james.waddell@cognitivewx.info

James C. Waddell is the President of Cognitive Corp, the creator of the Building Constitution and CST-1 governance framework, and a board member of the IFMA ITC.