The Multi-Tenant Governance Problem: When 47 Tenants Share One Ungoverned AI

The Multi-Stakeholder Governance Problem: When 47 Tenants Share One Ungoverned AI

Your property accommodates numerous tenants. Your AI is designed to optimize its function. But what happens when these interests conflict?

---

The Conflict Nobody Configured For

Imagine a Class A office tower in the heart of Manhattan, hosting 47 tenants across 52 floors. The building's AI-driven systems operate under a single optimization model, functionally balancing energy costs, comfort levels, and equipment longevity.

On a hot afternoon in July, the AI must make critical decisions. With outdoor temperatures soaring to 98°F, grid demands peak and energy prices skyrocket 340% above baseline. The optimization algorithm jumps into action, appropriately reducing the energy load.

Yet, this decision comes at a cost. Floor 31’s temperature rises by 4°F during this critical juncture. This floor houses a law firm embroiled in a deposition with 14 people crammed into a conference room meant for 10. Meanwhile, Floor 38 holds a private equity fund whose managing partner had already voiced temperature complaints to the building manager last week.

No bias was programmed into the AI. It merely learned from historical override data showing that the system was trained to prioritize comfort for the fund on Floor 38 due to consistent adjustments made by the building manager. Meanwhile, the law firm on Floor 31 remains unaware of this underlying priority shift — all they see is their conference room at an uncomfortable 78°F while paying a premium of $89 per square foot.

This scenario illustrates the multi-stakeholder governance problem, prevalent in multi-tenant properties where AI makes pivotal allocation decisions.

Why Multi-Stakeholder Structures Complicate AI Governance

While single-tenant buildings grapple with governance challenges, multi-tenant environments face governance conflicts. The distinction is not merely incremental; it’s structural.

Competing optimization objectives. A single-tenant building aligns its focus with one organization’s specific needs. In contrast, multi-tenant environments must navigate various objectives: some tenants prioritize maximum cooling, others focus on cost savings, while medical tenants have specific regulatory ventilation necessities, and data rooms prioritize equipment cooling above all else. Achieving simultaneous optimization for all becomes impossible. Hence, someone must be deprioritized. The pressing question is: according to what guidelines, devised by whom, and recorded where?

Hidden allocation decisions. Every optimization represents an allocation. The AI’s choice to truncate cooling during peak pricing leads to discomfort shared among tenants. Each time maintenance is scheduled or elevator dispatch algorithms are altered, disruptions and wait times are allocated invisibly, without tenant awareness or a governing framework ensuring fairness.

Bias operationalized through training data. The situation on Floor 31 versus Floor 38 isn’t merely an anomaly — it’s a byproduct of the AI absorbing training patterns from historical management decisions. If the building manager historically favored response times for anchor tenants, the AI replicates this bias, prioritizing certain floors based on previous override data. The AI doesn’t create biases; it amplifies existing systemic inequities at scale, often rendering them invisible.

Silence in lease agreements. Standard commercial leases fail to mention "artificial intelligence," "algorithmic," or "optimization." They outline service hours, maintenance obligations, and temperature ranges without addressing how AI allocates resources within those parameters. If a tenant consistently receives the low end of the acceptable temperature range, they possess no recourse as the lease lacks any mention of AI-driven allocation mechanisms.

Privacy in shared systems. Increasingly, multi-tenant environments utilize occupancy analytics and access control data to inform AI decisions. However, this information crosses tenant boundaries without consent. If the AI discovers that Tenant A’s space becomes vacant at 4 PM every Friday and leverages this pattern to pre-condition Tenant B’s space for an afternoon event, Tenant A’s behavior is essentially exploited without acknowledgment or authorization, placing tenant privacy at risk.

The Scale of the Problem

The stakes are substantial:

Organizations like BXP (formerly Boston Properties) oversee 54 million square feet across 190+ properties, mainly in multi-tenant Class A office categories. Brookfield Asset Management’s real estate portfolio exceeds $250 billion globally, encompassing numerous commercial properties. Similarly, JLL manages 5.6 billion square feet internationally, predominantly in multi-tenant settings. CBRE's portfolio stretches beyond 8 billion square feet.

Each of these properties with AI-driven systems contends with the multi-stakeholder governance issue. Furthermore, the difficulty escalates with tenant numbers; a five-tenant property may face manageable conflicts, while a 47-tenant scenario results in combinatorial complexities.

Cognitive Corp’s BAGI assessments of multi-tenant properties reveal governance scores that range from 12 to 22 on a 100-point scale. Specifically, the dimension assessing fairness in AI-driven resource allocation among tenants averages below 8. Most properties lack a documented framework explaining their AI resource distribution processes.

What Governed Multi-Stakeholder AI Looks Like

The Building Constitution framework addresses multi-stakeholder governance through several of its principles:

Fairness. AI allocation must adhere to documented rules open for tenant review. This doesn't mean equal treatment; rather, it should ensure proportional treatment. A tenant occupying 12% of a building should receive a corresponding share of resource benefits while factoring in specific lease terms and spatial needs.

Transparency. Tenants have the right to understand the impact of AI decisions on their space. This doesn’t necessitate sharing the source code but involves providing clear explanations such as: "Your floor was 2°F above target for 47 minutes yesterday due to a grid demand response event, impacting your space relative to the building’s average." Achieving such transparency necessitates comprehensive decision logging, often absent in current building AI frameworks.

Accountability. If AI allocation decisions yield tangible impacts on tenant operations, clear accountability mechanisms must be established. Who approved the optimization parameters? Who sets priority rules? Who reviews allocation fairness periodically? Without accountability, tenants are left with complaints as their only recourse — echoing how feedback from impacted tenants can further cement inequities within the AI system as it learns from incoming data.

The Governance Architecture

To achieve a governed multi-stakeholder AI system, five core elements must be implemented:

1. Tenant allocation transparency reports. Monthly reports should delineate the effect of AI allocation on individual spaces, relative to (a) lease terms, (b) building averages, and (c) past months. This enhances visibility into patterns obscured by complaint-driven feedback alone.

2. Allocation fairness audits. Conduct quarterly reviews of AI decisions across all tenants, searching for consistent biases or favoritism. Is a particular tenant receiving preferential treatment? Are specific floors systematically overlooked? These audits need not measure fairness against arbitrary standards but should reflect the documented allocation principles.

3. Tenant-specific governance parameters. Each tenant must have defined allocation requirements — covering not just temperature ranges, but also priority levels for various scenarios, such as demand response or emergencies. These parameters should serve as inputs to the AI, being documented in the governance schedule attached to the tenant's lease rather than obscure vendor settings.

4. Cross-tenant privacy boundaries. Regulations must be in place detailing how tenant occupancy, access behaviors, and data can inform or affect AI-driven decisions for other tenants, ensuring that such usage is both transparent and consensual.

5. Multi-stakeholder governance committee. Form a representative body — independently from the landlord — that evaluates AI allocation fairness metrics, approves amendments to allocation norms, and addresses tenant concerns regarding AI strategies. This accountability structure transforms decisions from being perceived as "AI determinations" into actions formalized by a governing committee.

The Lease Gap

The commercial reality surrounding this issue is pressing: lease renewals are currently taking place in buildings where AI governance is unaddressed.

A tenant considering a 10-year lease at $89/sqft in an AI-driven property is committing to a potential $4.45 million agreement (for 5,000 sqft) without understanding how AI will influence resource allocation for the next decade. There’s no clause for allocation transparency, no rights to audit fairness, and no role defined for tenant involvement in governance.

Landlords who prioritize these concerns will secure a competitive edge. In an environment where Class A office vacancies are rising, claiming "Our building embraces AI governance principles" is not a luxury; it’s a market differentiator. As tenants evaluate their options, superior AI governance strategies can emerge as pivotal attraction and retention tools.

Leading operators like BXP, Brookfield, JLL, and CBRE are well-positioned to set the industry standard. The first major commercial real estate operator to unveil a multi-stakeholder AI governance charter won’t merely protect their tenants; they will carve out a significant pioneering role.

What to Do Before Your Next Lease Negotiation

For building owners and operators:

1. Conduct thorough audits of AI allocation. Proactively analyze three months of HVAC, lighting, and elevator allocation data to identify whether any tenants receive systematically advantageous or unfavorable treatment.

2. Draft a tenant AI governance addendum. Create a succinct document outlining operational AI systems, decision-making processes, allocation mechanisms, transparency measures, and recourse opportunities. Attach this to new and renewed leases.

3. Ensure proper manual override hygiene. If building management’s manual overrides foster AI biases favoring certain tenants, establish a robust policy requiring all overrides to be documented, with patterns reviewed quarterly for fairness.

For tenants:

1. Inquire directly. Prior to signing or renewing a lease, ask questions like: "What AI systems influence my space, and how are resource allocations determined?" The landlord's response reveals their governance sophistication.

2. Request allocation reports. If buildings cannot provide documentation detailing the effects of AI decisions on your space against building averages, it may indicate transparency deficiencies.

3. Negotiate governance clauses. Suggest adding terms for allocation transparency reports, fairness audit participation rights, and non-discrimination clauses focused on AI resource distribution. If reluctant, seek clarification on their hesitance to commit to equitable treatment.

Buildings that effectively govern multi-stakeholder AI allocations prove to be more efficiently managed. They emerge as sound investments — experiencing lower tenant turnover, minimizing disputes, achieving robust lease economics, and establishing a defensible market presence in an increasingly critical area.

---

James C. Waddell is President of Cognitive Corp. Cognitive Corp specializes in AI governance assessments for commercial real estate portfolios, providing insights into fairness audits for multi-stakeholder allocations and developing effective governance charters.

→ Request a multi-stakeholder governance assessment: [link]

#AIGovernance #CRE #MultiStakeholder #BuildingAI #BuildingConstitution #SmartBuildings