There is a particular conversation the firm keeps having with operators of physical assets — property managers, energy operators, facilities directors. They have been pitched AI by four different vendors in the last quarter. Each pitch was confident. Each promised dramatic operational transformation. None of the pitches resembled each other. The operator is somewhere between hopeful and exhausted, and the question they are really asking, underneath the polite framing, is straightforward: which of these is real, and which is marketing?
It is a fair question. It deserves a more honest answer than the market is currently providing.
The shortest version of the answer is this. AI is genuinely transformative in a narrow set of operational contexts, marginally useful in a wider set, and actively harmful when it is introduced in places where the underlying operational data and processes are not ready for it. The hard part of doing this well is not the AI. The hard part is knowing which of those three categories a specific operation falls into. Most vendors will not say, because their incentive is to sell the same product regardless of the answer.
This brief is the unsold version of that conversation. It reflects how Bodenhamer & Co. thinks about the topic across the engagements the firm has run, the operators it has advised, and the patterns it sees consistently across portfolios in Europe and globally.
I. Where AI delivers operational value today
The firm has identified three patterns where AI produces real, measurable, durable value in property and energy operations. Each shares a structural feature: the conversion of unstructured human work into structured operational signal, in a domain where the cost of doing that work manually is genuinely high.
Tenant and customer communication at scale
A mid-size property operator running 800 units processes thousands of tenant communications per month — maintenance requests, billing questions, lease inquiries, complaints, repeated questions about the same building system that breaks every winter. Most of this volume is handled by human staff who are simultaneously trying to do everything else. The work is repetitive, emotionally draining, and a significant portion of operational labour cost.
An AI-assisted communications layer that drafts responses, classifies urgency, routes to the right person, and surfaces patterns across the portfolio is not a marginal improvement. It is a structural shift in what one operations person can handle. Operators who deploy it well report cost reductions in the range of 25 to 40 percent on communications-related labour. Tenant satisfaction generally improves rather than declines, because response times collapse from days to minutes.
Document processing for compliance and reporting
Property and energy operators in Europe are buried in document workflows: lease agreements, energy performance certificates, compliance inspections, ESG and CSRD reporting, contractor invoices, regulatory filings. Most of this processing is still done by humans copying values from PDFs into spreadsheets.
AI is genuinely good at this work now — extraction, classification, summarisation, anomaly detection across document corpora. The savings are real, the risk is manageable when humans remain in the loop on consequential decisions, and the compliance burden is only growing. This is one of the few places where the firm advises operators to move quickly rather than wait.
Technical knowledge access for field staff
Every operator has a deep pool of institutional knowledge — system manuals, historical maintenance records, vendor contacts, regulatory requirements, operational procedures — and almost none of it is accessible to the people who need it in the moment. A field technician troubleshooting a heat exchanger at 7 a.m. is doing it from memory, from a phone call, or from whatever PDF can be located on a shared drive.
AI-assisted access to internal operational knowledge — a system that lets staff ask questions in natural language and receive cited answers from the operator's own documentation — produces meaningful productivity gains. More importantly, it captures the operational expertise of senior staff before they retire. The demographic mathematics on this are unfavourable across the European property and energy sectors, and most operators are underestimating how much it will matter by 2030.
These three categories share something else worth naming. In each, AI does work that humans are currently doing badly because there is too much of it. AI is not replacing judgment. It is removing the volume of low-judgment work that prevents humans from doing the high-judgment work well. That distinction is the difference between AI that holds up in production and AI that gets quietly turned off six months after deployment.
II. Where AI is marginally useful
There is a second category where AI can produce value but where the value is real-but-modest, and where the wrong implementation choices easily turn it negative. Most operators get into trouble here, because the marketing pitches are at their most confident in this band.
Predictive maintenance is the headline example. The pitch is seductive: sensors on equipment, machine learning on the data, prediction of failures before they happen, dramatic reduction in downtime. The reality is more nuanced. Predictive maintenance produces real value when three conditions hold simultaneously — the equipment is expensive enough that downtime is genuinely costly, the failure modes are observable in sensor data, and the operator has the maintenance organisation in place to act on predictions. In commercial real estate and most building operations, those conditions hold for a small subset of equipment — major HVAC systems, large pumps, critical electrical infrastructure — and not for most of the rest. Vendors selling portfolio-wide predictive maintenance are usually overselling. Operators who succeed deploy it narrowly and ignore the rest.
Energy optimisation through AI is similar. There are real gains in dynamic load management, demand response participation, HVAC scheduling, and tariff arbitrage. But many of those gains are also available through good rule-based control systems that have existed for decades. The marginal benefit of layering AI on top of well-tuned rules is often smaller than vendors suggest. The honest position the firm takes is that most operators should fix their basic energy management before considering AI optimisation. AI applied to a portfolio with poorly commissioned building management systems and unreliable sub-metering will produce confident-looking dashboards that do not translate into actual savings.
Tenant experience personalisation, dynamic pricing for amenities, AI-assisted leasing — these are where the firm sees the most marketing and the least durable value. Some applications will work eventually. Most current implementations are solving for problems operators do not actually have, at the expense of attention that should go elsewhere.
III. Where AI is actively harmful
The third category is the one nobody likes to discuss. There are operational contexts where introducing AI before the underlying operations are ready will produce worse outcomes than not introducing it at all.
Operations with poor underlying data. AI systems amplify whatever signal exists in their input data. If sub-metering is unreliable, maintenance records are inconsistent, tenant data is fragmented across three systems, and vendor invoicing is a mix of PDFs and spreadsheets — applying AI to that environment will produce confident outputs based on bad inputs. The danger is not that the AI fails obviously. It is that it produces plausible-looking results that drive decisions, and the errors do not surface until the consequences are already in place. Operators in this situation should fix their data foundation before they layer AI on top. This is unglamorous work, and most consultants will not recommend it because there is no AI product to sell.
Operations with high regulatory or liability exposure. Anything that touches tenant rights, fair housing compliance, lease enforcement, regulatory reporting, or safety-critical systems should not be where an operator pilots their first AI deployment. The failure modes are too consequential and the regulatory environment around AI in these contexts is evolving quickly. Use AI in lower-stakes operational contexts first, build organisational capability, and only then consider higher-stakes applications. The firm declines engagements that would invert this sequence.
Operations where the team is not ready. This sounds soft and is not. AI deployments fail in production not because the technology fails but because the human workflows around the technology fail. If the operations team does not trust the new system, does not know when to override it, does not have a clear escalation path when it produces unexpected results, and does not see how it fits with the rest of their work — the deployment will be quietly abandoned regardless of how good the underlying AI is. Vendors do not talk about this because it is not their problem after the contract is signed. It is, however, the single most common reason real-world AI deployments do not deliver promised value, and Bodenhamer & Co. designs every engagement around it.
IV. The connected hardware piece
A separate observation, because it keeps appearing in the firm's engagements. Most of the value in property and energy operations is not actually about AI in isolation. It is about the integration of AI with the connected hardware operators already have or are installing — meters, sensors, building management systems, EV charging infrastructure, heat pump controls, distributed energy assets.
The hard problem in most operational contexts is not the AI. It is the integration layer. Operators have data scattered across building management systems running BACnet or Modbus, energy meters reporting to one vendor's portal, tenant systems on a different platform, contractor management on yet another, and ESG reporting being assembled manually from all of these once a quarter. AI gets to be useful only after that integration problem is solved. Most operators underestimate how much of the work and cost in any serious operational intelligence project is integration, not intelligence.
This is where the next several years of operational software in this sector will be defined. The operators who win will be those who treat their physical assets as a single coherent operational system — building, energy, tenant, regulatory — rather than five separate systems that happen to coexist. AI is one component of that. It is a component, not the centre.
This integration discipline is, in practice, what most distinguishes the firm's engagements from those of the AI consultancies and the building technology vendors operators encounter elsewhere. Neither side is wrong about their own piece. They are wrong about the assumption that their piece can be optimised independently.
V. What an operator should actually do
The sequence the firm recommends to operators at the beginning of this journey is unglamorous and is the one that works.
Start with an honest read on the operational data foundation. Are the basic systems — metering, work order management, tenant data, document storage — capturing reliable, consistent, accessible data? If not, this is the first investment, and it has nothing to do with AI. Operators frequently want to skip this step. They should not.
Identify two or three operational pain points where the volume of low-judgment work is genuinely consuming senior staff attention. Tenant communications is usually one. Document processing for compliance is often another. Field staff knowledge access is often a third. Pilot AI in those specific contexts, with clear success metrics, with humans in the loop on consequential decisions, and with explicit plans for what happens if the deployment does not work. Run those pilots for at least six months before scaling.
Treat connected hardware strategy as a portfolio decision rather than a series of vendor selections. The integration layer matters more than any individual component. Operators who think clearly about this in 2026 will be in a structurally better position by 2030 than those who acquire systems piecemeal.
Avoid the temptation to launch a portfolio-wide AI initiative. Those programmes consistently underperform narrow, well-scoped, operationally embedded deployments. The operators producing real results are doing three things well, not thirty things badly.
And resist the consultants and vendors who insist that any of this is straightforward. It is not. Anyone claiming otherwise is selling.
Closing
AI is real, valuable, and applicable to property and energy operations — but in a more specific way than the marketing suggests, and on a longer timeline than the vendors want operators to believe. The operators who get this right will quietly outperform their peers over the next decade. The ones who get it wrong will spend a great deal of money on confident-looking dashboards that do not translate into operational reality.
The difference between those two outcomes is rarely the AI. It is almost always the operational thinking around it.
Which is, perhaps, why Bodenhamer & Co. exists.
Bodenhamer & Co. is an independent advisory firm working with property and energy operators on AI, connected systems, and operational software. Headquartered in Sweden, practising across Europe and globally.