Skip to content
Jaanika
7 min readagentic ai customer experiencebuildings industry cxautonomous buildingschannel abandonmentphone vs digital supportcx governancehuman-ai escalationfacility management customer experience

Customer Experience in the Agentic AI Era: What the Buildings Industry Still Gets Wrong

The buildings industry is pouring energy into building autonomy (self-optimizing systems, predictive maintenance) while under-designing the human experience layer around it. Two things are true at once: expectations for AI-driven service are rising fast, but tolerance for AI failure is razor-thin, especially in B2B where a bad automated interaction means a missed SLA, not just an annoyed shopper.

Updated 21 August 2026

The buildings industry has spent the last two years talking about autonomy: self-optimizing HVAC, predictive maintenance, agents that renegotiate energy contracts without a human in the loop. Frost & Sullivan now describes buildings as moving from “passive assets” to “intelligent, adaptive ecosystems capable of autonomous decision-making.” That’s the right framing for the asset. It’s the wrong framing for the customer.

Buildings don’t buy things. People do: facility managers, tenants, procurement teams, service technicians, the property owner’s ops director who still calls at 7am because a chiller alarm woke them up. Somewhere between “autonomous building” and “loyal customer” sits an experience layer that most B2B buildings companies haven’t finished designing, and the gap is starting to show.

Where the market actually is

Two things are true at once right now, and most vendors only plan for one of them.

First, expectations have shifted fast. Industry surveys this year put the share of consumers who expect AI to make service faster and better at over three-quarters. Gartner projects that by 2028, 60% of brands will be using agentic AI for one-to-one customer interactions, and early industrial deployments (a European energy provider is one cited example) are already reporting double-digit satisfaction gains from well-governed agentic rollouts.

Second, patience for AI that fails is almost nonexistent. Most consumers give a virtual agent two or three tries before giving up, and a meaningful share say they’d switch providers after just two or three bad automated interactions with a brand they otherwise liked. In a B2B buildings context, where a “bad interaction” might mean a misdiagnosed fault, a missed SLA, or an agent that closes a ticket a tenant considers unresolved, that tolerance is even thinner, because the cost of getting it wrong isn’t an annoyed shopper, it’s a facility manager who has to explain an outage to their own boss.

Only about a third of CX infrastructure across industries is fully cloud-native today, which is the quiet reason so many “agentic AI” pilots stall: you can’t orchestrate a real-time, cross-channel journey on top of a ticketing system, a separate BMS platform, and a phone tree that don’t share data. The bottleneck usually isn’t the AI. It’s the plumbing underneath it.

The channel-abandonment trap

Here’s the part that gets missed in most agentic-AI strategy decks: digital adoption curves are real, but they are not uniform, and the buildings customer base is one of the least uniform out there.

The data on channel preference is genuinely messy: different studies report anywhere from a third to nearly half of customers naming phone as their top channel, and the range depends heavily on how the question is asked. But a few patterns hold up across almost every 2025-2026 report:

  • Preference splits sharply by age. Consumers over 55 name phone as their default channel roughly half the time in most surveys, and some put it much higher for the boomer cohort specifically. Younger customers say they prefer digital-first channels, but even Gen Z, the generation assumed to hate phone calls, shows usage rates in the 70-80% range and often rates phone as the fastest way to resolve something urgent or high-stakes.
  • Phone ownership doesn’t track phone preference. Roughly 88% of consumers across age groups still use the phone for at least some service interaction, even though only about a third rank it as their favorite channel. Phone has effectively become the escalation lane: people arrive there after self-service or chat has already failed them, which means the calls that do land are disproportionately the hard ones (angry, urgent, or complex). That’s a bad channel to under-resource.
  • The failure mode is asymmetric. Forrester’s CX Index has tracked a multi-year decline in overall CX quality across brands, and older, long-tenured customers, the kind who in commercial real estate and facilities often include the property managers and building owners who control renewal decisions, are typically the least forgiving of that decline when a brand shifts too hard toward digital-only support.

For a B2B buildings company, this isn’t an abstract stat. Your customer base spans a facility technician who wants a chat interface and instant diagnostics, a 60-year-old building owner who wants to talk to a person when a fire panel throws a fault, and a procurement lead at a REIT who wants an API. Digitizing the experience is the right long-term bet: digital channels are absorbing a growing share of interactions almost everywhere, and self-service and AI containment keep improving. But “the trend is upward” is not the same as “we can retire the phone team.” Running both in parallel, properly staffed, isn’t a transitional inefficiency to be engineered away as fast as possible; for a multi-year stretch, it is the strategy.

What actually builds or breaks trust with agentic systems

A few patterns are consistent across the industrial and CX literature this year:

Governance has to be visible, not just present. The deployments getting cited as successes build oversight into the agent from day one: scoped permissions, a decision log per action, human approval checkpoints above a defined risk threshold, and clear escalation paths when the agent hits ambiguity. In buildings specifically, that matters doubly: an agent that can adjust setpoints autonomously needs guardrails that are legible to the facility manager who’s ultimately accountable for the space, not just to the vendor’s engineering team.

Failure needs a fast, human off-ramp. Given how little patience customers have for repeated AI failure, the single highest-leverage design decision most companies can make is making the handoff to a human fast and low-friction, not making the AI perfect. A tenant who has to fight to reach a person after the bot fails twice is a tenant who remembers the fight, not the ten good interactions before it.

The economics push hard toward automation, so the discipline has to be deliberate. A human phone interaction can run well over $15 fully loaded, versus well under a dollar for an AI-handled equivalent, which is exactly why the temptation to over-rotate toward digital is so strong, and exactly why it has to be resisted where the customer base doesn’t support it yet. The cost savings are real; the customer segmentation work to know where they’re safe to take is the part that gets skipped under time pressure.

Trust in autonomous building operations is transitive. If the agentic layer gets the small things right (a correctly triaged fault, a maintenance visit scheduled before the tenant even reports the problem), customers extend more trust to the bigger autonomous decisions the building is making on their behalf. Get the small things wrong, and every subsequent “the building decided X” announcement gets read with suspicion.

What B2B buildings companies need to actually prepare

  • Segment the customer base by channel readiness, not by assumption. Age is a decent proxy but tenure, account complexity, and issue type matter more. Build the segmentation before you build the routing logic.
  • Fund the phone channel like a strategic asset for as long as your data says it needs to be one, not as a legacy cost center you’re trying to shrink to zero on a fixed timeline. Revisit the timeline against actual usage data, not the roadmap you wrote two years ago.
  • Design escalation paths before you design automation. The AI’s failure mode is the moment that actually determines customer loyalty.
  • Make agentic decision-making auditable and explainable to the people accountable for the building, not just to your own compliance team.
  • Treat CX and building operations as one system. The Genesys framing this year is the right one: the next era of CX isn’t defined by AI adoption alone, but by how well AI connects with the people, data, platforms, and governance needed to make it trustworthy. In buildings, “the people” includes both your own frontline team and the human on the other end who still, for a good long while, wants to be able to pick up the phone.

The autonomous building is coming. The autonomous customer relationship isn’t, not yet, and not uniformly. The companies that win this transition won’t be the ones with the most advanced agents. They’ll be the ones that kept both systems, digital and human, running well enough, long enough, that no customer ever had to find out the hard way which one they’d been quietly deprioritized into.

Sources

← All writing