AI agents in building tech: where the margin really hides

The paradox that applies to our industry too

Nearly nine companies out of ten, across all sectors, have adopted AI in at least one business function. Only one in fifteen claims to have derived "significant" value from it. For building tech — lighting, HVAC, electrical distribution, smart buildings — the paradox has a specific shape: AI technology is already in the products we sell (controls, sensors, cloud platforms) but rarely in the processes we use to sell, support and grow them through the channel.

AI agents — systems capable of planning, reasoning and executing multi-step tasks autonomously, not merely reactive copilots — promise to close that gap. Among industrial companies that have scaled them, two thirds report tangible productivity gains and over half report cost savings. But value is not evenly distributed. It concentrates on those who made a precise choice in the processes that really matter for margin — quoting, channel governance, after-sales, demand planning — and made it quickly.

This article is an operating map for that choice, applied to our industry.

Where agents generate value (and where they are just demos)

Three messages should guide the decision.

First, value is not where everyone is looking. The industry's attention is almost monopolised by predictive product diagnostics — Trane Technologies with BrainBox AI reports HVAC consumption reductions of 15–25%; Johnson Controls OpenBlue claims up to 30% energy savings on HVAC and lighting. These are real numbers, but they go to the end customer, not to the manufacturer's or distributor's P&L. For those selling building tech products and solutions, margin is played elsewhere: in the speed and quality of quoting, in the governance of the distribution network, in the reduction of post-sale cost-to-serve, in the alignment between forecast and production capacity. That is where AI agents should be aimed to generate EBIT, not just brochureware.

Second, pure efficiency is a trap. Companies that genuinely capture value from AI — the 6% identified by McKinsey as high performers — do not chase cost cutting. They simultaneously target growth and innovation, and are 3.6 times more likely than others to pursue enterprise-wide transformation. Translated for our industry: those who set up AI agents as a "back-office automation" project obtain modest, isolated savings. Those who set them up as go-to-market and channel-model redesign get the multiplier.

Third, the constraint is not technological, it is organisational. Industry data is fragmented across six to eight systems: ERP, CRM, configurators, price lists, BIM, ticketing, e-commerce, distributor portal. Those who layer agents over non-integrated silos get demos. Those who first tackle the cleansing and unification of product and channel data get leverage.

Five families of high-return use cases for building tech

1. Quoting, specifications and tenders. This is the first point of margin compression across the industry. An AI agent can read a tender specification or a technical request, extract requirements — luminous flux, IP ratings, required certifications, energy classes, compatibility constraints — and come back with a product configuration, cross-sell alternatives and a draft offer in minutes instead of days. The impact is not only on time: it is on win-rate, because in tenders where the first to quote with technical quality wins the edge, reducing time-to-quote from five days to four hours changes conversion structurally.

2. Channel governance and distributor support. Sonepar — the world's leading distributor, €33.6 billion in 2025 revenue — brought its omnichannel platform Spark to over €11 billion in digital sales in 2024 and explicitly put AI at the centre to predict customer needs, optimise inventory and logistics, personalise recommendations. Rexel, with its Axelerate 2028 plan, already has 33% of sales in digital. Those who sell through these distributors — or through smaller networks — can use AI agents for automatic lead routing by territory and specialisation, anomaly alerts on sell-through, generation of distributor-specific commercial proposals, contractual performance monitoring. It is the difference between "managing the channel by feel" and "managing the channel with a nervous system".

3. After-sales, technical support and field service. This is the area where cost cutting is most documented and immediate. A public case illustrates the principle: an office tower in Tokyo used AI to intercept anomalous fan vibrations and compressor problems in advance, avoiding $40,000 in emergency repairs. Honeywell reports a 60% reduction in fault detection and resolution times through Forge. Deloitte estimates that predictive maintenance reduces maintenance costs by 25-30% and eliminates 70-75% of unplanned downtime. The principle also applies to manufacturers: an agent that triages technical support tickets, identifies serial defects, routes the right technician and suggests the correct spare part can reduce post-sale cost-to-serve by 20-30%, recovering margin otherwise eroded by call centres and travel.

4. Technical specification and designer support. Schneider Electric integrated AI directly into its SpaceLogic Touchscreen Room Controller, with up to 23% reduction in occupant complaints. But more relevant for sellers: an agent that assists the design firm with product selection — lighting calculations, HVAC sizing, compliance checks, alternatives research on discontinued or critical-lead-time products — shifts the company from supplier role to specification-partner role. It is the most undervalued leverage point in the industry: those who control the specification control the conversion, because the specified product wins the tender before the distributor even quotes.

5. Demand planning and lead time management. Signify, with its SmartBright All-In line, explicitly designed products that reduce distributor warehouse SKUs. It is a product innovation, but the strategic pattern is identical on the AI side: agents that integrate signals from project pipeline, channel sell-through, lead times of critical electronic components and production capacity, and that dynamically reallocate production priority. In an industry where component lead times can determine whether an order closes or not, this translates directly into win-rate and cash conversion.

The decision framework: three horizons

Horizon 1 — Efficiency of existing processes (0–6 months). Insert agents into already-mapped flows, with measurable KPIs: average quoting time, win-rate by value bracket, post-sale cost-to-serve, three-month forecast accuracy. Objective: validate the ROI and build the learning curve. Contained investment, fast return, important internal narrative.

Horizon 2 — Flow redesign (6–18 months). Redesign processes assuming the agent is a node of the workflow. Concrete example: instead of automating the reading of a specification, redesign the entire specification-to-order cycle with agents orchestrating technical reading, configuration, dynamic pricing, offer generation, commercial follow-up and human escalation only where needed. This is where you enter high performer territory.

Horizon 3 — New service models (12–36 months). Offerings enabled by agentic AI that today would not be economically sustainable. Net Zero Buildings as a Service — the model Johnson Controls launched — is exactly this: the customer pays for the energy savings generated instead of CapEx. For a mid-sized manufacturer or distributor the equivalent is offering the end customer a continuous optimisation subscription service, economically sustainable only thanks to agent autonomy. This is where long-term competitive advantage is played.

The typical mistake is staying glued to Horizon 1, mistaking a collection of successful pilots for a strategy. The industry companies that capture value are all pushing toward Horizon 2 and 3.

The operating blueprint: 90 days, 6 months, 12 months

First 90 days — Setting the ground

Select two use cases with structured data and strong functional sponsors. For building tech, the safest choices are quoting (data in CRM/ERP) and post-sale ticket triage (data in ticketing system). Form an agile team of 4–6 people with a business product owner — commercial director or service director, not just IT. Define metrics upfront: time-to-quote, win-rate, FCR (first contact resolution) on service. Set up minimum governance: human-in-the-loop on offers above a certain threshold, complete audit trail, ownership of exceptions.

From 90 days to 6 months — Redesigning the flow

Once the pilot is validated, stop before scaling. The question is not "how do we deploy the agent faster" but "how does the sales process, or the service process, change if we take the agent as the starting point". Reallocate sales engineer and support technician time toward high-value activities — relationship with specifiers, key account development, channel training. Build a minimum internal capability — an AI product owner, a data engineer, a governance lead — to reduce dependence on external consultants.

From 6 to 12 months — Scaling the platform

Standardise a single agentic infrastructure for commercial and service. Extend to Horizon 2 use cases: integrated project-distributor-installer management, demand planning linked to project pipeline, distributor churn prediction. Start measuring impact at EBIT level by business unit, not just by project. This is the shift from "sum of pilots" to "transformed line of business".

Three risks specific to building tech

Data fragmentation risk. The industry has product data distributed across ERP, configurators, PDF datasheets, BIM, distributor price lists, CRM. An agent that queries six systems without a unified product layer produces inconsistent answers, and a single wrong answer on a specification is enough to lose the tender. Those who skip the Product Information Management phase burn the economic case.

Compliance and technical liability risk. Lighting, HVAC and electrical distribution are safety-critical and heavily regulated sectors (CEI, EN, IEC, fire safety regulations, EPBD). An agent that proposes configurations or suggests substitutions without traceability of applied standards is a reputational and legal risk disproportionate to the saving. The principle "human-in-the-loop where it counts, autonomous where it is safe" is not optional — and must be coded from the pilot phase.

Channel rejection risk. When an agent automates quoting, configuration or technical support, it is entering activities that some channel actors consider core to their own value add. Without an explicit channel enablement strategy — not replacement, but upgrade — scaling stops at the resistance of commercial agents, installer partners or in-network specifiers.

The question to bring to the board

For those leading a lighting, HVAC, electrical distribution or building tech company, the question is no longer "are we using AI?". It is: what percentage of our EBIT is reasonably attributable to AI initiatives twelve months from now, and in which commercial and service processes are we looking for it?

If the answer is "some back-office savings" or "we have a pilot on predictive maintenance for an important customer", the trajectory is clear — and it is the same as the 94% of companies that today report not seeing significant value from their investments.

If instead the answer includes quoting, channel governance, after-sales and demand planning fundamentally rethought over the next six to twelve months, with explicit EBIT KPIs and a structured team to lead them, the company is positioning itself in the small group that will capture the disproportionate share of value — and in the next cycles of tendering, restocking and contract renewal, it will be felt.

AI agents in building tech are not a technological solution. They are a management choice. And like all management choices that count, they reward those who decide early and well — and penalise those who wait for the market to settle.