Building Analytics and FDD— answered

Clockworks Analytics is a cloud-based fault detection and diagnostics (FDD) platform for commercial buildings. Founded in 2008 within MIT’s Building Science Department, Clockworks plugs into existing building management systems (BMS), continuously analyzes thousands of equipment data points, and pinpoints the highest-impact performance issues — complete with root-cause diagnostics, 0–10 prioritization, and projected cost savings.

The platform monitors more than 600 million square feet across 3,500+ buildings in 30+ countries, and clients have documented over $60 million in annual avoidable costs.

Fault detection and diagnostics (FDD) is software that continuously analyzes data from building systems — air handlers, chillers, boilers, VAV boxes, pumps, and more — to automatically identify equipment faults and energy-saving opportunities. Detection tells you something is wrong; diagnostics tells you why it’s wrong, what it costs, and what to do about it.

FDD lets facilities teams shift from reactive maintenance to proactive, data-driven operations.

BMS alarms and custom if-this-then-that fault rules generate alerts but leave the investigation to your team — often producing false positives and alarm fatigue. Clockworks instead builds a digital model of every piece of equipment, combining BMS data points, control sequences, and mechanical schedules, and applies a global analysis engine refined over more than a decade across thousands of buildings.

There are no per-building rule libraries to write or maintain, and results arrive as diagnostics — root cause, avoidable cost, and recommended actions — rather than raw alerts.

Most facilities teams are stuck reacting to complaints, critical alarms, and scheduled rounds. Clockworks replaces that cycle with a daily, prioritized list of the issues that matter most — found by continuous diagnostics, not occupant complaints. Degrading equipment is caught early, before failure or comfort impact, and every issue arrives with root cause and recommended action so limited staff time goes to fixing problems instead of finding them.

Condition-based maintenance means servicing equipment based on its actual measured condition rather than fixed calendar schedules. Clockworks enables condition-based maintenance by continuously monitoring HVAC and building equipment through the BMS and flagging the specific faults and inefficiencies that need attention — so teams spend time where it has the most impact instead of on unnecessary scheduled rounds.

Clockworks works in five steps:

Connect. A software gateway installed in the building extracts thousands of data points from existing BMS and metering systems every five minutes and sends them to the Clockworks cloud.
Model. A digital model of every piece of equipment is created from data points, control sequences, and mechanical schedules.
Diagnose. A continuously improving analysis engine reviews equipment operation to find and diagnose performance issues and optimization opportunities.
Prioritize. Every issue is scored 0–10 for its impact on energy, indoor environment, and equipment reliability, with avoidable costs quantified.
Act. Your team receives a prioritized task list with AI-generated recommended actions.

Clients use Clockworks to lower energy costs, improve occupant comfort and indoor environmental quality, extend equipment life, and focus limited staff on the highest-impact work. Across the platform, clients have documented over $60 million in annual avoidable costs and completed more than 60,000 maintenance tasks.

For example, Clockworks partner Wendel identified $616,000 in annual energy savings for a single private university client. Every finding includes a quantified avoidable cost so savings can be tracked and verified over time.

Clockworks serves healthcare systems, universities, corporate real estate organizations, and facility-management and mechanical/controls service providers in more than 30 countries. The platform is used both directly by in-house facilities and energy teams, and by service-provider partners who deliver analytics-driven maintenance to their clients.

Clockworks uses a layered AI stack — expert systems at the core, machine learning for asset inference, and large language models for document parsing and orchestrated workflows. Each layer is applied where it adds the most value:

When diagnostic accuracy matters most, expert systems do the heavy lifting. Physics-based logic doesn’t guess.

There’s a lot of noise around AI in buildings. Before trusting a platform with your building operations, it’s worth understanding exactly what different AI technologies do and why Clockworks uses each one:

• Expert Systems (ES) use a domain-specific knowledge base to apply physics-based logic to complex problems. This is the core of Clockworks diagnostics — the engineering equations governing HVAC performance are well defined, so we encode them directly rather than training a model to learn them. The result is transparent, explainable root-cause diagnostics that technicians can act on.
• Machine Learning (ML) uses statistical algorithms to learn patterns from data without explicit programming. Clockworks applies ML for asset inference — recognizing equipment configurations and operating relationships from available BMS data.
• Deep Learning (DL) is a subset of ML that passes data through multiple model layers. It underpins large language models.
• Large Language Models (LLMs) are trained on massive text corpora to generate and understand language. Clockworks uses LLMs for document parsing during onboarding and for generating AI-powered task summaries.

For HVAC fault detection, physics-based expert systems outperform machine learning on the three things that matter most in building operations: accuracy, transparency, and reliability.

Accuracy. The engineering equations governing building performance are well established. You don’t need to train a neural net to learn thermodynamics — you apply the principles directly. Machine learning requires large historical datasets of labeled faults that simply don’t exist for most buildings.

Transparency. Expert system diagnostics tell you exactly why a fault was detected — root cause, co-occurring issues, and recommended action. A technician can verify the reasoning and act with confidence. Black-box ML models can’t offer that.

Reliability. Clockworks’ expert systems are aware of each other — if one analysis identifies a faulty sensor, it prevents false positives from propagating to other rules. This cross-checking is difficult to replicate in purely data-driven approaches.

Machine learning has a role in the Clockworks stack — specifically for inferring equipment configurations from available data — but when root-cause accuracy is the standard, expert systems are the right tool.

Clockworks uses fine-tuned Azure OpenAI large language models across four areas of the platform:

Document parser. Mechanical schedules, sequences of operations, and controls drawings are parsed automatically by fine-tuned LLMs that extract equipment names, types, ratings, and control parameters and convert them into structured JSON for the Clockworks information model — eliminating manual data entry during onboarding.

AI-generated diagnostic summaries. Azure OpenAI generates natural language summaries of diagnostic results and recommended actions, translating complex FDD output into clear, actionable insights for building operators and technicians.

Natural language data exploration. Users can query building data in plain English — “How much energy did I waste in the last 30 days?” — and an AI agent automatically generates the underlying KQL query against Clockworks’ data model and produces the corresponding visualization.

Orchestrated AI workflows. A LangGraph-based agent platform orchestrates multiple purpose-built AI agents — for assets, diagnostics, model metadata, raw data insights, and KPIs — allowing complex multi-step tasks like generating a full diagnostic report across a portfolio of buildings from a single natural language prompt.

LLMs handle language, inference, and orchestration. Physics-based expert systems handle the diagnostics. Together, they make Clockworks both accurate and genuinely easy to use.

When a diagnostic finding becomes a task in Clockworks, an AI summary automatically drafts the task description and recommended actions. It synthesizes:

• The specific fault identified and its root cause
• The equipment’s configuration, class, and history
• Co-occurring faults on the same piece of equipment
• The energy, comfort, or maintenance priority and cost impact
• Recommended corrective actions drawn from the global knowledge base

The result is a work order that arrives pre-populated with context — so technicians spend time fixing problems, not writing tickets. This feature was introduced in the September 2025 platform release as part of the redesigned Diagnostic Reports experience.

Yes. Clockworks’ natural language data exploration feature lets users type questions in plain English — for example, “How much energy did I waste in the last 30 days? Plot as an area chart” — and an AI agent automatically generates the accurate KQL query aligned to the Clockworks information model, retrieves the data, and produces the appropriate visualization.

This removes the need for users to know query syntax or data structures to get answers from their building data. It sits alongside the full diagnostic and task workflows, giving teams a fast way to investigate specific questions or spot trends without opening a support ticket or writing code.

Clockworks’ orchestrated workflow platform uses a LangGraph-based multi-agent architecture to execute complex, multi-step tasks from a single natural language prompt. Purpose-built AI agents are available for assets, diagnostics, model metadata, raw data insights, quality assurance, and KPIs — each with access to the Clockworks API via a secure MCP server gateway.

A prompt like “Generate a diagnostic report for the Hospital and South Office buildings” automatically triggers a coordinated sequence of API calls: equipment retrieval, diagnostic results retrieval, and building score calculation across both buildings, compiled into a formatted report — with full observability via LangSmith tracing.

This is how Clockworks is bringing generative AI into building operations without sacrificing the accuracy and reliability that facilities teams depend on. Orchestration handles the workflow; expert systems still own the diagnostics.

Clockworks includes an automated QA layer powered by expert systems that continuously checks the information model for completeness and accuracy before and during diagnostic runs. Two types of flags are generated automatically:

• Configuration flags detect issues in equipment and point configuration — for example, a fractional point value outside its expected range.
• Missing information flags detect gaps that limit diagnostic coverage — for example, a missing supply fan rated power or economizer high limit that would prevent certain analyses from running.

This means data quality issues are surfaced proactively, not discovered after a false positive or missed fault. The QA layer is the same expert system technology as the diagnostics — it knows the information model well enough to know what’s missing from it.

Onboarding a building into Clockworks involves extracting configuration data from mechanical schedules, sequences of operations, controls drawings, and utility rate sheets — documentation that can run to hundreds of pages. Clockworks uses fine-tuned Azure OpenAI LLMs to parse this documentation automatically, identifying equipment names, classes, ratings, control sequences, and variable mappings and converting them to structured data for the information model — without requiring manual data entry for each field.

Separately, a machine learning inference engine uses a nearest-neighbor strategy to automatically identify point types and equipment types from raw BMS point names, further reducing manual configuration. Early diagnostic results are available at go-live so clients start seeing value from day one.

One of the most powerful capabilities of the Clockworks AI is its ability to infer how equipment is configured and controlled from available BMS data points alone — without requiring manual modeling of each unit. For example:

• Clockworks can infer an AHU’s full configuration — ducts, fans, coils, dampers — using only available point types, without requiring someone to model every component.
• If an enthalpy point is present, Clockworks infers the economizer is controlled by enthalpy rather than dry bulb or dew point.
• If an RH Max point is present, Clockworks infers a dehumidification sequence even if no explicit “dehumidification mode” point exists.
• If a leaking heating coil valve is suspected but heating is not enabled, the system infers sensor drift rather than a valve fault — suppressing the false positive automatically.

This inference layer — powered by a combination of expert system logic and machine learning — is what allows Clockworks to run accurate diagnostics across thousands of unique building configurations without custom rule-writing for each one.

The Clockworks Information Model is the brain of the platform — a four-pronged global data model combining equipment variables, point types, inferred relationships, and calculated data. Unlike simple point tagging or per-building rule libraries, it captures how equipment is configured and controlled, identifies the relationships between issues, and produces root-cause diagnostics with recommended actions.

It is built for the imperfect data of real buildings, requires no custom rule-writing, and because Clockworks is a cloud-based SaaS platform, every model improvement benefits every customer immediately.

Yes. Clockworks is vendor-neutral and connects to your existing building management system and metering infrastructure through a software gateway — no rip-and-replace of controls hardware is required. The platform connects to virtually any protocol or controls platform: BACnet, Siemens, Honeywell, Schneider Electric, Johnson Controls, and more.

Data from diverse BMS vendors, equipment variants, and controller configurations is normalized into one uniform diagnostic output across your whole portfolio.

The gateway extracts time-series data from BMS points and utility meters every five minutes and sends it securely to the Clockworks cloud. That live data is combined with static configuration — equipment ratings, control sequences, utility rates, and occupancy schedules — to model and diagnose each piece of equipment.

Every diagnostic finding is scored on three independent 0-to-10 priorities. Energy priority reflects the annualized avoidable cost of the issue relative to a configurable cost threshold. Comfort priority reflects how far and how long conditions deviate from setpoints affecting occupants. Maintenance priority reflects the severity and duration of broken or abnormal operation, with the highest scores reserved for potentially dangerous or damaging problems.

This lets teams sort one combined list and act on what matters most.

Avoidable cost is the estimated energy cost attributable to a specific fault or inefficiency — what you could save by fixing it. Clockworks calculates it automatically using engineering energy models and regression techniques that normalize for weather, equipment runtime, and seasonal heating and cooling schedules, then projects an annualized figure used for prioritization and savings tracking.

This avoids the double counting and rule-of-thumb estimates common in basic fault detection tools.

Yes. Diagnostic findings become tasks in Clockworks, and CMMS integration through the REST API allows those tasks to generate work orders automatically in your existing maintenance workflow — so technicians receive actionable, prioritized work without changing systems.

Yes. The Clockworks KPI REST API provides programmatic access to diagnostics, tasks, equipment configuration, avoidable costs, and time-series data. It supports custom queries and integrates with business intelligence tools such as Power BI for portfolio-level reporting.

Every piece of equipment in Clockworks has a digital model of its configuration — class, type, ratings, and design variables — plus a continuous, dated history of diagnostics, faults, priorities, avoidable costs, and completed tasks. Over time this becomes a longitudinal health record for each asset: documented evidence of condition, chronic and recurring issues, and the corrective work performed.

Teams use it to inform maintenance decisions, support repair-versus-replace analysis, and preserve institutional knowledge through staff transitions. All of it is accessible in the platform and through the KPI REST API.

Optimizing existing building systems is often the fastest, lowest-cost path to emissions reduction — and that’s exactly what Clockworks finds. The platform pinpoints operational waste such as simultaneous heating and cooling, schedule and economizer faults, and leaking valves, then quantifies the avoidable energy, cost, and associated carbon impact of each fix.

Completed tasks are tracked with measurement and verification built in, giving sustainability teams documented, defensible savings for carbon reporting — without waiting on multi-year capital projects.

Clockworks distinguishes issues that can be fixed operationally from those requiring capital investment — capital projects are a distinct diagnostic result category in the platform. The longitudinal asset health record shows which equipment is chronically underperforming despite repairs, and quantified avoidable costs provide the financial evidence for repair-versus-replace decisions and capital business cases.

An equipment’s age tells you very little. Its measured condition tells you everything. Teams enter budget season with data, not anecdotes.

Hospitals, laboratories, and life-science facilities use Clockworks to continuously verify conditions in critical spaces — temperature, humidity, pressurization, and ventilation — and to catch deviations early, before they become compliance or safety events. The most severe maintenance priorities are reserved for potentially dangerous conditions, and the platform’s dated diagnostic and task records provide documentation that supports audits and regulatory reporting.

First-time fix rate — the percentage of work orders resolved on the first visit — is one of the highest-leverage metrics in facilities operations. Industry averages hover around 75%; best-in-class teams reach 90%. For a team completing 1,000 work orders per year, moving from 50% to 90% FTFR frees up roughly 800 hours of labor annually for proactive work.

Clockworks improves FTFR by sending technicians to the field with root cause already identified, co-occurring faults surfaced, and recommended actions pre-populated — eliminating the diagnostic guesswork that causes repeat visits and incorrect first fixes.

Implementation has three main phases: installing the software gateway and connecting it to your BMS and meters; configuring the digital model by classifying each building and piece of equipment and mapping BMS data points into the Clockworks data model; and validating data quality before diagnostics go live.

AI-assisted document parsing accelerates configuration by automatically extracting equipment ratings, control sequences, and occupancy schedules from project documentation. Early diagnostic results are available at go-live so your team can start acting on findings from day one.

Clockworks is offered as an annual software subscription scoped to the equipment connected to the platform — so cost scales with what is actually monitored, rather than building square footage alone. Because every portfolio differs in systems and instrumentation, pricing is quoted per project. Request a demo to get a scoped proposal.

Every client gets expert guidance and training from the Clockworks team, access to a dedicated support portal, and membership in the Clockworks Community of Practice, where users across organizations share workflows and results. The goal: every user on the platform has the support needed to transform building operations — without a steep learning curve.

Yes. Mechanical, engineering, and controls service providers — along with smart-buildings technology organizations and consultants — join the Clockworks Partner Program to deliver analytics-driven fault detection and diagnostics to their clients. Partners use the platform, including continuous monitoring and intelligent dashboards, to power data-driven service delivery across their client portfolios.

Partners use Clockworks to transform service delivery with remote, predictive monitoring; shift from time-based to analytics-based service contracts; demonstrate quantifiable results with impact reports and client dashboards; and increase contract profitability and pull-through revenue — competitive differentiation backed by diagnostics clients can see and trust.