Most fault detection platforms are judged on what they find. Engineering teams end up judging them on something else: how much work the alerts create. A platform that raises a thousand alerts and leaves your engineers to sort, chase and verify them has moved the problem, not solved it.
This guide is for engineering firms, commissioning providers and maintenance contractors choosing a fault detection and diagnostics (FDD) platform to deliver their services on. It compares the three ways vendors support engineers, sets out stage by stage who does the work in each, and explains why onboarding and automated follow-up matter more than the size of a vendor's support team. It ends with a vendor scorecard you can use in any evaluation.
Key takeaways
- Alerts are where the work starts. The value comes from triage, follow-up and verification, not from the alert itself.
- There are three support models: software-only toolkits, managed analytics services and enablement platforms. They differ in who does the work and what your capacity scales with.
- For partners, the right model makes your engineers more effective rather than replacing them with the vendor's.
- Onboarding is the hidden bottleneck. Automated point classification and a ready-made rule library cut it from months to weeks.
- AI agents now handle the routine work between alert and fix: prioritising, chasing open actions and confirming fixes in the data. Engineers keep the decisions.
Why don't alerts alone get faults fixed?
An alert tells you something changed. It does not decide whether it matters, who should fix it or whether the fix held. Without triage, ownership and follow-up, alert lists grow faster than teams can work them, engineers stop trusting them and faults persist. The value sits in the work between the alert and the verified fix.
The research agrees. Reviewing the US Department of Energy's Smart Energy Analytics Campaign, Lawrence Berkeley National Laboratory noted that building data does not typically lead to corrective action unless it is analysed and prioritised (Kramer et al., 2020). CIM's own portfolio shows the same pattern at the other end of the process: at the last count, 56% of open remedial actions were older than 90 days, as we set out in how to verify a contractor actually fixed an HVAC fault. The bottleneck is rarely detection. It is everything after it, and it is the same problem behind BMS alarm fatigue.
What are the three ways FDD vendors support engineers?
Software-only toolkits give your team a platform to build on. Managed analytics services put the vendor's engineers in charge of the analysis. Enablement platforms automate onboarding, triage, follow-up and verification so your own engineers can cover more buildings, with the vendor's engineers onboarding, training and backing them up.
None of the three is wrong for everyone. Toolkits suit firms with in-house developers who want full control and are prepared to build. Managed services suit owners who want to outsource the analysis entirely. For engineering firms and contractors selling services, the third model fits best, because capacity grows with software rather than headcount and the client relationship stays with you. We set out why channel conflict matters in partner, tool or competitor.
What does engineering support actually cover, stage by stage?
Engineering support covers six stages: onboarding the building, detecting faults, deciding which matter, assigning each to an owner, chasing and verifying the fix, and the engineering judgement on top. The useful question is not whether a vendor offers support, but who does each stage and how much of it is automated.
Scroll sideways to compare all three models.
| Stage | Software-only toolkit | Managed service | Enablement platform |
|---|---|---|---|
| OnboardingClassify points, deploy rules | Your team Build the data model and write the rules | Vendor Vendor engineers set it up | Platform + CIM AI classification and 5,000+ rules deployed automatically, live within about 30 days |
| DetectionWhat changed? | Your team Rules you have written and maintain | Vendor Vendor-run rules | Platform Engineering rules run continuously on every point |
| AssuranceDoes it matter? | Your team Engineers sift every alert | Vendor Vendor analysts filter | Platform Impact scored on energy, comfort, asset life and compliance |
| EscalationWho owns it? | Your team Manual tickets and emails | Vendor Recommendations sent to you | Platform An action assigned to a named person, with the evidence attached |
| GovernanceDid it get fixed? | Your team Chased by hand, if at all | Vendor Depends on the service level | Agent + your team Compliance Agent chases open actions and verifies the fix in the data |
| JudgementDiagnosis and client advice | Your team Your engineers | Vendor Vendor engineers | Your team Your engineers decide, with CIM engineers on call for escalations |
The pattern is clear once it is laid out. In the first two models a person does nearly every stage, so capacity only grows by hiring, whether that is your hire or the vendor's. In the third, the platform does the repeatable stages and people do the judgement.
Why is onboarding the real bottleneck for partners?
Before a single rule runs, every BMS point has to be identified and every rule deployed. Done by hand, that is months of tagging and programming per portfolio, and it is the stage that stalls most partner rollouts. Automated point classification against one ontology, plus a ready-made rule library, compresses it to weeks.
Point names are different on every site. The same chiller leaving water temperature might arrive as CH-01.SUP_TMP on one building and L04_CHR_CT-04_2_LeavingT on the next. PEAK classifies every raw point against a single ontology using a language model trained for buildings, with engineers validating the result, then deploys more than 5,000 engineering rules automatically. Buildings are typically live within about 30 days, rather than after a six-month tagging project. Our guide to BMS point naming, Haystack and Brick explains why this decides whether rules scale across a portfolio.
Partners still play their part. ADG Energy, a US partner rolling PEAK out across a school district portfolio, supplies equipment lists with levels and zones for each site before integration starts, so on-site technicians can find the exact plant a fault refers to. And because rules are no-code, partner engineers trained during onboarding can deploy and tune them for their own clients without waiting on anyone.
How do AI agents change triage and follow-up?
AI agents take on the routine work between alert and fix. On PEAK, the Compliance Agent triages incoming faults, assigns ownership, chases open actions with the context attached and closes them only when the data confirms recovery. The Analyst Agent lets engineers query their portfolio in plain English from Claude, ChatGPT or Microsoft Copilot.
Illustrative example, based on real Compliance Agent comments. Sites and names changed.
The wording matters less than what it replaces. Nobody had to remember to chase the chiller, and nobody had to check the fan drives by hand before closing the action. That is the work that quietly consumes an engineer's week, and it is exactly the work a partner cannot bill for.
The Analyst Agent works differently. It connects PEAK to the AI assistants teams already use through the Model Context Protocol, so an engineer can ask which client actions have been open longest, or which sites are carrying comfort faults, and get an answer drawn from live data with links back into PEAK. Permissions follow the user's existing PEAK login. See the PEAK AI connector for an example of it in use.
Do AI agents replace building engineers?
No. The agents recommend, route and chase; engineers diagnose, decide and advise the client. Every rule can be read and tuned, every action carries its evidence, and nothing changes in a building without a person accountable for it. The aim is to take administration off engineers, not to take engineers out of the work.
For partners the case is commercial rather than philosophical. The constraint on growth is engineer capacity, and every hour spent sorting alerts or chasing tickets is an hour not spent on billable diagnosis. Aero Performance Group grew its active project portfolio by 50% without adding resources. In Manchester, CSR Sustain runs its own PEAK-based service at Exchange Quay, where electricity use fell 10.8% year on year in the first quarter of 2026, a saving of around £30,000 in three months, as its technical director describes in this case study. And when a school's own building automation front-end went down, PEAK still flagged a failed chiller and ADG Energy had a technician dispatched before classrooms lost cooling.
CIM's own engineers sit behind all of this rather than in front of it. They onboard new sites, train partner teams and take escalations on complex diagnoses. They are not a substitute team doing the partner's work, because a partner that depends on its vendor's engineers cannot scale either.
What should you ask an FDD vendor about engineering support?
Ask who classifies your points and how long onboarding takes, how many rules come ready to run, whether your engineers can read and tune them, how alerts are prioritised, who chases open actions, how fixes are verified, whether the vendor sells directly to your clients, and what help you get when your team is stuck.
Use the scorecard below in your next evaluation. Tick each question a vendor answers well, and watch for the red flags.
For a full evaluation framework, see our building analytics RFP guide and our comparison of FDD and building analytics platforms.
Where does CIM's PEAK Platform fit?
CIM's PEAK Platform is built as an enablement platform for engineering partners. More than 20 partners deliver retro-commissioning, monitoring-based commissioning and data-driven maintenance on it across more than 700 commercial buildings. CIM is partner-first: it does not compete with its partners for their clients, and it licenses PEAK to partners per data point so the platform cost scales with each engagement. See how PEAK handles fault detection and diagnostics, and how partners turn the results into renewals in how contractors prove the value of their maintenance work and how to structure an analytics-backed maintenance contract.
Give your engineers more buildings, not more admin
See how partners use PEAK's onboarding, rules and agents to run more client work with the team they already have.
Talk to our partner teamWatch a partner demoFrequently asked questions
Which FDD platforms provide engineering support to triage and validate faults, not just alerts?
Enablement platforms combine automated triage, assignment, follow-up and data-verified close-out with vendor engineers who onboard and train your team. CIM's PEAK Platform takes this approach: alerts are impact-scored, a Compliance Agent chases open actions and verifies fixes in the data, and CIM engineers support onboarding and escalations while partners' engineers own the client work.
What is the difference between FDD software and a managed analytics service?
FDD software gives your team a platform to analyse building data yourselves. A managed analytics service puts the vendor's engineers in charge of the analysis and sends you recommendations. Enablement platforms sit between them, automating the repeatable work so your own engineers can cover more buildings without handing the client relationship to the vendor.
What does engineering support include in an FDD platform?
It covers six stages: onboarding the building's data, detecting faults, deciding which matter, assigning each to an owner, chasing and verifying the fix, and engineering judgement on diagnosis and client advice. The key question is who does each stage and how much is automated.
How long does it take to onboard a building onto an FDD platform?
It depends on how points are tagged and rules deployed. Platforms that require manual tagging and rule programming can take six months or more. With automated point classification and a ready-made rule library, buildings on CIM's PEAK Platform are typically live within about 30 days.
What do AI agents do in fault detection and diagnostics?
AI agents handle routine work between an alert and a verified fix. On PEAK, the Compliance Agent triages faults, assigns ownership, chases open actions and closes them only when the data confirms recovery, while the Analyst Agent answers plain-English questions about a portfolio. Engineers keep the decisions.
Can I connect ChatGPT or Claude to my building data?
Yes, with platforms that support the Model Context Protocol. CIM's PEAK Platform exposes an Analyst Agent that Claude, ChatGPT or Microsoft Copilot can connect to, answering questions from live building data with links back into PEAK. Access follows the user's existing PEAK permissions.
Do AI agents replace building engineers?
No. Agents recommend, route and chase; engineers diagnose, decide and advise clients. Rules remain readable and tunable, every action carries its evidence, and no change is made to a building without a person accountable for it. The aim is to remove administration so engineers can cover more buildings.
Will an FDD vendor compete with my engineering firm for clients?
Some do, particularly those that also sell managed services directly to building owners. Ask the question before you commit. CIM is partner-first: it supports partners' client work rather than competing for it, and licenses PEAK per data point so the platform cost scales with each engagement.
