Skip to content

Construction safety software

Construction safety software for PPE detection alerts

Camera detections become alerts, investigations and reports, with an AI assistant.

Ailixar builds PPE detection alert software for construction safety: it turns camera detections, such as a missing hard hat or a forklift too close to a worker, into rule-based alerts, investigations and reports. Our own product, SiteGuard, is the web control room for this, with zone checks, an alert inbox and a site-scoped AI assistant. The detection models run separately and send detections in through an API. Prices start at $2,500 for an AI assistant in a safety system you already have and $4,500 for a first AI workflow, which usually takes three to six weeks.

Book a 30-minute call

The risks behind the alerts

Construction and extraction jobs accounted for about one in five US workplace deaths in 2024. That year, "construction and extraction workers experienced 1,032 fatalities," out of 5,070 fatal work injuries recorded nationwide, according to the Bureau of Labor Statistics. Fatal falls, slips and trips among these workers fell "to 370 in 2024 from 400 in 2023."

Head injuries are part of that picture. The National Institute for Occupational Safety and Health (NIOSH) says "the construction industry has the greatest number of both fatal and nonfatal WR TBI among U.S. workplaces," where WR TBI means work-related traumatic brain injury. From 2003 to 2010, "2,210 construction workers died because of a TBI," according to the same NIOSH bulletin.

These are the hazards camera models watch for: missing hard hats, work at height and vehicles close to people. Software doesn't prevent an injury by itself. Its job is to make sure a detection reaches someone who can act.

Why PPE detection alone isn't enough

A detection model can flag a missing helmet, but a flag isn't a safety process. Site safety teams can't watch every camera, and a raw stream of detections is just more noise.

Detections need to land somewhere useful:

  • Matched to the right zone on the site
  • Turned into an alert only when a rule says so
  • Followed through to an investigation and a report

Managers also want to ask plain questions, such as "what happened on the north tower this week?", without building a report by hand.

How SiteGuard works

SiteGuard is the control plane between the vision models and the people who act on alerts. It's our own internal product, built on Laravel 13, React 19, OpenAI through Laravel AI, and queues. Full details are in the SiteGuard case study.

Each detection follows the same path:

  1. The camera sends a detection. Each camera has its own ingest token, which can be revoked on its own.
  2. Duplicates are rejected. The detection API drops repeats so one event doesn't become ten alerts.
  3. The snapshot is stored, so every alert has an image behind it.
  4. The zone is checked. SiteGuard works out which of the zones set for that camera the person or vehicle is in.
  5. Rules decide. A queued rules check opens a new alert or updates an existing one.
  6. People act. Staff work an alert inbox, acknowledge or dismiss alerts, link them to investigations and export reports.

The AI assistant for safety managers

The SiteGuard assistant answers questions about one site from the site's own records. It uses ten read-only tools over alerts, trends, cameras, rules and investigations.

  • Site-scoped. It only sees the site it's asked about.
  • Read-only. It can report and summarize, but it can't dismiss alerts or change rules.
  • Logged. Every AI call is recorded.

A manager can ask what happened on a part of the site this week and get an answer built from the alerts and investigations on record.

What SiteGuard doesn't do

We would rather you know the limits up front.

  • It doesn't include the detection models. They run in separate vision workers outside SiteGuard and send detections in through the API. Choosing and running those models is separate work.
  • It's our own product. We show it as an internal build, not a client deployment, and we don't publish accuracy figures for it.

Safety software for your sites

There are three common starting points, depending on what you already have.

  • You have cameras and a detection model, but alerts go nowhere useful. A control room like SiteGuard, with zones, rules, an alert inbox and investigations.
  • You already have a safety system with an API. An AI assistant added to it, so managers can ask questions in plain language.
  • You track inspections and certificates on paper or spreadsheets. Safety and compliance software that turns them into alerts with audit trails. Our NEMT platform tracks compliance documents with expiry alerts in the same way.

The same alert-and-follow-through pattern appears in our school pickup safety system, where supervisors can hold or freeze pickup in an emergency and security staff log incidents.

Data and access

Safety data often includes images of workers, so access is limited by design.

  • Each camera has its own revocable token
  • The assistant reads through read-only tools scoped to one site
  • A model receives only what a task needs, and we never use your data to train models
  • Every AI call is logged

Read more on security and data.

Cost of construction safety software

Our prices are published, and you approve a fixed price before work starts.

  • AI Workflow Audit, $950: two weeks to map your alert workflow, check your cameras, data and systems, and get a fixed quote. Credited if you build within 60 days. See the audit.
  • Feature Build, from $2,500: an AI assistant or similar feature added to a safety system you already have.
  • First AI Workflow, from $4,500: one agent or workflow connected to your systems, with tools, tests, logging and handover.
  • Extended Build, from $6,500: a wider build with an admin console or a Flutter mobile app for site staff, and more integrations.

See pricing for every plan, or read how we approach AI agent development.

Sources

Work

Related work.

Builds from our portfolio that use the same pieces.

Read next

Keep reading.

FAQs

Questions about cost, timing and ownership.

Cost, timing, ownership and your data.

  • No. The detection models run in separate vision workers outside SiteGuard. SiteGuard receives their detections through an API, checks zones and rules, and turns them into alerts, investigations and reports.

  • No. SiteGuard is our own internal product, and we label it that way. It shows how we build safety alert software and AI assistants, and we can build the same approach for your sites.

  • No. The assistant is scoped to one site and answers through ten read-only tools over that site's alerts, trends, cameras, rules and investigations. Every AI call is logged.

  • A source of detections, such as vision models watching your cameras, and clear rules for when a detection should become an alert. The AI Workflow Audit checks both and gives you a fixed quote.

  • Our published starting points are $950 for an audit, $2,500 for one feature in an existing system, $4,500 for a first AI workflow and $6,500 for a wider build. Model and hosting costs are billed at cost.

  • Often, yes, if the system has an API or a database we can reach. A Feature Build starts at $2,500. The AI Workflow Audit, $950, checks access first and gives you a fixed quote.

Contact

Tell us what you want to build.

A few lines is enough. We'll reply within one business day with next steps or a time to talk.

We use these details only to reply to you. See our privacy policy.

We only use the cookies this site needs to work, such as keeping forms secure. Cookie policy