Automation for the plant and the field, built on site.
We automate manufacturing and field service work with three things. Agentic workflows run the multi-step jobs people do by hand, custom models learn from your own images, sensor data and history, and a knowledge layer answers from your own documents. Our forward-deployed engineers build it on site, on your data, measure it against your own baseline and leave it running on our platform.
The engineers come to the work
Forward-deployed means our software engineers work where the job is done, on site two to three days a week. They ride along with technicians. They sit with dispatch and the warranty desk. They stand at the line during a changeover and read a year of your work orders.
Every week they put a new version in front of the people who use it, and fix what those people say is wrong. That is how software ends up in production instead of in a pilot that nobody opens.
You get a working system, measured against your own numbers, instead of a slide deck and a recommendation.
75%
of manufacturers using generative AI in an RSM survey said they need outside help to get the most out of it.1
54%
of field service engineers say they would call a colleague when they run into a service issue.2
1 RSM US, manufacturing AI survey, 2025, 116 manufacturers using generative AI, accessed October 20, 2026.
2 Service Council, "Good, but not enough: reflections from the 2026 Smarter Services Executive Symposium", September 23, 2026, accessed October 20, 2026.
Start with a First Build
Every engagement starts with one problem, one process owner and one written measure, and delivers software your people can use by the end of week six.
Six weeks. Fixed scope, fixed price, from USD 35,000.
Working software on your data.
These prices are for custom engagements for larger operations, such as automating work in parts distribution or on a production line. They are not the price of Zylergy Field, Zylergy Solar or Zylergy Facilities, which shops and installers join through early access.
Weeks 1 and 2 · Scope
We ride along, interview the people who do the work, get access to the data and measure the baseline on your own numbers. We set the target with you in writing.
Weeks 3 to 6 · First working version
Our engineers build on site with your team. Three to ten of your users work with the software every week. Any model is tested on your own held-out cases.
End of week 6 · Decision
You see the software working, the baseline and the early numbers. You get a fixed price and a date for production. Then you decide: go to production, change the scope, or stop.
What you get
- Working software in the hands of your users
- Your baseline, measured, and a target agreed in writing
- Model results on your own cases, where a model is part of the job
- A fixed price, scope and date for production
What we need from you
- An executive sponsor
- A process owner for about a day a week
- One or two of the people who know the work best
- An IT contact for data and system access
- Three to ten users willing to work with early versions
Semiconductor fabs, defense suppliers and other secure sites need a security review before any data moves. For those, the First Build runs longer, and we set the length in writing before we start.
Scope, Build, Run
-
Scope
2 weeks, longer for secure sites
Site visits and ride-alongs. Interviews with the people who do the work. Data inventory and access. Security review. The baseline, measured on your own numbers. The target, agreed in writing.
You provide: an executive sponsor, a process owner, data access and site access.
It ends with: a written scope with the problem, baseline, target, data, plan and price.
-
Build
6 to 14 weeks by project, longer for secure sites
Engineers on site with your team. Working software in users' hands by week six of the engagement. A weekly working session with the process owner. Models tested on your held-out cases. Integration with your system of record. Cutover to production against written acceptance criteria.
You provide: a process owner for about a day a week, one or two subject matter experts, an IT contact and three to ten users.
It ends with: software in production and a measured result against your baseline.
-
Run
Ongoing, annual term
Support and monitoring. Models monitored against the evaluation set and retrained when accuracy drifts. A quarterly review of the measure against your baseline. Product releases. The next problem scoped from what we learned.
You provide: a product owner and admin users.
It ends with: a renewal decision based on the measure.
How it is priced
The First Build, which covers Scope and the first working version, is one fixed price, starting at USD 35,000. The rest of Build is fixed scope, fixed price and fixed length, agreed at the end of the First Build. Run is a subscription to the product your project runs on, with an optional engineer retainer for the next problem.
Who you work with
Two forward-deployed engineers
On site two to three days a week during Build. They own the integration, the workflows and agents, and the relationship with your users. One of them leads the engagement from Scope to cutover.
A machine learning engineer
Builds and tests the models, owns the evaluation set and the monitoring that carries into Run. Mostly remote, on site when we capture data.
A product manager, part time
Decides what is your configuration and what is standard product, and writes the handover.
A Zylergy executive
In Scope and at every decision point.
On your side: a process owner, the people who know the work best, an IT contact and a small group of users. We name the people and the hours for each engagement before Build begins.
Agentic workflows, custom models and knowledge
These are engagements we scope and build, in plants, in offices and in the field. Each reaches people as an application they use every day: on the line, at the desk or on a phone at the equipment. The measures are targets we set with you during Scope and track against your own baseline.
Agentic workflows
Agents that run a multi-step job your office or plant does by hand today, with a person approving wherever money, safety or a customer commitment is at stake.
Order entry and quote preparation
For customer service, inside sales and estimating teams at manufacturers that receive orders and requests for quote by email, PDF and portal.
Runs on Company Brain, connected to your ERP · Build 6 to 8 weeks
What we build: An intake agent reads incoming orders and requests for quote in any format and extracts part numbers, quantities, part revisions and dates. It checks them against your item master, price lists and open orders. A quoting agent drafts the quote from your costing rules and past quotes for similar parts. A person reviews and releases every order and quote.
What we measure: Hours from request to quote. Order entry errors that reach production. Touches per order.
Nonconformance and corrective action paperwork
For quality managers and quality engineers at plants under a customer or industry quality system.
Runs on Company Brain, in your environment · Build 8 to 10 weeks
What we build: An agent opens the nonconformance report from the inspection record, photos and operator notes, then pulls the drawing, specification and similar past cases. It drafts containment and the corrective action write-up in your format and routes it for approval.
What we measure: Days to close a nonconformance. Time quality engineers spend on paperwork. Repeat defects on the same part.
Service call intake and dispatch
For service and operations managers at commercial HVAC and mechanical contractors.
Runs on Zylergy Field · Build 6 to 8 weeks
What we build: An intake agent that works the call, email or web form and asks for what dispatch needs: site, unit, model and serial, a nameplate photo, the symptom and any error code. A diagnosis agent proposes the two or three likely causes from the manual and that unit's history, with the source. A parts agent checks van and warehouse stock. A dispatch agent proposes the technician by skill, certification, location and parts on hand. The dispatcher approves.
What we measure: First-time fix rate. Share of jobs that need a second trip for parts. Minutes from call to scheduled visit. Dispatcher touches per job.
Warranty claim pre-check
For warranty and service leaders at equipment manufacturers with dealer networks.
Runs on Zylergy Field and Company Brain
What we build: A pre-check agent that reads each incoming warranty claim, checks it for completeness and coverage, and returns an incomplete claim at once with exactly what is missing. A person reviews every claim before it is decided.
What we measure: Claim cycle time. Claims returned for missing information. Reviewer minutes per claim.
Inspection deficiency to quote
For branch managers and service sales managers at fire protection and life safety companies.
Runs on Zylergy Facilities · Build 6 to 8 weeks
What we build: An agent that reads the completed inspection report, photos and inspector notes, and turns each deficiency into a line item with its code reference. A pricing agent builds the quote from your price book and labor standards. A follow-up agent sends it, tracks approval and schedules the work. A person approves every quote before it leaves.
What we measure: Days from inspection to quote sent. Quote approval rate. Count and age of open deficiencies per customer. Quotes produced per estimator hour.
Maintenance job packages
For maintenance and reliability managers and planners at food and beverage plants.
Runs on Zylergy Facilities and Company Brain · Build 8 to 10 weeks
What we build: A planning agent turns each PM or corrective work order into a job package. It holds procedure steps from the OEM manual, lockout and tagout points, sanitation requirements, parts and kits with stock status, estimated duration and the last three jobs on that asset. A scheduling agent fits packages into sanitation and changeover windows. The planner approves each package.
What we measure: PM schedule compliance. Wrench time as a share of the shift. Jobs delayed for missing parts or information. Planner hours per week spent assembling packages.
Inverter alert to service job
For solar installers and O&M providers.
Runs on Zylergy Solar
What we build: Agents that turn inverter and battery telemetry into packaged, quotable service jobs. One agent separates the real fault from the noise. Another builds the repair package from the equipment manual and your own rates: the action plan, the bill of materials, the labor and the price. The office reviews the quote before it goes to the homeowner.
What we measure: Share of alerts that become a quoted job. Hours from alert to quote. Truck rolls per resolved fault.
Custom models
Models trained on your own equipment, images and history, where a general-purpose model is not accurate enough. Each one is tested on a held-out set of your real cases before it goes to the field.
Visual quality inspection on the line
For plant managers and quality leads where inspection is manual, sampled or late in the process.
Runs on Company Brain, on your plant network · Build 10 to 14 weeks
What we build: A vision model trained on images of your own good and defective parts, running on cameras at the station. It flags defects with the image and the reason, and learns from every call your inspectors confirm or overturn.
What we measure: Escape rate to the next station and to the customer. Inspection time per part. False reject rate.
Process anomaly detection from line sensors
For maintenance, process and reliability engineers with sensor and historian data that nobody watches in real time.
Runs on Company Brain, in your environment · Build 10 to 12 weeks
What we build: Models trained on your own process historian and sensor data learn normal behavior per machine and product. They flag drift before it becomes scrap or a stop, and attach the likely cause and the procedure from your documents.
What we measure: Unplanned stops. Scrap from process drift. Lead time between the alert and the failure.
Parts identification from a photo for the dealer network
For aftermarket and parts leaders at equipment manufacturers with a large installed base.
Runs on Zylergy Field · Build 10 to 14 weeks
What we build: A vision model trained on your catalog images, exploded views and labeled field photos. It identifies the part from a photo and matches it to the bill of materials for that exact model and serial. When it is not certain, it returns a ranked short list with the reason and shows the part on the exploded view. If labeled photos do not exist yet, we organize the capture with your technicians.
What we measure: Top-1 and top-3 accuracy on held-out field photos. Wrong-part return rate. Parts counter calls per week. Time from photo to confirmed part number.
Callback reduction model
For service leaders at elevator and escalator service companies.
Runs on Zylergy Field · Build 10 to 12 weeks
What we build: A diagnostic model trained on controller fault logs, callback history, maintenance records and parts replaced. It ranks the likely root cause for a reported fault on that specific unit and flags units at risk of a repeat callback. The mechanic gets the procedure and the controller manual page on the phone.
What we measure: Callbacks per unit per year. Repeat callbacks within 30 days. Top-3 cause accuracy on held-out tickets. Mean time to restore service.
Tool-down diagnosis
For semiconductor fabs, and the field service teams of fab equipment suppliers.
Runs on Company Brain, in your environment · Build 12 to 16 weeks, after a longer Scope for security review
What we build: A diagnostic model trained on your alarm and event logs and past repairs that ranks the likely cause of a tool-down and links the procedure from your documents.
What we measure: Mean time to repair on the target tool set. Unscheduled downtime hours. Repeat alarms within 7 days. Night-shift events closed without engineer escalation.
Assembly verification by camera
For quality engineering and plant managers at energy equipment plants: switchgear, transformers, generator sets.
Runs on Company Brain, on your plant network · Build 10 to 14 weeks
What we build: A vision model at one or two assembly stations that checks the build against the order: components present, torque marks, wire routing, terminal labels and nameplate data. The assembler sees the result before the unit moves. Every exception is stored with its image against the serial number.
What we measure: Defects found at final test per 100 units. Rework hours per unit. False-reject rate at the station. Inspection seconds per unit.
Custom applications
Applications for work that still runs on paper, spreadsheets or a generic form, built on our mobile app, web console and backend.
Digital startup and commissioning
For field service and startup services leaders at power and HVAC equipment manufacturers.
Runs on Zylergy Field · Build 8 to 10 weeks
What we build: A commissioning app that walks the startup technician through your procedure for that model. It captures every required reading with range checks, nameplate and installation photos and signatures, generates the startup report, registers the warranty and flags out-of-range values to the factory before the technician leaves site. Authorized third-party contractors use the same flow.
What we measure: Share of startups with a complete report. Warranty registrations at startup. Days from startup to report received. Early warranty claims attributed to installation or startup.
Asset register and contractor portal
For facilities directors at multi-site owners: school districts, hospital systems, property managers, municipal portfolios.
Runs on Zylergy Facilities · Build 6 to 8 weeks
What we build: A QR-tagged asset register built from a site walk and merged with your existing records. A compliance calendar for the inspections each asset requires. A contractor portal where vendors scan the asset, see its history, check in with current license and insurance, and close the work with photos. Audit packs per site or asset class in one click.
What we measure: Share of in-scope assets with a complete record. Overdue compliance items. Hours to assemble an audit pack. Share of contractor visits with verified credentials.
Line-side troubleshooting and shift handover
For plant and maintenance managers at automotive and EV plants and their Tier 1 and Tier 2 suppliers.
Runs on the Zylergy Field mobile app and Company Brain · Build 8 to 12 weeks
What we build: A mobile app for technicians and line leads to log a stop by voice, photo and fault code at the station. It structures the entry into cause, action and parts, searches past stops on that station and similar equipment, and shows what worked before with the source. At shift change it writes the handover from the logged events.
What we measure: Unplanned downtime minutes per shift on the target line. Repeat stops with the same cause within 14 days. Handover completion rate. Time from stop to restart for recurring faults.
Knowledge: the Company Brain
Manuals, specifications, work instructions, quality records, service records, technician notes and photos in one governed knowledge layer that answers on the floor and in the field with its source. Many engagements draw on it; the ones below are built around it.
Technical support knowledge layer for dealers
For technical support and service leaders at equipment manufacturers whose dealers call the factory for help.
Runs on Company Brain and Zylergy Field · Build 6 to 10 weeks
What we build: Your manuals, service bulletins, wiring diagrams, parts catalogs and solved support tickets, each linked to the models and serial ranges it applies to, with access rules by audience. Dealers and support staff ask a question and get the answer with the document and page or the ticket it came from. When there is no source, it says so, and the question goes to a support engineer whose answer is added.
What we measure: Dealer calls to factory support per 100 units in service. Time to answer. Share of answers accepted without escalation. Share of questions with no source.
Retiring-expert capture
For operations leaders and plant managers at aerospace and defense suppliers.
Runs on Company Brain, in your environment · Build 8 to 12 weeks
What we build: Recorded interviews and walk-throughs at the machine with your senior people, turned into draft procedures, troubleshooting trees and setup notes that the expert reviews and signs off. Together with your work instructions, nonconformance history and machine manuals, they become a knowledge layer newer staff can ask at the machine. Deployed in your environment, access limited to authorized people.
What we measure: Share of critical tasks with a reviewed procedure. Weeks for a new hire to reach independent sign-off. Questions to the experts per week. Repeat nonconformances on the target part families.
Senior-technician know-how for service contractors
For owners and service managers at HVAC and mechanical contractors whose most experienced technicians are close to retirement.
Runs on Company Brain and Zylergy Field · Build 6 to 10 weeks
What we build: Closed work orders, technician notes and job photos turned into a cited history for every unit and model you service. Short recorded walk-throughs with your senior technicians become reviewed troubleshooting notes. Technicians get the answers in Zylergy Field on the job, with the work order or note they came from.
What we measure: Calls from the field to senior technicians per week. Callbacks within 30 days. First-time fix rate on the target equipment. Share of questions with no source.
From engagement to product
What we build for you runs on our platform from the first day, so there is no migration at the end.
Zylergy Field, Zylergy Solar and Zylergy Facilities are what our method produces when we point it at one problem. The method brings agents that run the job, models trained on the equipment and the parts, and a knowledge layer that answers with its source.
We bring the same building blocks to custom projects anywhere in manufacturing and field service: the order desk, the quality lab, the line, the maintenance shop, the service van. Your project starts from those building blocks and ends on a product you can keep running after we leave.
At cutover, the product manager gives you a handover note: what is standard product, what is your configuration, what is specific to you, such as your models and connectors, and who supports each part.
From then on you run it on a subscription to the product your project landed on: Zylergy Field, Zylergy Solar, Zylergy Facilities, or platform modules such as the Company Brain. Your custom models are supported under the subscription with a stated retraining schedule.
What we learn building with you makes our products better for companies like yours.
Becomes product
Code, software connectors, AI agent designs, workflow templates and evaluation methods that hold none of your data. What Zylergy's vision models learn from photos and confirmed answers in our products.
Stays yours
Your documents, records, images and telemetry.
You own your data.
Your documents, records, images and telemetry remain yours. In our products, though, the photos technicians take and the answers they confirm help train Zylergy's vision models, which recognize parts. Our trust page describes this.
Your data stays where you decide.
By default it is processed on US cloud infrastructure, kept separate from every other customer's. On request we deploy in your own cloud account or on your premises.
What we build is set in the agreement.
Who owns the models and other work built in an engagement depends on the engagement. Code, software connectors, AI agent designs, workflow templates and evaluation methods that hold none of your data become part of our products.
Access follows the source.
Answers respect the permissions of the documents they draw on.
Every answer is traceable.
We cite sources and log actions and approvals. People approve wherever money, safety or a customer commitment is involved.
Security terms are agreed before the first engineer arrives.
Single sign-on, role-based access, encryption in transit and at rest, audit logs, a named security contact and a data processing agreement are written into the contract before any data moves.
Questions about working with us
Is this consulting or software?
Both, in that order. Our engineers work inside your operation for a fixed period and build software that runs in production on our platform. When the engagement ends, you keep using it under a subscription. You receive a working system and a measured result, not a report.
How long until something is in production?
The First Build puts working software in your users' hands in six weeks. Depending on the project, production comes 8 to 16 weeks after our first day on site. Projects that need a security review inside your network, such as fabs and defense suppliers, take longer, and we tell you so before we start.
What do you need from our team?
An executive sponsor, a process owner for about a day a week, one or two of the people who know the work best and an IT contact for access. We also need a small group of users willing to work with early versions. We list the exact data and people for each engagement before Build begins.
What happens if the target is not met?
The baseline and the target are agreed in writing at the end of Scope, and we measure against them together throughout Build. If the measure is not moving by the midpoint, we say so and agree to change the approach, change the scope or end the engagement, before the rest of the budget is spent.
Bring us one problem worth solving.
Tell us which costs you most: the intake that sends the wrong technician, the warranty queue or the line stop nobody wrote down. An engineer will reply.