Agentic layer for physical operations
The problems you don't see are the expensive ones. Many systems. One picture of the operation. Machines report. Agents triage. People act.
Operations run behind systems that do not talk to each other, and your staff are left with what falls in the cracks. AEDDIX reads those systems, models your operation and raises what an alarm misses. We read. We never write to a controller or a machine.
Taking on a small number of pilot projects at a heavily discounted rate.
AEDDIXLogisticsDelivery fleet
Scene 1 of 6
- SmartSensecargo temperature4.6 °C · rising
- SamsaraGPS · rear dooropen 4 min at stop 3
- Cooling unitcompressor · run timerunning 96 %
- OptimoRoutetomorrow's routes06:00 · 14 stops
Van 07 · cargo box · rear door · route
- Trend
- Cargo warms at every stop, faster each time
- Cause
- The rear door stays open while unloading
- From SmartSense · Samsara
- Impact
- 14 refrigerated drops before 09:00
Van 07's cargo leaves 2 to 8 °C before the last drops.
Avoided€2,10014 refrigerated drops kept
Value · Est. exampleMonth €2,100Year €11,400
Brief the driver before the 06:00 run
DispatcherTonight 21:00
Via MCPFiled as a task in your work system. Never a controller.
4 sources read · 2 joined1 raisedDispatcher · Tonight 21:00
Running today at
Customers with live devices and open tasks.
Connect what you run. We model the rest.
Your sensors and vendor accounts stay where they are. We read them, work out what is there and how it fits together, and you approve the result.
Four steps, two of them yours
- SmartSense••••READ-ONLY✓ CONNECTEDSamsaraVENDOR CLOUDTiveVENDOR CLOUDAEDDIX devicesVENDOR CLOUDRingCentralVENDOR CLOUDMicrosoft 365VENDOR CLOUDAWSYOUR CLOUDAzureYOUR CLOUDBusiness CentralVENDOR CLOUD
- YOU
- Pick a source and paste a read-only key.
- WE
- Check the connection.
- 12devices38sensors96signals
- FRIDGE-03temperature · door
- FRIDGE-04temperature · door
- COLD-ROOM-01temperature · humidity
- COMPRESSOR-Acurrent · run state
- YOU
- Nothing.
- WE
- Find every device, sensor and signal it reports.
- YOU
- One click.
- WE
- Draft a blueprint: the machines, groups, processes and rules that fit what we found.
- ✓Machines4 accepted
- ✓Sensor groups6 accepted
- ✓Processes2 accepted
- ✓Rules9 accepted
FIRST TASKDUE 18:00Cold Room 01 drifting toward 8 °C
Owner: pharmacy lead
RESOLVE IN AEDDIX · OR IN YOUR TOOLS
- YOU
- Accept, edit or reject each item.
- WE
- Put it live. Every problem it finds arrives with an owner, to resolve in AEDDIX or in your own tools.
Describe the next site. Then ask how it runs.
- Blueprint
- An agentic topology, inferred from your data. A person approves each item.
- Playbook
- An agentic set of policies, from finding to action. A person approves before it runs.
AIoT Copilot · Example Brewery
Step 1 of 5 · Describe
We're adding two fermenters to Cellar 2, FV-07 and FV-08, on the same glycol loop. Set them up.
Read devices Cellar 2
4 new sensors: 2 tank probes, 2 jacket valves
Draft blueprint Cellar 2 · v2
inferred from the new sensors and how FV-05 and FV-06 run
+2 machines · +4 placements · rules extended · replaces v1 once applied
3 of 3 changes accepted by the head brewer
Apply blueprint · Cellar 2 · v2
Adds FV-07 and FV-08 with their sensors, under the same rules as FV-05 and FV-06.
Cellar 2 runs four tanks now. The same rules watch all of them.
Next morning · 06:40
Anything I should know about Cellar 2?
Read trends Cellar 2 · this batch
FV-07 +1.8 °C over its recipe profile since Tue 02:00 · outside the band of its last 9 batches · valve reads open
Read triage verdicts Cellar 2
1 incident · FV-07 · recommended playbook: Investigate on site
Render figure FV-07 · against recipe profile
FV-07 · Batch 244 · recipe profile · last 9 batches
FV-07Recipe profile9 batches · ±2σOff profileFV-07 left its recipe profile at 02:00 and sits outside what nine batches have done, while its valve reads open: likely a stuck valve or an air lock on the new install. Triage recommends Investigate on site. Apply it?
Apply playbook · Investigate on site · FV-07
Dispatch a technician to inspect the asset on site.
Fig. 02 · The session is the product; the 3D view is ours
Live Copilot flows · our 3D view
Many sources. One model. Someone responsible for what it finds.
We read your systems, model how they relate, apply your rules and policies, let agents weigh what is left, and hand it to a named person.
Pick an industry, then step through its three cases.
- 01
Your systems
Each one reads normal on its own.
- SmartSenseCold Room 01 · 6.9 °C
- AEDDIX door sensor14 openings · usual 5
- Compressor controllerrunning 92 %
- Stock system€38k of vaccines inside
- 02
Modelled
Adds relations, rules and policies.
- Relations
- Cold Room 01 · door · compressor · stock
- Rule
- Vaccine storage 2–8 °C
- Policy
- Pharmacy lead approves stock moves
- 03
Triaged
Agents weigh it; noise closes here.
6.9 °C is inside the limit. With 14 door openings and a compressor at 92 %, the stock crosses 8 °C in about 40 minutes.
- 04
Your staff
One named person decides.
Pharmacy lead
Check the door seal and move the stock
by 18:00
- 05
Closed
Done where your team works.
Closed 17:48
Seal replaced. Back to 4.8 °C.
Via MCP
Filed as a task in your work system. Never a controller.
On record · Pharmacy lead
Fig. 01 · How a problem travels
Example events
It reads. Your people decide.
- Read-only from every source
- Runs in your own cloud account
- Your data, exportable from day one
- One named person decides, on record
The questions we hear first
- How do you stop the AI from making things up?
- Your rules decide the clear cases, not the AI. It looks only at the unclear ones, reads only your own data, and shows the readings behind every suggestion. When it is unsure, a person sees it as “needs review”. It never acts on its own.
- Where does our data go?
- It stays in your own space, kept apart from every other customer. It runs in your own cloud account, in the region you choose; running on your premises is in development. You can export all of it from day one.
- Can it change anything on our machines?
- No. We read. We never write to a controller or a machine. Tasks reach your own tools only through a hand-off you switch on.
- Who answers for a decision?
- Seen02:14Compressor running 92 % · 14 door openings
- Suggested02:15Check the door seal · needs review
- Decided06:12Approved, edited · Pharmacy lead
- Rulesv2Unchanged since
One named person, every time. Later edits do not rewrite the past.
Deployment
- Your cloud regionYou pick where it runsLive
- Agents on your premisesBring your own cloudIn development
- Tasks back into your work systemsOver MCP: a ticket or a work order, never a controllerIn development
Compliance
Compliance, deployment by deployment.
We claim no certification today; every deployment runs in your own account, mapped to the rules that apply to it.
- GDPREuropean Union · privacyPER DEPLOYMENT
- Québec Law 25Québec · privacyPER DEPLOYMENT
- PIPEDACanada · privacyPER DEPLOYMENT
- HIPAAUnited States · health informationPER DEPLOYMENT
- ISO/IEC 27001:2022Information securityPRIMARY TARGET
- SOC 2 Type IIService assuranceROADMAP
- ISO 28000 / 28001Supply-chain securityALIGNMENT
What is live and what is not
Plans, not promises.
- LiveTODAY
Sources, blueprints, the model, AI triage and playbooks, your rules, hand-offs to your tools, email, SMS and tests on a copy.
- In developmentUNDER WAY
An on-site runner that keeps working when the connection drops.
- Planned18 TO 24 MONTHS
Approved actions back to a machine. Only after a person signs off, and only after a test on a copy.
Running today, in places where a missed problem costs real money.
Two years in at a brewery with sixteen fermenters. Pharmacies, fleets and kitchens in four countries. The same product runs a cellar, a pharmacy fridge and a truck: one model of an operation, across five industries. The figures below come from the platform and carry their date.
“Two years in, and AEDDIX systems have become integral to our success, helping ensure the quality, consistency, and reliability of our operations. Over 16 fermenters and our taproom, the platform has proven simple to use and highly effective in supporting our team day to day.”
- industry sectors in production
- 5
- live devices
- 275
- countries
- 4
- source connectors live today
- 9
Production figures as of 18 October 2026.
Why it compounds
Every source you connect, every rule and policy you approve and every decision on record adds to one model of your operation. It gets sharper the longer it runs; a new tool would start from zero.
“From infrastructure migrations and custom application development to cold chain monitoring and operational digitization, the team has consistently supported our vision every step of the way.”
“With telematic GPS fleet monitoring and custom cold chain management devices, we now benefit from stronger traceability, better operational control, and greater confidence in our logistics environment.”
“Over the years, the platform has delivered real value through stronger customer engagement, better data transparency, and reliable operational support.”