People now ask, they do not browse
Owners, technicians and auditors put their question to an assistant. The answer has to come from the product itself, not from a manual nobody can find.
TwinAIReview drafts
Physical products are becomingsomething software can read.
Assistants and agents are moving into the places where products are sold, used, serviced and audited. They can only be useful there if each item carries structured, permissioned, trustworthy data. TwinAI is the layer that makes that possible on Twintag.
TwinAI · Item assistant
EM
TwinAI assistant
Home Battery ESS-4820 · Serial TT-8842-117
Conversation
The unit shows a temperature warning after installation. Is this covered by warranty and what should the technician check first?
Yes. This unit is within its 5-year warranty until March 2029. The warning matches a known ventilation clearance issue: check 30 cm clearance on the intake side and confirm firmware 2.4 or later. Both steps are in the installation manual, section 4.
Warranty terms · Document · v3.2Installation manual · Document · ENService history · 2 events
Answers are generated from verified item data and reviewed before publishing.
Generated experience
Warranty status cardPublished
Troubleshooting checklistDraft
Technician booking formDraft
Drafted from the same item record. Nothing goes live without your team's approval.
AI is only as good
as the identity underneath it.
The hard part is not the model. It is having an item that can be identified, resolved and read in the first place.
Agents need a source they can trust
A model can only be useful about a physical product if there is one record behind it, with a known owner, a known version and a known scope.
Most product data is still unreadable
Specifications sit in PDFs, service history in spreadsheets, compliance data in email. None of it is structured per item, so none of it is machine readable.
What changes when an item becomes AI readable.
A product stops being a label and becomes a record that software can reason about. The identifier resolves to one item, the item holds structured fields, documents and events, and every one of them has a permission attached. That is the difference between a chatbot guessing about a product category and an assistant answering about this exact unit.
Identified
One identifier, one item, resolvable anywhere.
Structured
Fields, documents and events instead of loose files.
Permissioned
Who may read what travels with the record.
TwinAI inside Twintag,
working on the item record.
It is not a separate assistant bolted on the side. TwinAI works on the same records TwinOS holds and the same experiences TwinStudio publishes.
TwinStudio · TwinAI
EM
TwinAI composer
Home Battery ESS-4820 · Owner experience
Prompt and context
Prompt
Build an owner experience for the Home Battery ESS-4820 with specifications, manuals, warranty status and a repair request flow.
Draft composed
HeroProduct name, serial, status
Specifications6 fields bound
DocumentsManual, conformity, datasheet
WarrantyCoverage and expiry
ServiceRepair request form
Draft ready in 8 seconds, every block editable
Compose experiences from item data
Describe the experience you need and TwinAI builds it from the live record, in the structure your team would have made by hand.
Answer the question behind the scan
Owners, technicians and auditors ask in their own words and get the answer their role is allowed to see.
Keep the record in shape
Mapping, enrichment and consistency checks across large catalogues, so item data stays usable at scale.
Why enterprises can let AI
near their product data.
Trust is a design decision. TwinAI inherits the scope, permissions and provenance that already live on the record.
Scoped to the record
TwinAI answers from the item record, not from the open web. If the data is not there, it says so instead of inventing it.
Permissions carry over
An owner, a technician and an auditor get different answers, because they already have different rights on the same item.
Every answer has a source
Fields, documents and events stay traceable, so a claim about warranty or material origin can be checked.
People stay in the loop
Drafts and generated experiences go through the same review and approval your teams use today.
Where TwinAI sits
in the stack.
Identity first, then the experience, then the systems around it. TwinAI runs across all three.
400M+
Products & assets connected
8+ years
Connecting physical products
99.95%
Platform uptime
Unlimited
Experiences per physical item