TagIQ · business overview

The register gates handover. Fixing it does not have to break anything.

A tag register is not documentation — it is the document that gates commissioning and, on most contracts, payment. Everyone already knows theirs does not meet its own naming standard. The reason it never gets fixed is that renaming tags breaks every drawing, datasheet and maintenance reference pointing at the old names, so it is rational for the team not to touch it.

TagIQ renames the register to the standard while every old reference still resolves, fills the gaps with an engineer approving each change, and hands the result to the systems that operate the plant. One working session rather than one quarter — and nothing on either side of it is replaced. It ‘just’ gets better.

Handover accepted 15 → 165 of 180 records the receiving maintenance system took, before and after — its verdict, not ours
Rework avoided 165 records each one otherwise opened, researched and retyped by hand — at four minutes apiece, weeks of engineering
Field completeness 77.2% → 84.8% across the nine fields the handover is judged on — measured before and after, not estimated
References broken 0 every renamed tag kept its source identifier, so drawings and the owner's register still resolve

Measured live in the recorded demo against an eight-unit gas plant, not modelled — the same run the presenter script quotes. Your register will produce its own numbers.

Part one

Where it sits

Read the arrows, not the boxes. The systems on either side are the customer's existing investment, and they are the same after as before.

Design side Unchanged
Engineering design tools 3D plant design · P&ID authoring
instrument index · electrical schematics
Document control the issued package, by revision
Discipline deliverables process — P&IDs, line & equipment lists
instruments — index, loop diagrams, datasheets
piping — isometrics, specs
electrical — single-line diagrams, cable schedules
3D — IFC, OBJ, native models
TagIQ · the checking step
1 Read REST API, CSV, XLSX, JSON, DEXPI Proteus XML, drawings, models
2 Check naming convention per class · field completeness · class and parent structure · CFIHOS / DEXPI / 3D audit · required documents by project phase
3 Reconcile against the owner's register, against the P&ID, against the model — matched on tag name or source alias
4 Repair, under human control AI proposes, a person approves — attributed, audited, and reversible field by field
Operations side Unchanged
SAP PM / CMMS functional locations · TPLNR · PLTXT · EQART
Owner's AIM asset information register · read as an extract
Pull — TagIQ reads; nothing is written back up this line Delivery — per record, on approval Verdict — the receiving system accepts or refuses, with a reason
Part two

FAQ — what that answers

The objection is never "does it work". It is "so we are replacing something, and who owns the data now?" — followed, once that is settled, by whichever version of the question fits the situation the buyer is actually in.

QuestionAnswer
Does it write back into our design system? No. The issued package is pulled and read. Nothing goes back up that arrow.
So it becomes the new master? No. It holds the register for the handover window — the interval where the data is between two owners and today belongs to nobody.
What happens to our tag numbers? Renaming keeps the source identifier as an alias, so every drawing, purchase order and SAP reference that already exists still resolves.
Who decides what is correct? For handover, the receiving system does. TagIQ submits per record and logs SAP's own accept-or-refuse reason. It does not grade its own homework.
What if the AI gets it wrong? Every change names who proposed it and who approved it, and any run can be taken back field by field.
What do we have to switch off? Nothing. Both ends keep doing what they do. The manual checking step in the middle is what changes.
Our plant is running — there is no project. Does this still apply? That is the more common case. The register is read out of the system that already holds it, measured, corrected and augmented, and the richer version goes back to the same system. No handover required.
We have several contractors. Will they see each other's data? No. Each package sits in its own project boundary, enforced where the data is read rather than requested politely — a query that does not name one project is refused. What crosses the boundary is the score and the named gaps, which is what the owner is entitled to and nothing more.
Can we try it without connecting it to anything? Yes, and it is usually the right first step. One extract in, a read-only assessment out: how far the register is from the standard it claims to follow, and what remediation actually consists of. Nothing is configured to receive, so nothing can be written.
What if we have no engineering register — tags start life in a spreadsheet? Then TagIQ can be where they start instead. Numbers are issued from the class's own convention and a project counter rather than typed, and the same rulebook that grades a contractor's package grades yours. What that replaces is the spreadsheet, not a system.
If you hold our register, are we locked in? It leaves whole, in open formats — CSV, Excel, JSON, a printable register — or per record into a receiving system. Holding the data for a project is a service, not a hostage position, and you are right to ask which one it is.
Is this only useful at handover? No. Handover is where the pain is sharpest and easiest to price, which is why the demo runs it. The cases above are illustrations rather than a catalogue — the same checking step is worth having anywhere a register has to be trusted by someone who did not build it.

When presenting this, point at what stays before pointing at what is new. The boxes on the left and right are the customer's existing investment, and they are the same after as before.

Part three

Use cases

The handover case above is the one the demo runs. It is not the only one, and these are not a menu — what changes between them is only where the data comes from and who it goes back to. Anywhere a register crosses a boundary, or has to be trusted by someone who did not create it, is a candidate.

01

Brownfield round-trip — the AIM system is both ends

Operating owner

No project, no handover. The register already lives in the owner's asset information system and has decayed for fifteen years. TagIQ reads it, measures it, corrects and augments what can be inferred, and delivers the richer dataset back to the same system it came from.

Source Owner's AIM REST API · CSV · XLSX
as-built drawings · models
pull
TagIQ Check · reconcile · augment naming per class · completeness
class & parent · CFIHOS audit
document coverage per tag
enriched
per record
Destination Owner's AIM the same system
· enriched fields
· aliases preserved
What's different Source and destination are one system, so the round trip is configured deliberately — an export target pointed back at the system that fed it. Nothing about that is automatic, and it should not be.
Why it's safe Every corrected tag keeps its original identifier as an alias, so work orders, permits and history already written against the old name still resolve. On an operating plant that is not a nicety — it is the whole risk.
The measure A baseline is captured before anything changes, so the improvement is a delta the owner can audit rather than a claim.
02

One owner, several EPCs, one rulebook

Owner · programme level

Three contractors deliver three packages in three house styles, each one arguing its data is fine. Each package lands as its own project, and every one is measured against the same conventions, the same required documents and the same standards audit — so the packages become comparable even though the contractors are not.

Package A EPC A API · DEXPI P&IDs
Package B EPC B XLSX · DWG/DXF
Package C EPC C CSV · IFC models
pull
per package
TagIQ One project per package,
scored on one rulebook
completeness · convention conformance
safety-critical gaps
required documents, by phase
score &
punch list
Destination The owner's acceptance decision per package: a number,
and the named gaps behind it
What's different Nothing is delivered onward yet. The output is a gate: what each package is missing, by class and by tag, before anyone signs for it.
Why it lands "Seventy percent complete" is unarguable and useless. "Ten piperuns with no isometric, and fabrication cannot start" names the tags and survives the meeting.
Scope isolation Each package is a separate project boundary, enforced in the engine — every query against project data is rejected unless it filters to one project, and that applies to the AI's own SQL as much as to a person's. Three competing contractors can feed the same owner without either of them appearing in the other's view. What crosses the boundary is the score and the named gaps, which is precisely what the owner is entitled to and nothing else.
03

Assessment only — nothing is written anywhere

First engagement

The lowest-commitment version, and often the right first conversation. A copy of the register goes in and nothing comes out except a verdict: how far it is from the standard it claims to follow, which safety-critical items carry gaps, and what the remediation actually consists of.

Source A register extract one CSV is enough
drawings optional
pull
TagIQ Read-only audit agent autonomy set to read-only
CFIHOS / DEXPI / 3D checks
scorecard · coverage · punch list
report
only
No system written A findings report PDF register · CSV · XLSX
nothing pushed anywhere
What's different There is no destination system at all. The agent's autonomy control has a read-only position, and no export target is configured — so there is nothing to switch off, and nothing to be nervous about.
Why it works as an entry It answers "is our register as bad as we think" with a number the buyer's own standard produced, before anyone has to trust the writing half of the product.
04

Revamp — new scope merged into a live register

Operating owner + contractor

A modification delivers a few hundred new tags into a plant that is running. The hard part is never the new tags; it is knowing which of them are genuinely new, which restate something the register already has under a different name, and which contradict it.

Live register Operating AIM / CMMS as-built · in service
New scope Revamp package DEXPI P&IDs · models · lists
pull
both sides
TagIQ Reconcile, then merge matched · drifted
only-in-revamp · only-in-register
matched on name or alias
approved
delta
Destination The operating register only what a person approved
· reversible as one run
What's different Two live sources that disagree, rather than one source and one destination. The reconciliation is the product here — the import is trivial.
Why it's safe An import lands as a run that can be reverted whole. If the merge is wrong, the operating register goes back to what it held before, rather than being cleaned up by hand afterwards.
05

TagIQ as the tag management tool

Greenfield · no incumbent register

The cases above all preserve something that already exists. This one is different: there is no upstream register to protect, because the tags do not exist yet. They are created in TagIQ, numbered by the project's own convention, reviewed by name, and handed to the systems that will operate the plant. What this replaces is not a system — it is the spreadsheet.

No upstream system Tags originate here the project's classes,
conventions and required
deliverables come first
created,
not imported
TagIQ Author · number · review name built from the class rule
sequence issued atomically
Draft → Pending → In Review → Approved
owners, open questions, punch list
approved
per record
Destination SAP PM / CMMS functional locations
Destination Contractors & document control CSV · XLSX · JSON · PDF register
What's different This is the one scenario where TagIQ holds the register rather than checking someone else's. That is a deliberate choice, made where there is no engineering register to defer to — and the operating systems downstream are still exactly as unchanged as everywhere else.
Why numbers hold A tag number is issued, not typed: built from the class's own convention and taken from the project's counter atomically, so two engineers working at once cannot mint the same tag and nobody invents a format on a Friday afternoon.
Why it stays honest The same rulebook that grades an EPC's package grades this one. A register authored here is measured against its own conventions and its own required documents, by phase — so "we made it ourselves" is not a reason it passes.
Nothing is captive The register leaves whole and in open formats — CSV, XLSX, JSON, a printable PDF register, or per record to a receiving system. Holding the data for the project is a service, not a hostage position, and a buyer is right to ask which one it is.
Part four

What the AI agent actually does

It sits inside the middle box in every shape above. What makes it usable near master data is not how clever it is — it is how little it is allowed to do without being told.

Setting one Read-only It can look and answer. It cannot change anything, and with no destination configured there is nowhere for a change to go.
Setting two DEFAULT Ask before changes It works out the change, shows it resolved against your live data, and waits. This is the position the demo runs in, and the one most customers stay in.
Setting three Autonomous It does the work unattended — and the entire run can still be taken back afterwards, field by field.
Reads · always allowed
The register itself counts, filters and live queries — so an answer is something you can check
Drawings and models what a tag appears on, and what it stands next to
The other systems the owner's extract, and the delivery log of what was refused
TagIQ agent · orchestrator It does not answer in one move. It works.
  1. 01Reads what you asked for
  2. 02Chooses a tool and uses it
  3. 03Looks at what came back
  4. 04Goes round again, or answers
↻ repeats until the job is done — and every step is listed on screen as it happens
Approval gate Nothing on this side happens until the autonomy setting allows it — and in the default position, until a person approves the exact change.
Writes · only through the gate
Change tags shown first as a resolved diff — every tag, every field, now and after
Deliver to the receiving system per record, and it reads back what was refused
Publish a report or chart drawn from the query it just ran, not written from memory
Remember a project rule stored against the project, so the next session starts knowing it

And the boundary it cannot talk its way past: the agent can only see the project it is working in. That is enforced where the data is read, not asked for in its instructions — a question about another project's register comes back refused, however it is phrased.

Part five

Core concepts — the same in every use case

These are not separate products. The differences are entirely in the wiring at each end; the core concepts in the middle are the same every time.

What is identical in every one

These are not separate products. The differences above are entirely in the wiring at each end; the core concepts in the middle are the same every time, and they are what makes any of it acceptable near master data.

Reads are pulls TagIQ fetches; a source system is never written to unless someone deliberately configures it as a destination.
Identity survives A renamed tag keeps its source identifier as an alias, and reconciles on either — in every scenario, indefinitely.
A person approves The agent proposes a resolved diff — every tag, every field, before and after — and waits. Autonomy is a setting, not an assumption.
Everything is attributable Who proposed it, who approved it, when. Human and AI writes land in the same audit trail, with no side channel.
Any run comes back An agent run or an import reverts field by field to what each one held before it started — and the revert is itself audited.
Scopes stay separate Project isolation is enforced at query level, so commercially sensitive packages — competing contractors, licensor IP, a client's register — never share a view.
The receiver decides Where there is a destination system, it accepts or refuses per record with its own reason. TagIQ does not grade its own homework.