Application Engineering Intelligence
The working model of the industrial supply chain behind every answer. It cites its source or it asks.
A request read into a structured table: 4 lines, each with its quantity, and 15 attributes, each cited to the customer's own words.
Packs chosen before any model runs
7 categories over 15 knowledge packs, the same for every workspace.
A trigger scan, not a guess
A deterministic scan of the request's own words picks up to three packs, five on escalation, before any model runs.
BF-4472 rev A, gusset, 6mm S275JR, 40 off, plasma cut, no folding
- Cutting tools
- Abrasives and grinding
- Machine tools and CNC
- Workholding and toolholding
- Hydraulics and pneumatics
- Power transmission
- Welding and joining
- Sheet metal and fabrication
- Fasteners and fixings
- Metrology and inspection
- Coatings and heat treatment
- Materials and stock
- Plastics and composites
- Fluids and lubricants
- MRO and workshop
Profiling line: 2 of 15 packs selected before any model runs
Designations route on their own
A standard designation, such as a BS 970 steel or an ISO viscosity grade, reaches its pack even when no product is named.
Choosing costs nothing
The scan is plain matching, so picking the packs spends no model time and gives the same answer every time.
- Cutting tools and abrasives
- Cutting tools, Abrasives and grinding
- Machines, workholding and automation
- Machine tools and CNC, Workholding and toolholding, Hydraulics and pneumatics, Power transmission
- Fabrication and joining
- Welding and joining, Sheet metal and fabrication, Fasteners and fixings
- Measurement and quality
- Metrology and inspection
- Surface engineering and heat treatment
- Coatings and heat treatment
- Materials and stock
- Materials and stock, Plastics and composites
- Fluids, consumables and workshop
- Fluids and lubricants, MRO and workshop
Cite or abstain
Missing stays missing
A value the request does not state stays empty and becomes a question. Nothing is filled in to look complete.
A vague request read without guessing: 2 lines with the quantity marked missing, only what was written cited, and 6 clarifying questions asked instead.
Every value points back
Each extracted attribute carries the span of the customer's words it came from, shown on its chip.
Assumptions are declared
Where the customer's last order or an engineering norm fills a gap, the value is marked as an assumption you can change.
Unsure means review
A match the AI is not sure of is offered for review with the closest options, never slipped into the quote.
Who is asking
The customer comes first
Pin a customer and the AI reads their counts, dates and your own fields before it answers, never a price or another customer's rows.
What the AI reads about a demonstration customer before it answers: orders, lines, typical quantities, most ordered products and terms from the database, beside the organisation's own fields on the record, one of them not yet filled.
Pinned, never guessed
Only the customer chip pins a customer; a name typed in the request never does.