Answers whether you can make a described part — process, material, size envelope and tolerance — checked against your actual equipment rather than a capabilities page.
An AI agent for capability questions handles the first thing an engineer or buyer wants to know: can you make this. It asks about the process, the material, the size envelope, the tolerances and any finishing required, checks that combination against your equipment and process list, answers plainly where you can and cannot, and routes borderline parts to an engineer. Manufacturing websites list capabilities as prose and equipment as a table, and neither answers the actual question — which is whether a specific part, in a specific alloy, at a specific tolerance, fits inside what the shop can hold and hit.
Checks a described part against real equipment.
Asks the process, material, size envelope and tolerances
Checks that combination against your machine and process list
Answers plainly where a part is within or outside capability
Identifies which constraint fails when a part does not fit
Routes borderline parts to an engineer rather than guessing
Records what buyers are asking for that you cannot make
An engineer sourcing a part is comparing several shops in an afternoon, and the one that answers whether the part fits gets the drawing. The information exists — the machine list, the envelopes, the tolerance the shop routinely holds — but it is published as a capability table that requires the reader to do the mapping, and most will not. The second cost lands on the quoting team. Requests for quotation arrive for parts that were never manufacturable in that shop, and somebody spends an hour establishing that before declining. Answering the fit question at the point of asking removes both, and it is a question with a genuine answer rather than a matter of positioning.
A three-step flow from a described part to a straight answer.
The agent establishes process, material, size envelope, tolerances and finishing — the five facts that determine manufacturability.
It tests that combination against your machine list and process capability rather than against a marketing description.
Clear fits and clear misses are answered directly, with the failing constraint named. Borderline parts go to an engineer.
An inquiry for a part just outside the envelope.
Scenario: a machining shop was receiving requests for quotation for parts it could not hold and spending quoting time to decline them. An engineer asks whether the shop can machine a part in a specific stainless alloy. The agent asks the envelope and the tightest tolerance on the drawing. The alloy is routine and the tolerance is within what the shop holds, but the longest dimension exceeds the travel on every machine that could do the work. Rather than a vague no, the agent says the tolerance and material are fine and names the specific constraint that fails, which lets the engineer decide whether the part can be split or whether to go elsewhere. That answer took a minute and cost no quoting time. Over a quarter the shop can see how often the same envelope constraint is what turns work away, which is a concrete input to a machine purchase decision.
Anybody quoting parts against fixed equipment.
Capability you cannot demonstrate is capability you do not have.
Quoting time is spent declining work that never fitted.
You are pulled in to assess parts a screen could have filtered.
Envelope and tolerance decide most of what you can take.
Buyers compare several suppliers in a single afternoon.
Turned-away demand is the case for a machine purchase.
Reads the equipment list, answers the fit, records the gap.
Holds the machine list, envelopes, tolerances and process capability.
Records the inquiry with the part parameters attached.
Receives drawings and carries the written answer.
Stores drawings and specifications against the inquiry.
Routes borderline parts to a manufacturing engineer.
Reports which constraints most often turn work away.
The fit questions asked before any drawing is sent.
Questions about answering manufacturability quickly.
An AI agent for capability questions answers whether you can make a described part: it asks the process, material, size envelope, tolerances and finishing, checks that against your equipment and process list, answers plainly, and routes borderline parts to an engineer.
It can accept one and attach it to the inquiry, but the fit answer should come from the parameters the buyer states. Interpreting a drawing is engineering work, and a wrong reading that produces a confident yes is worse than asking three questions.
Because a bare no ends the conversation and a specific one sometimes does not. An engineer told that the tolerance is fine but the length is not can split the part or change the design, and that is a quote you would otherwise have lost.
Anything at the edge of an envelope, at the tightest tolerance you hold, or in a material you run rarely. Those should reach an engineer, because the cost of a wrong yes is a job you cannot deliver.
Your machine list, and it needs envelopes and realistic tolerances rather than manufacturer specifications. The tolerance a shop routinely holds in production is not the number on the machine's datasheet, and using the latter will produce promises the floor cannot keep.
It is one of the better inputs to a capital decision. Most shops decide on machine purchases from a general sense of what they are missing, and a count of quotes lost to a specific constraint makes the case concrete.
No, it filters what reaches quoting. Price, lead time and process planning still require people; what changes is that they stop spending time on parts that were never manufacturable in your shop.
Answers whether you can make a described part — process, material, size envelope and tolerance — checked against your actual equipment rather than a capabilities page.