30-min call with RobBook a proof-of-concept call
Job 01 · Applications desk

// Industrial filter manufacturers · Spec sheet quotes

Filter quotes drafted from the customer's own specification, for your applications person to check.

A plant emails a filter spec, a contractor forwards a section, or a maintenance tech reads the numbers off a worn unit. A free AI proof of concept has AI draft the quote sheet from copies of 20 to 30 requests your desk already answered. Your reviewer, the applications person who knows these specs best, holds each draft against the quote that went out. The match stays with your people.

Book a proof-of-concept call30 minutes with Rob
Past requests
20 to 30
Filters selected by AI
None
Live quotes touched
None

// The one job

From a customer's stated specification to a quote sheet a person can check

A filter request arrives as a spec page, a bid section, a schedule of air handling units or the data off a used cartridge, and it rarely arrives complete. Someone at your applications desk reads it, works out what the customer actually said, chases what is missing and then decides which of your products answers it. The reading comes before any selection.

In the test, AI reads copies of those same requests and drafts a quote sheet: dimensions as stated, media and efficiency rating as stated, flow and pressure drop fields, end caps and gaskets, temperature and compatibility fields as stated, and a short list of what a person must confirm before matching a part. Your own selection on each closed quote goes in as an input. Nothing goes to a customer, and your reviewer sets every draft beside the quote your desk sent.

Customer spec or labelAs received
Customer
Application
Nominal size
Stated rating
FieldAs statedUnit
Illustration, not a real specification
Read as
written
Quote sheetDraft
Customer
Request no.
Method cited
To confirm
FieldStatedTo confirm
Made-up sheet, for illustration

// What makes it hard

Where a filter spec stops meaning what it seems to say

Nominal is a name, not a measurement. North Carolina's air filter specification lets a filter's actual face run up to five-eighths of an inch under nominal, and it sets a minimum depth for each nominal depth: 7/8 inch for a 1 inch filter, 1-3/4 for a 2, 3-3/4 for a 4, 5-7/8 for a 6 and 11-1/2 for a 12. The University of Michigan's master specification calls a V-cell 24 by 24 by 12 inches, nominal and without gaskets. UMBC's 2024 filter schedule lists 22-3/4x24x2 and 9-7/8x42-1/2x1 beside the ordinary sizes, along with rows that read TBD by 2. A tape measure on a worn unit gives a third answer. The draft holds nominal, actual and measured in separate columns and leaves TBD as an open question.

The rating line can be written in an older language. UMBC's pleated filter section asks for a dust spot efficiency as determined by ASHRAE 52.2 test methods, then requires a MERV in the next sentence, while the National Air Filtration Association's January 2026 guide says dust spot testing was removed from the standards in 2009. The University of Texas at Arlington's 2020 filter guideline asks for MERV 8 and rates filters to ASHRAE Test Standard 52-76. UMBC's same section also states an average resistance in the units arrestance normally takes, which reads like a typing slip. North Carolina lists MERV 13 and MERV 13-A as separate ratings, since MERV-A comes from the optional Appendix J conditioning step. The draft copies every rating, method and edition as stated and marks any line that doesn't hang together. It translates nothing.

A rating without a velocity is half a line. The same NAFA guide says the MERV is used along with the air velocity at which the test ran, and one independent lab's service sheet notes that ASHRAE 52.2 recognizes only seven face velocities for MERV evaluations. Specs tie their pressure fields to velocity as well, and they name them differently: New York State's master specification gives initial resistance at 500 fpm and a recommended final resistance, North Carolina has a minimum rated final resistance, and UMBC's ridged mini-pleat lists rated initial and final resistance at 2,000 cfm with a minimum operating velocity of 750 fpm. The draft carries flow, face velocity, initial and final resistance as separate fields, each with its unit and the customer's own label, and writes not stated wherever the sheet is silent.

End caps, gaskets and seals decide whether it fits at all. New York's HEPA paragraphs offer a face gasket of neoprene expanded rubber, ceramic fiber or silicone and five different media-to-frame bonds. Michigan puts a neoprene gasket of set width on the downstream face of a V-cell header, or an endless poured polyurethane seal instead. UMBC's mini-pleat has to run up to 176 degrees Fahrenheit with no metal components at all. EPA's fact sheet on pulse-jet cartridge collectors says temperature is limited by the filter media and the sealant together, so a temperature on a sheet is not only a media question. The draft records end cap style, gasket material and position, sealant, temperature and any fluid or chemical named, exactly as stated, and takes no position on compatibility.

Some fields only a person can close. North Carolina wants each filter marked with the maker's name or trade name, series, nominal size, MERV, MERV-A if tested and the UL flammability mark, but ring, link and bag filters without frames carry those markings on the packaging, so a used bag with its carton gone has nothing to read. The same state's ordering data asks for the number of pockets and the header size. Every draft therefore ends with a list for your applications person: actual size confirmed, rating and test velocity, media and construction, gasket side, temperature and fluid, pockets or header, and anything the customer left blank. Your reviewer will check that no stated value changed on the way across, no blank got filled and the list caught what your desk really had to ask.

// What you send

Requests your applications desk already answered

Copies only
After the NDA
01

20 to 30 closed requests

Customer specifications, bid sections, schedules and label readings, with the quote each one became. Include one your desk got wrong at first, one with a blank somebody had to chase and one you declined.

02

The selection your person made

Which of your products answered each request, as recorded on the quote. It goes in as given, so the test never has to choose.

03

Your product data from those dates

Data sheets for the products quoted, in the versions that applied when each quote went out, so a draft can show what your literature said then.

04

Your desk's own checklist

Any list your people use for size, seal, temperature or application questions. It all moves only after the NDA, and if specifications can't leave the building we build look-alikes with you.

// Before it starts

Agreed before a single spec is copied

The owner settles each of these before any request leaves your files.

Terms for the quote testThe owner's decisions
Your reviewer
The owner's pick: usually a senior applications person
Success criteria
Written down: which fields must match and how open questions score
Sample
Chosen by your reviewer across air, liquid and dust collection
Who decides
Your team: every match, every price, every quote
Price
A live-job range from Rob, before any request leaves your files
NDA
Signed first

// How the test is judged

The quote your desk sent beside each draft

Request by request
On your own closed work
Across the same 20 to 30 requestsWhat your desk quotedDraft from the test
Nominal, actual and measured dimensions, kept apartScored on your requestsScored on your requests
Media and rating as the customer stated themScored on your requestsScored on your requests
Test method and edition as citedScored on your requestsScored on your requests
Flow, face velocity and pressure drop fields, with unitsScored on your requestsScored on your requests
End caps, gaskets and seal positionScored on your requestsScored on your requests
Temperature and compatibility lines, unchangedScored on your requestsScored on your requests
Items listed for a person to confirm, against what your desk had to askScored on your requestsScored on your requests
Nothing filled in yet. Cells stay open until your own closed requests have run.Your applications reviewer gives every score.

// After a yes

About 30 days on new requests

Once your reviewer says yes, new specifications get the same drafting for about 30 days.

Your applications person approves every quote before it reaches a customer, and the match stays with that person throughout. Scoring carries on against live requests. Any further job is a separate decision at its own price, with monthly support after that.

// Jobs that could follow

After quotes, if the first job earns it

Possibilities, ranked
Priced one at a time
  1. 01

    Filter test report packages

    The documentation a customer asks for on a lot, drawn from lab and plant records for your quality manager to sign.

    Next
  2. 02

    Filter schedules from bids

    A campus or plant list of units, sizes and change frequency collected into one line list for whoever prices the contract.

    Then
  3. 03

    Label and carton data checks

    What each customer wants printed on a frame or carton, compared with your label file before a run goes out.

    After that

// One hard limit

It never selects a filter, and never touches a live quote.

The draft records what the customer stated and the selection your person already made. It doesn't choose a filter, match a part, judge a test result, sign a certificate, say a product meets a rating or rule on a temperature or chemical line. It never opens your ERP, a live quote or a live certificate, and an applications person approves every quote before it goes anywhere. Defense and aerospace requests stay out of the test. This page describes public documents, and none of it is legal or engineering advice.

Priced per job. You get a range before the proof of concept and a fixed price before you sign.

Rob gives the range on the call, once he knows how many specifications your desk answers in a month and how many product families they cover.

// Questions

Spec sheet quoting questions

Will the AI decide which of our filters to quote?
No. Your person's choice on each closed quote is handed to the test as a given. The draft reads the customer's lines and lists what is unsettled, and choosing the product stays at your desk.
Can it work from a worn filter's label?
Yes, as far as the label goes. The draft records the size, rating and any other data the label states, and everything it doesn't state becomes a question for your applications person. Matching a part to that reading is a person's job.
Does it convert a MERV to an ISO 16890 class, or the other way round?
No. The two come from different test procedures, and ASHRAE itself calls its table between them approximate. The draft keeps whichever rating the customer named and leaves any comparison to your staff.
What happens when a spec leaves a bracket or a TBD?
It stays open. The draft never fills a blank from a typical value, and each one shows up as a question for your applications person. Your reviewer scores that too.
Can it read liquid cartridge and dust collector specs as well as panel filters?
Yes, if they are in your sample. Micron ratings, end cap styles, bag or cartridge dimensions and media are read as written and placed in their own fields. Whether any of it suits an application is your applications person's decision.
Do you need access to our ERP or product data?
No access at all. Copies and exports are the whole of what the test uses, and drafts come back as files. What a live month would do about your systems gets settled only after a yes.
Rob Eason

Rob Eason, founder

// Book a proof of concept call

Bring the spec your desk had to call back about.

Half an hour with Rob. Tell him what you make, your role and roughly how many specifications your desk answers monthly. Every spec stays with you until the NDA is signed.

  • Co-owned a 60-employee financial services firm and sold his share to his partner.
  • Ran GOSO Technologies, a Boston solar company rated 5/5 across 24 Google reviews.
  • 5/5 across 83 reviews in HighLevel's Certified Admin Directory.
  • 4.3/5 from 6 Google reviews of Massmarket AI.
  • Every directory and Google review, critical ones included, is shown word for word at massmarket.ai/reviews.
  • Takes every proof-of-concept call himself.