Considering Paperless Parts? Five alternatives, and what its own compliance pages actually claim.

Paperless Parts sells the front of a custom manufacturing business. Not the shop floor, not the job costing, not the machines: the part of the operation where a request for quote arrives as a drawing and a deadline, and somebody has to turn it into a number that wins the work without losing money on it. Its own product pages describe costing templates, geometrically driven routers, customizable pricing templates, live pricing and availability for metals and purchased components, automatic design for manufacturability warnings, and sheet metal nesting and unfolding.

The AI inside it is called Wingman, and the company describes it as reading and interpreting technical documents the way an experienced estimator would. The specific claim underneath that is more useful than the slogan: Wingman is said to recognize over ten thousand ASTM, AMS, MIL SPEC, NADCAP, geometric dimensioning and tolerancing, and original equipment manufacturer specifications. No methodology sits behind that count, so treat it as a vendor claim rather than a measurement, but the shape of the claim tells you what the feature is for. It is a callout reader. If your estimators lose an hour per quote decoding what a customer wrote in the notes block of a drawing, that is the hour the product is aimed at.

There is a second piece that gets less attention and probably deserves more. The Smart RFQ Form embeds on your own website, so inbound requests arrive as structured records instead of as attachments in a shared mailbox. Shops that have never had a front door for quoting often get more from that than from the model, because it removes a wait rather than compressing a task. That distinction runs through the rest of this page, and it is the same one we draw in the manufacturing automation consulting guide.

Company facts worth having before a call: the head office is at 101 Federal Street in Boston, with a second office in Encinitas, California. The founder's earlier company, a separate business, started in 2001; Paperless Parts does not publish its own founding year on its company page. We looked, and we would rather say we could not find it than round it off.

Paperless Parts publishes no price. Its pricing page loads normally, and what it contains is a form: name, company, employee count, email, phone. There is no dollar figure, no tier name and no per user rate anywhere on it, and no plans page exists elsewhere on the site. The only pricing signal published is a description of who the tiers are for, running from a five person shop to a five hundred person enterprise. That tells you the segmentation exists. It does not tell you what happens to your bill when you add a third estimator.

What makes this worth a section rather than a sentence is that it is the category norm, and buyers rarely get to see the whole category at once. We checked Paperless Parts and eight competitors on the same day, on each vendor's own site, in a browser.

Read on August 11, 2026, on each vendor's own site. Third party research sites publish estimated figures for several of the vendors in the no column, and some of those estimates circulate widely enough to look official. They do not come from the vendors, so they are not repeated here. A number you cannot trace to the company that will invoice you is not a price, it is a rumor with a decimal point.

The practical consequence for a shop owner is that comparing this category on cost requires entering a separate sales process for every vendor on the shortlist, and each number that comes back is anchored to whatever the salesperson learned about you on the discovery call. That is a real cost of evaluation, and it is one of the reasons a fixed fee commission quoted before work starts reads differently to operators who have been through it. Ours are published at what AI consulting costs for a specialty manufacturer and on the pricing page.

If you quote aerospace or defense parts, the compliance section of a vendor's site is not boilerplate, it is the reason the evaluation either continues or stops. So it is worth reading the exact words, and on this site the exact words change depending on which page you land on.

The security page says the platform is designed for CMMC compliance. The dedicated compliance page raises the question in a heading and answers it by saying the company is FedRAMP Moderate Equivalent and is designed to support manufacturers looking to achieve CMMC Level 2. The AI principles page describes the company, without qualification, as a CMMC compliant cloud service provider. A fourth page lists it as an ITAR and CMMC compliant solution in a bullet.

Those are not the same statement. Designed to support a customer pursuing Level 2 is a claim about the platform's suitability. CMMC compliant is a claim about the vendor. Underneath all four, the one dated, third party anchored claim is a FedRAMP Moderate Equivalency audit the company says it completed in 2023, assessed by a FedRAMP recognized third party assessment organization. The platform is separately described as ITAR registered, hosted on AWS GovCloud, with FIPS validated encryption in transit and at rest.

None of this makes the product unsuitable. FedRAMP Moderate Equivalency plus GovCloud plus ITAR registration is a stronger posture than most software a mid sized shop runs, and it is more than several of the competitors in the table above publish at all. The point is narrower and it is about procurement rather than engineering: when your prime contractor's supplier questionnaire asks whether your quoting system is CMMC certified, you need to know which of your vendor's four sentences is the one that appears on the order form, and you need it in writing before you sign rather than after. Ask for it by name.

Two things we looked for and did not find on the security, compliance or AI principles pages: a SOC 2 report and an ISO 27001 certificate. Neither is mentioned. That is not evidence that no report exists, because many vendors release SOC 2 only under a mutual non disclosure agreement. It does mean your own security questionnaire cannot be answered from the public site, and it is exactly the sort of item worth putting on the list in the security questions to ask before an AI build.

The pricing gap is normal for this category. The document gap is not, and it is the finding on this page we would want a buyer to carry into the room.

The only agreement linked from the site footer is a privacy policy carrying an effective date of August 22, 2018. Read it and it does not describe a business software relationship at all; it refers to ordering products and to shipping information, in the register of a consumer ecommerce policy. There is no terms of service linked, no data processing agreement, and no trust center. The terms of use path returns a 404.

So every commitment that would settle an argument two years from now arrives with the order form rather than before it. Contract term and renewal mechanics. Uptime, if any is promised. Limitation of liability. The list of subprocessors that will touch your customers' drawings. And the one that matters most in this category, what happens to your quote history, part library and customer records after cancellation.

That last item is worth holding next to something the vendor does say clearly. Its compliance page states that customers always retain ownership of the data they upload. We take that at face value; it is a plainer ownership statement than several vendors in this comparison series make. But ownership and access are different properties. Nothing published describes an export format, an offboarding process, or how long you have to retrieve anything. A shop that has run five years of quoting through a system holds its own pricing history inside that system, and the day it wants to leave is the day it discovers whether ownership was a title or a mechanism. The general form of that argument is at why code handoff matters, and the questions to put in a scope document are in how to write an RFP for a custom AI build.

Three sentences to put in an email before you sign: what is the initial term and the renewal notice period, what is the documented process and format for exporting our full quote and part history if we do not renew, and which of your published compliance statements is the one warranted in this agreement. If those come back cleanly, the rest of this page is a comparison rather than a warning.

This series exists to help an operator choose, and a page that only lists gaps is a sales document wearing a lab coat. So here is the other side, and on one dimension it is genuinely ahead of the field.

Its AI principles page, updated on May 1, 2026, says three things most vendors leave vague. The foundation of its custom models is synthetic data. A random sample of technical drawings uploaded by customers has been used to train proprietary models, which it states rather than buries. And customer data is not used to train models for generative applications. It then offers an opt out in plain language: a customer who would rather their data stayed out of aggregated model improvement can ask, and the request will be honored.

Compare that with the pattern this series keeps finding, where a vendor's terms grant a broad training license by default and the opt out costs you features. Paperless Parts has written the sentence a defense subcontractor's counsel needs to see, and it has dated the page. That is a real credit and it should count in the evaluation.

The integrations page deserves similar credit. Ten enterprise resource planning systems get their own description plus a deeper page or case study, including JobBOSS and JobBOSS squared, M1, Epicor 10, 11 and Kinetic, Fulcrum, Genius ERP, Global Shop Solutions, MIE Trak Pro, ProShop and StartProto. Three more are named with a single line and nothing behind it, and one links only to a press release. NetSuite does not appear. A page that lets a reader tell those tiers apart is more useful than a wall of logos, and most vendors in this category ship the wall of logos.

Worth noticing on the way past: several of the systems on that integrations list, ProShop and Global Shop Solutions and MIE Trak Pro and the ECI products, also sell quoting. They are partners and competitors at once. That is not a criticism of anybody, it is a fact that changes how you should read a reference call arranged by either side.

How to tell which one you are.

The decision is usually settled by one question, and it is not a question about AI. Is the thing that wins you work expressible as a template?

For a large share of shops the honest answer is yes. Quoting is slow because requests arrive unstructured, because material prices are looked up by hand, because there is no shared cost model and two estimators price the same part differently. Every one of those is a solved problem in a product, and a product will beat a build on cost, on time to first result and on the number of things that can go wrong. Buy, implement properly, and spend the saved money on the shop floor. The vendor's own published onboarding window for its implementation is eight to twelve weeks, which is a reasonable planning assumption for the category rather than a promise.

For a smaller share the answer is no, and it is usually obvious once stated aloud. The pricing logic that wins work is a part library built over three decades, a margin policy that varies by customer and by lead time, a capability rule your foreman can state but nobody has written down. A template approximates that. Approximation is exactly the wrong instrument for the number you send a customer, because the error shows up as either lost work or lost margin and you cannot tell which from the outside.

There is a third case that deserves naming because it is the one operators talk themselves out of. Some shops have both: ordinary intake problems and one genuinely proprietary pricing rule. The sequence for them is buy the product for intake, run it for two quarters, and only then decide whether the remaining gap is worth a build. The order of operations is the decision, not the vendor. The way to structure that so it ends in a decision somebody signs is set out in how to run an AI pilot that produces a decision, and the readiness question is scored at the maturity index.

What we would build, and what we have not built.

Our commissions run as fixed fee engagements between 45,000 and 180,000 dollars, on a six to seven week build cycle, with the operator owning the code at handoff. In a shop the shape is consistent: a layer that reads the inbound request, matches it against prior jobs in the system you already run, checks capability and current capacity, drafts a price on your own rules, and hands a senior estimator something to approve rather than something to assemble. The system of record stays where it is. We read and suggest; we do not write to the plant floor. The reasoning behind that boundary is in the production planning guide and the architecture is in the Epicor Kinetic playbook.

We have not shipped a commission for a specialty manufacturer yet. That is stated on the specialty manufacturing page and it is stated here, because a comparison page that quietly implies a track record it does not have would fail the same test we are applying to everybody else on this page. The adjacent production work and what it actually proves is listed on the manufacturers page. If you want the eleven firm landscape rather than our view of it, that is the best AI consultants for specialty manufacturers, and the way to vet any of us is the manufacturer's guide to choosing a consultant.

What is Paperless Parts, and what is Wingman?

Paperless Parts is quoting and estimating software for custom manufacturers, sold to job shops, machine shops, sheet metal fabricators and contract manufacturers. Wingman is the AI layer inside it rather than a separate product, and the company describes it as reading and interpreting technical documents the way an experienced estimator would. The published feature set around it includes costing templates and geometrically driven routers, customizable pricing templates, real time pricing and availability for metals and purchased components, automatic design for manufacturability warnings, and sheet metal nesting and unfolding. A customer facing intake form called the Smart RFQ Form embeds on your own website so inbound requests arrive structured instead of as email attachments. All of that description comes from the vendor's own product pages, read on August 11, 2026.

How much does Paperless Parts cost?

Paperless Parts publishes no price. Its pricing page loads normally and contains a lead capture form asking for name, company, employee count, email and phone, and no dollar figure, tier name or per user rate appears anywhere on it. No plans or packages page exists elsewhere on the site. The company will say that its tiers are meant to suit anything from a five person shop to a five hundred person enterprise, which tells you the shape of the segmentation without telling you the number. Whether the license is per seat, per site or usage based is not published either, so a buyer cannot model the cost of adding estimators without entering a sales conversation.

Which competitors actually publish a price?

Two of the eight competitors we checked on August 11, 2026. DigiFabster lists a Bare Essential tier starting at 415 dollars a month with a 4.99 percent processing fee on annual billing, and a Growth tier starting at 840 dollars a month with a lower processing fee. Katana lists a free plan and a Core plan starting at 299 dollars a month, with itemized add ons and an optional onboarding fee. Paperless Parts itself, Fulcrum, ProShop, JobBOSS squared from ECI, Epicor Kinetic, Global Shop Solutions and MIE Trak Pro all gate pricing behind a form. Third party aggregator sites publish estimates for several of these, and none of those numbers come from the vendors, so we do not repeat them here.

Does Paperless Parts train its AI on my drawings?

Not for generative models, according to its own AI principles page, which was updated on May 1, 2026. That page says the foundation of its custom models is synthetic data, acknowledges that a random sample of uploaded technical drawings has been used to train proprietary models, and states that customer data is not used to train models for generative applications. It also offers an opt out: a customer who prefers their data stay out of aggregated model improvement can ask, and the company says it will honor the request. That is a clearer position than most vendors covered in this comparison series take, and it is worth crediting rather than glossing over.

Who owns the data I upload to Paperless Parts?

You do, in the vendor's own words. Its compliance page states that customers always retain ownership of the data they upload. What is missing is the mechanism. Nothing on the security page, the compliance page, the AI principles page or the privacy policy describes how you get your quote history, part library and customer records back out in a usable format if you leave. Ownership without a documented export path is a claim about title rather than about access, and the difference only becomes visible on the day you try to switch.

Is Paperless Parts CMMC compliant?

Its own pages give three different answers, and the difference matters if you quote defense work. The security page says the platform is designed for CMMC compliance. The dedicated compliance page asks the question in a heading and answers that the company is FedRAMP Moderate Equivalent and is designed to support manufacturers looking to achieve CMMC Level 2. The AI principles page describes the company flatly as a CMMC compliant cloud service provider. The one dated third party claim underneath all of that is a FedRAMP Moderate Equivalency audit completed in 2023, and the platform is described as ITAR registered and hosted on AWS GovCloud with FIPS validated encryption. Ask the vendor which of the three sentences is the one on the order form.

Does Paperless Parts publish a SOC 2 report or ISO 27001 certificate?

Neither appears on the security page, the compliance page or the AI principles page as of August 11, 2026. That is not the same as saying no report exists, because plenty of vendors hand a SOC 2 report over under a mutual non disclosure agreement rather than advertising it. It does mean the buyer has to ask, and it means an information security questionnaire from your own largest customer cannot be answered from the public site alone.

There is no terms of service on the site. Does that matter?

It matters more than the pricing gap does. The only agreement linked in the footer is a privacy policy carrying an effective date of August 22, 2018, and it reads as a general consumer policy with references to ordering products and shipping information rather than as a business software agreement. There is no terms of service, no data processing agreement and no trust center linked from the site, and the terms of use path returns a 404. Every commitment that would decide a dispute later, meaning term length, renewal, uptime, liability, subprocessors and what happens to your data after cancellation, arrives in the order form rather than in advance. Read it with the same care you would give a supply agreement.

Should a shop with two senior estimators buy this or commission something?

Buy first, in most cases. If your quoting is slow because requests arrive as unstructured email and nobody has a shared cost model, a product that already does intake, geometry reading and templated pricing will beat a custom build on both cost and time to first result. The case for commissioning appears when the pricing logic that wins you work is specific enough that no template holds it: a part library built over three decades, a customer by customer margin policy, capacity rules only your foreman can state. At that point you are paying a vendor to approximate something you already own, and the honest comparison is against a fixed fee build you keep.

What does an integration with our ERP actually get us?

It depends which ERP, and the integrations page is unusually honest about the difference. Ten systems get their own description plus a deeper page or case study, including JobBOSS and JobBOSS squared, M1, Epicor 10, 11 and Kinetic, Fulcrum, Genius ERP, Global Shop Solutions, MIE Trak Pro, ProShop and StartProto. Three more are named with a single line and nothing behind it, and Infor VISUAL links only to a press release. NetSuite is not listed at all. Before signing, ask which of those categories your system falls into, ask what the connector writes as opposed to reads, and ask to see it running against a shop the size of yours.

What would you build instead, and what would it cost?

We commission custom AI as a fixed fee engagement between 45,000 and 180,000 dollars, on a six to seven week build cycle, with the code owned by the operator at handoff. For a shop the pattern is usually a quoting layer that reads the inbound request, pulls matching prior jobs out of the system you already run, checks capability and capacity, drafts a price on your own rules, and hands a senior estimator something to approve rather than something to build. We have not shipped a commission for a specialty manufacturer yet, and we say so on the vertical page rather than implying otherwise here.

How should we run the evaluation so it produces a decision?

Measure your own baseline before a demo is booked. Pull the last ninety days of requests out of your existing system and record three numbers: elapsed hours from request received to quote sent, quotes issued per estimator per week, and win rate on quotes returned inside the customer's decision window. Write down what each would have to reach for the purchase to be worth renewing, and write it down before anyone sees a demo. Then run one vendor against one real week of quoting rather than a curated sample. A pilot designed this way ends in a decision somebody signs, which is the only output that matters.

Related reading.

More on AI in custom manufacturing: AI consulting for manufacturing, what we would build for a specialty manufacturer, and who actually automates a manual back office workflow in manufacturing. The integration paths are documented in the Epicor Kinetic playbook and the Sage Intacct playbook. To size your own numbers before a call, the manufacturing calculator takes a few minutes, and operators willing to contribute data can see the 2027 manufacturing AI benchmark. If the vocabulary is new, start at what custom AI means and what AI commissioning means, then what a mid market engagement costs and how we work. The generic version of the buy against build question is generic SaaS AI versus a custom commission, and the rest of this series sits on the comparisons hub.