Turn every inquiry into revenue.

Why the first hour decides it

Most RFQs in MRO are not planned purchases. A conveyor motor quit on second shift, or a drive is throwing a fault nobody on the floor has seen. The maintenance manager sends the same request to three or four suppliers and takes the first answer that is complete.

In that situation, speed is the sale. The chief commercial officer at the industrial automation distributor put his target plainly: zero backlog for every inside sales rep, and every response in under an hour. Read how they got there.

Where the hours actually go

The typing is the easy part. The time goes into everything a rep has to work out before a price can go back:

• What the customer means. A partial description, a competitor's part number, a model number read off a label, or a spreadsheet with 40 lines and no headers.
• Which price applies. The manufacturer list, the account's contract tier, a quantity break, or the special someone agreed to on the phone last spring.
• Whether you can ship it. Which branch has it, whether it leaves today, and the real lead time on the lines you don't stock.
• Getting it into the ERP. Building the quote in P21 or SAP, attaching the PDF, and writing the reply.

Then there are the requests nobody owns. The web form queue. The shared sales inbox. The 200-line bid list that lands at 4:47 PM and goes to whoever is still at a desk. We regularly talk to distributors who have stopped asking outside reps to bring in more quote requests, because the inside team cannot answer the ones it already has.

Which RFQs to automate first

Start with the smallest, cleanest queue you have, and prove it before you touch anything harder.

The industrial automation distributor started with requests from its website contact form, about a tenth of inbound volume. In the first month, AI agents answered more than 6,000 emails and returned roughly 1,500 quotes, handing a request to a person only when it needed one. Within two months the company counted more than 10,000 hours a year returned and about six inside sales reps' worth of capacity, without hiring.

A sensible order after that:

1. Requests with a clear part number and standard pricing.
2. Description-only requests and competitor cross-references.
3. Multi-line bids and repeat orders from contract accounts.
4. Engineered and project quotes stay with your most experienced people.

Pricing the quote without the lookup

MRO pricing rarely lives in one place. Lily builds each quote from your price lists, contract terms and quantity breaks in the ERP, then checks margin before it goes out. When the price she would send falls below your floor or conflicts with a contract, she flags it for a person instead of sending it. In the example above, a complete $142,731 quote built from an inbound RFQ was priced and margin-checked in about three minutes.

Most accounts also carry rules that were never written down. InstaQuote covers how Lily learns those from your quote history.

What to measure

• Median time to first response, split by channel. Email, web form and phone behave differently.
• Conversion on quotes created, not just quotes sent. The honest denominator includes the ones that went stale.
• Share of RFQs answered without a person touching them. This tells you more than line-level match accuracy.
• Backlog at close of business, per rep.

One warning: quotes given over the phone and never entered in the ERP are invisible. If a third of your quoting happens that way, your win rate is a guess.

Questions MRO sales leaders ask

Why part identification is the real bottleneck

A large share of MRO requests arrive in words, not numbers. "Same motor as last time, east line." "The 3-phase starter for a 40 horse." The shorthand makes sense to the person on the floor and appears in no catalog.

Your rep is also working across two catalogs. The manufacturer's catalog holds every part number that brand ever made. Your item catalog holds what you stock and what you charge. The job is to connect the customer's words to the first, then the first to the second.

The equipment makes it harder. Plants run motors, drives and switchgear for decades, so the part being asked for has often been replaced twice. And the people who can sort that out quickly are usually one or two senior reps. Everyone newer learns by walking over and asking them.

How Lily resolves a part

1. She reads the request as written: part numbers, competitor numbers, model and serial numbers, and plain descriptions.
2. She checks it against your item catalog, the manufacturer catalogs and any cross-reference data you already keep.
3. She looks at your order and quote history to see what was sold the last time someone asked for the same thing.
4. She returns the most likely match and the alternates, each with the reason.
5. When confidence is low or the part is safety-critical, the request goes to a person with the candidates attached.

Step three is what separates this from catalog search. The parts distributor's agent was trained on millions of parts drawn from past work orders, so it works from what actually broke and what actually fixed it. On a benchmark against GPT-4.1, it scored 97% parts accuracy.

Substitutes without breaking the system

When a part is discontinued or out of stock, the next question is what you can send instead. Lily can offer the equivalents you have approved, with the reason each one qualifies.

Some parts only work as a set. If a kit has seven components and you are short two, mixing brands inside it can create a fit or safety problem. Lily keeps substitutions inside the approved system and flags the gaps for a person.

What part ID changes in the field

Faster lookup helps, but the bigger gains come from changing when it happens. The parts distributor ran a pilot in its service divisions with the worst first-time fix rates. Technicians had to run the lookup before the visit and load the likely part on the van. First-time fix in those divisions rose five to six points, from 61% to 66% in one example. Read the case study.

Across the business, roughly 1,600 technicians had run more than half a million searches by early 2026, returning more than 80,000 hours.

One more habit worth building: track the requests you cannot match. A part you get asked for ten times a month and do not stock is a purchasing decision waiting to be made.

What to measure

• Time to identify, from request received to part confirmed.
• Wrong-part returns as a share of orders.
• First-time fix rate, if you or your customers run field service.
• Unmatched requests per week, and which product lines they fall in.

Questions about part identification

What Lily reads and what she writes

Epicor Prophet 21

Many MRO distributors run P21 on premises, with years of user-defined fields built up on customer and item records. Those fields carry real rules. One says an account always gets a packaging and handling fee. Another says it never gets the default extended warranty. Lily maps them during setup, so her quotes come out the way your best rep would build them.

An on-premises P21 connects through database access or scheduled exports. You do not need to move to the cloud first. Quotes and orders write back as P21 documents, and customers get the same quote PDF they get today.

SAP

Distributors on SAP usually route every write through an integration layer, with a separate team that owns customer and material master data. Lily works through the interfaces your SAP team already governs. She creates quotes and orders through the same approved path, and leaves master data changes with the people who own them.

In practice, write access is often the longest approval in an SAP project. Starting read-only lets the work begin while that approval moves.

Infor, NetSuite, JD Edwards and others

Lily connects to Infor, Oracle NetSuite and Oracle JD Edwards the same way: read first, then write-back scoped to quotes and orders. On the CRM side she works with Salesforce and HubSpot, and with Snowflake where your data team keeps a warehouse. Tell us your version and how it is hosted, and we will tell you plainly what is supported today.

More than one ERP after an acquisition

A lot of MRO platforms are built by acquisition, and the systems show it. One company might run a different ERP in each business it bought, with part numbers that do not match across them. That mismatch hides cross-sell, because nobody can see that two divisions sell the same item to the same customer.

Lily reads across those systems and keeps one view of the customer and the product, while each business keeps its own ERP. Nobody has to finish a consolidation project first.

What about the AI agents your ERP vendor ships?

Most ERP vendors now ship AI agents of their own, and they are good at what the ERP already models. The contract exceptions, substitution rules, branch habits and account notes that decide a real quote usually are not in that data model. Lily sits on top of the ERP, not in front of it, and works alongside the vendor's features rather than replacing them.

Security and where the data lives

Each customer runs in its own single-tenant environment: in our cloud, in yours, or on premises. InstaLILY holds SOC 2 Type II attestation. At a multibillion-dollar healthcare products distributor, the work runs inside the customer's own managed cloud, under its own access controls, cleared through its own IT and security review. The data never leaves. Read the case study or see Security at InstaLILY.

Questions from IT and operations

What bad product data costs you

• Search only works with the exact part number, so reps and customers give up and call.
• Quotes slow down, because a rep has to open the spec sheet to confirm fit before pricing it.
• Cross-sell stays hidden when the same item carries different numbers in different divisions.
• Your web store returns nothing for parts sitting on your shelves.
• Every acquisition adds another catalog with its own naming conventions.

Why manual cleanup never finishes

The usual plan is a project. The product team, or a group of temps, works through the catalog line by line from manufacturer websites and PDFs. The estimates we hear for a large MRO catalog usually run to a year or more of full-time work.

The catalog does not wait. By the time the team finishes pumps, suppliers have changed the valve line, and the company has bought a business with 20,000 SKUs of its own.

How Lily enriches a catalog

1. Agree what matters. Work with your product and sales teams to decide which attributes actually drive selling for each product family.
2. Classify. Place each SKU in a taxonomy designed to absorb the next acquisition.
3. Fill the gaps. Pull missing specs from manufacturer sites, spec sheets and vendor files, and normalize them. The same pressure rating written five different ways becomes one value.
4. Check. Values Lily is unsure of go to a reviewer instead of into the catalog.
5. Write back. Approved values go to your ERP, PIM or catalog, with a record of where each one came from.

What clean data makes possible

Once the attributes are there, search can answer the question a buyer actually asks. "Two-inch PVC ball valve, 150 psi, socket ends" returns the right items, filtered, with what is in stock. Part identification and cross-reference get more accurate too, because there is more to match on. See how part identification works.

Clean data also shows which parts sell together. At the industrial process solutions distributor, the enriched catalog now feeds AI sales workflows for about 300 reps. The company recovered $491K in revenue within three months.

What to measure

• Attribute completeness by product family: the share of SKUs with every required field filled.
• Searches that end with no result, on your web store and in your reps' tools.
• Time from new SKU received to sellable.
• Share of values that needed a reviewer, which should fall as the rules settle.

Questions about product data

Why good reps still miss accounts

A rep carrying a couple hundred accounts cannot check every one each week. Attention goes to the loudest customers and the ones they like, and the quiet decline at a steady account goes unnoticed until the orders stop.

The signals are there, just scattered. Order history is in the ERP, contacts are in the CRM, margin is in a report, and the context is in someone's inbox. And field reps are rarely at a desk to go looking. As one field rep at the fastener distributor put it: "Ninety-nine percent of the time it's not over email."

Cross-sell is the hardest to see in MRO, because the catalog is so wide. A plant that buys bearings from you every month and has never bought a belt or a seal from you is buying them from someone else.

What goes on the list

• An account buying below its usual pattern, before it turns into a lost customer.
• A regular repurchase that is overdue.
• An open quote that is going stale.
• A product family the account buys elsewhere, based on what similar accounts buy from you.
• A customer who has not bought in a year.

Each item carries the data behind it and a drafted next step: talking points for a call, a text, or an email in the rep's own voice. The rep decides what to send.

Following up on the quotes that can still close

Not every open quote deserves a call. Some are budget numbers for a project that will not buy this year. Some went to an engineer who does not sign purchase orders. Lily separates the quotes worth following up, puts them on the rep's list, and records what happened: won, lost with a reason, or not ready with a new date.

That record is what makes quote conversion measurable. See what to measure on quote response.

Notes that log themselves

Field reps write up their visits in the parking lot, at the hotel at night, or not at all. Lily captures notes by voice or text right after the visit and turns them into structured CRM records and follow-up tasks.

At a national trailer manufacturer, reps were barely using the CRM, because logging a day of calls properly took about three hours in the evening after driving all day. Eight weeks after the company replaced it with one that logs itself, weekly logged activity had grown roughly sevenfold. Read the case study.

Getting senior reps to use it

At the fastener distributor, the reps who loved the platform were the new ones. The most senior reps would not touch it. The chief financial officer named the lever himself: "Outside salespeople are primarily driven by their commissions."

The company restructured commissions to reward what the platform surfaced: new customers, and gaps in existing ones. A new vice president of sales made it a standing topic in one-on-ones. In the middle of June, usage doubled. Over the following months, rep coverage across the active group rose by mid-teens percent, and company-wide growth for the year landed near 12%. Read how adoption turned into revenue.

Measuring the lift honestly

Comparing this year to last year flatters any pilot when the market is also rising. A territory-based manufacturer compared pilot reps against non-pilot reps selling the same products in the same period, and subtracted what the market was doing anyway. What survived was a lift of roughly five percent in sales per account. See how they measured it.

Questions from MRO sales leaders

Where order entry time goes

Keying the order is the short part. The time goes into making the PO fit your system:

• The PO lists the customer's part numbers, not yours.
• They ordered 10, meaning boxes. You sell the item each.
• The company name on the PO matches three ship-to accounts.
• The price doesn't match the quote, or the quote expired last week.
• There is no quote at all, so someone has to look up the contract price.

In a lot of distributors the same order is then keyed twice, once into the CRM and again into the ERP. And the customer calls to ask whether you got it.

How Lily processes a PO

1. She reads the PO from an email, a PDF attachment, a fax or a web form. Orders already on EDI stay on EDI.
2. She identifies the customer and ship-to, even when the name on the PO matches several accounts.
3. She matches each line to your item and unit of measure, using the customer's past orders and cross-references.
4. She checks price and terms against the quote or contract, and checks for a duplicate of a recent order.
5. She creates the sales order in the ERP and sends the confirmation.
6. Anything that doesn't match goes to a person, with the PO, the quote and the difference side by side.

The exceptions that decide whether it works

A clean PO is easy for any tool. What matters is what happens to the rest.

You set the tolerances: how far a price can drift, how long a quote stays valid after expiry, and which accounts always get a human check.

The second inbox: order status

After the PO come the follow-ups. Did you get my order? When does it ship? Can I get tracking? Those calls and emails pull the same reps away from quoting.

Lily answers status requests from the ERP and, where it is connected, carrier tracking. Returns, disputes and anything unusual go to the right team with the history attached. Routing matters more than it sounds. At a multibillion-dollar third-party logistics provider, emails that had to be passed between teams added about nine business hours to the first response. See what changed there.

What to measure

• Touchless rate: the share of POs that become orders without anyone editing them.
• Time from PO received to order confirmed.
• Order-entry errors found after the fact, in credits, returns and reshipments.
• Duplicate orders per month.
• Status calls and emails per hundred orders.

Questions about order entry

Eight questions to ask every vendor

Bring your hardest request to the demo

Pick the RFQ your team dreaded most last week: the 60-line bid with competitor part numbers, three lines you don't stock, and an account with a contract exception. Send the same one to every vendor. How each one handles the lines it can't resolve will tell you more than anything on its website.

Why first AI pilots stall in distribution

We hear the same four stories from distributors whose first pilot went nowhere:

1. It couldn't read the input. The demo used clean PDFs. Real requests arrived as email threads, scans and spreadsheets with no headers.
2. It read the request, and a person retyped the order anyway. The output never reached the ERP, so nobody saved any time.
3. Every exception landed on a person. The easy requests were automated, the hard ones still took all day, and the net saving was close to zero.
4. Nobody owned it after the vendor left. Prices changed, a product line was added, and accuracy slid until reps stopped trusting it.

If one of those sounds familiar, it is worth asking each vendor directly how they prevent it.

How to structure a pilot

• Start with one queue. The national distributor of industrial automation components in our quoting case study started with its website form, about a tenth of inbound volume, before working up to harder requests.
• Write down the baseline first. Response time, touchless rate, quotes per rep, and conversion on quotes created, measured before anything goes live.
• Compare against a control. A territory-based manufacturer compared pilot reps with non-pilot reps in the same market and period, and published the roughly five percent lift that survived. See how.
• Set a window and an exit. A defined pilot period, a success gate both sides signed, and a clear decision at the end.

When a narrower tool is the better choice

If all you need is emailed POs turned into ERP orders, a dedicated order-capture tool may be enough. Lily is built for distributors that want quoting, order entry, part identification and sales follow-up to run on one shared understanding of their customers, products and pricing. If that isn't the problem you have, say so early and save everyone the evaluation.

Questions about evaluating AI for MRO