AI illustrationIndustrial manufacturing · LinkedIn
One in three engineers replied to a foundry president’s AI agent
Lamothermic, a US investment-casting foundry, runs Sellinger on its president’s own LinkedIn. The agent asks engineers one real engineering question, follows up every five days, and hands RFQs, drawings and NDA requests straight to Garrett Noach.
- Client
- Lamothermic, Brewster NY
- Sender
- Garrett Noach, President
- Channel
- Window
- Oct 2025 to Oct 2026
- Data as of
- 9 October 2026
LinkedIn reply rate: 293 of 879 engineers and buyers messaged
buyer-side outcomes: quote requests, drawings, NDAs, referrals and calls
of all replies came from follow-ups, not the first message
median reply time to a prospect since May 2026
In short
- A foundry is sold engineer to engineer. The agent writes as the president and asks one short engineering question, never a pitch.
- Follow-ups lifted the reply rate from 17.2% to 30.9% and cost almost no goodwill: 7 negative first replies across 1,699 follow-up recipients.
- Every replied thread was read by hand. 42 conversations ended with a buyer raising a hand, and the agent handed each one to Garrett by email.
9 min read
A foundry is sold engineer to engineer
Casting buyers are design engineers, sourcing managers and quality leads at OEMs. They ignore vendor pitches, but they answer a peer who asks a sharp question about their parts. Garrett Noach, president of Lamothermic, is that peer. A foundry president, though, cannot prospect on LinkedIn every day.
Lamothermic is an investment-casting foundry in Brewster, New York. It pours steels, stainless and bronze, and machines and finishes parts in-house. The target buyers are engineers and buyers at US OEMs that use cast parts: valves and pumps, heavy equipment, defense and aerospace controls, firearms, oilfield and agriculture.
Sellinger runs an AI agent that writes as Garrett, from his own LinkedIn profile. It finds prospects, connects, asks one diagnostic question, follows up every five days, answers within minutes, qualifies part, alloy, volume and timing, and hands every buying signal to Garrett by email.
AI illustrationWhat
Investment castings, machined in-house
Steel, stainless and bronze parts for valves, pumps, heavy equipment and defense.
AI illustrationWho
Engineers and sourcing leads at US OEMs
People who live with porosity, tolerance stack-up and long lead times.
AI illustrationHow
One question on LinkedIn, from the president
A bump every five days, and a hand-off to email when a buyer raises a hand.
Conversion on this account means the prospect raises a hand: asks for a quote, sends drawings, requests an NDA, names the right buyer or agrees to a call. The agent then steps back and Garrett takes it to email, a median of two prospect messages into the thread.
The playbook: connect, ask one question, follow up every five days
The sequence the agent ran for every prospect. Timing comes from the data: the median gap between touches was 5.0 days.
Day 0
Connection request
Usually without a note (149 of 3,345 had one). 27% were accepted.
On accept
One diagnostic question
A median of 33 words. 81% contain a question. No pitch, no link.
+5 days
Bump
One line plus an easy out: “if not a priority, no worries”.
+10 days
New angle
Tariffs, tooling lead times, MRO spares, a two-minute foundry video.
+15 days
Soft close
“I’ll get out of your inbox.” It still drew late replies.
Any reply
Diagnose, then hand off
Part, alloy, volume, timeline. Then email, drawings, an NDA or a 15-minute call.
Where the prospects came from
| Lead source | Messaged | Replied | Buyer outcomes |
|---|---|---|---|
| Engineers and buyers at named target OEMs (LinkedIn search by company and title) | 449 | 27.6% | 24 (5.3%) |
| Title and keyword searches (buyers, sourcing, engineers) | 210 | 30.5% | 8 (3.8%) |
| People engaging with competitor foundries’ pages (Sellinger intent signals) | 141 | 58.2% | 6 (4.3%) |
| Imports, Garrett’s network, manual adds | 79 | 29.1% | 4 (5.1%) |
Intent-signal lists replied twice as often, but 57% of those replies came from recruiters, vendors and peers. Named-OEM searches produced 24 of the 42 buyer outcomes.
Rules the agent works by
- Write as Garrett: a practitioner and a smart filter, not a salesperson. A polite disqualification counts as a win.
- US manufacturing only. Never sales, marketing, HR, consultants, other foundries or personal connections.
- Ask about the part, alloy, volume, application and timeline. Mirror the prospect’s length, one to four sentences.
- Not the right person? Ask who handles sourcing, and for an email.
- State capabilities exactly: steels, stainless and bronze; no aluminum, no vacuum-melt alloys. Garrett added these guardrails as the campaign ran.
From 3,345 connection requests to 42 buyer-side outcomes
Every one of the 293 replied threads was read and classified by hand, because CRM stage labels overstate outcomes. A buyer-side outcome means a buyer or engineer at a target OEM asked for a quote or capabilities, sent drawings, asked for an NDA, named the right buyer, or agreed to a call.
AI illustrationThe funnel
LinkedIn, 27 Oct 2025 to 9 Oct 2026
- Connection requests3,345
- Accepted and messaged879
- Replied293 (33%)
- Advanced, any type63
- Buyer-side outcomes42 (4.8% of messaged)
The 42 buyer-side outcomes
- Asked for a quote, info or NDA23
- Referred us to the right buyer12
- Agreed to a call6
- Booked a call1
of all connection requests ended in a reply (293 of 3,345)
replies became a buyer-side outcome (42 of 293)
messages in a typical replied thread (median)
Volume fell after July because the lead list ran out, not because the agent slowed down: prospects first messaged in July still replied 51%. A short email test in June (47 leads) got no replies at all; for this audience, LinkedIn won.
Follow-ups added 79% more replies and cost almost no goodwill
Counting only replies to the first message, the reply rate would have been 17.2%. With the follow-ups it reached 30.9%. Across 1,699 follow-up recipients only 7 first replies were negative.
Reply rate at each touch
of prospects who got that message without having replied yet
- Intro17.2% (157/911)
- Follow-up 111.7% (82/701)
- Follow-up 23.7% (22/597)
- Follow-up 33.0% (12/401)
- Follow-up 43.9% (6/153)
Which message earned the first reply
281 prospects who replied to outreach
- Intro56%
- Follow-up 129%
- Follow-up 28%
- Follow-up 3+7%
Purchasing lead, aerospace and defense components group
The bump that found a live RFQ
- Garrett (agent)Day 0Dan, given [company]'s focus on mission-critical components for aerospace and defense, I'm curious. Do you ever find it a challenge to source complex investment castings that come fully machined and finished from a single, reliable partner?
One question tied to his employer’s products. No pitch, no link.
- Garrett (agent)Day 5Dan, just bumping this. Let me know if consolidating casting and machining suppliers is a priority for you this quarter.
Bump 1, on schedule, time-boxed.
- Dan, purchasing lead+11 hI am about to start an rfq for some castings. Send me email with your contact details.
The buying window was open. Without the bump this RFQ is missed.
of the 42 buyer outcomes first replied to a follow-up, not the intro
of 416 nudges sent after a prospect went quiet got an answer; 13 buyer outcomes came after one
negative first replies to 911 opening messages
Answers in minutes, including nights and weekends
Prospects replied a median 1.5 days after the first message. When they did, the agent was usually back before they had left the conversation.
How fast the agent answered
138 prospect messages since May 2026, share answered within
- 5 minutes75%
- 1 hour94%
- 24 hours96%
median reply since May 2026
of 690 prospect messages arrived in the evening, at night or on a weekend (US Eastern)
of those were answered before the next business morning
of replied threads since May ran with no human message
An engineer who writes back at 9 pm about a porosity problem gets a relevant follow-up question at 9 pm, while the problem is on their mind. Garrett’s own typed replies took a median of 29 minutes.
Six patterns behind the 63 conversations that moved forward
Shares are of the 63 advanced threads; a thread can show more than one pattern.
Ask an engineer about their own parts
The opener names the prospect’s product family and asks one question. Engineers answer with paragraphs about porosity, tolerance stack-up, tariffs or a caster that changed hands, and the agent mirrors that pain before asking for anything.
68% of advanced threads began with a reply to the first message.
Hand off within two messages
The most common win is a hand-raise: send capabilities, quote this part, sign an NDA. The agent collects the email and stops; Garrett continues by email.
25 hand-raises, a median of 2 prospect messages in.
Bump at the right moment
A one-line bump five days later finds the buyer when the need is live. Several RFQ-grade replies arrived hours after the first bump.
17 of 42 buyer outcomes first replied to a follow-up.
Nudge with the prospect’s own words
When a prospect goes quiet, the next message quotes the pain they named (“you mentioned cost and quality…”). NDAs and bookings arrived after the second to seventh nudge, across holidays.
13 buyer outcomes followed a nudge.
Turn the wrong person into a referral
“Who handles sourcing?” turned “not my department” into named buyers, supply-base managers and forwarded introductions.
14 referrals, 12 of them to buyers.
Peer-to-peer framing
Openers that named Garrett and Lamothermic, or spoke “as a fellow manufacturer”, drew more replies and curiosity (“I checked out Lamothermic after I saw your message”).
46% vs 35% reply rate (Q4 2025 intros).
Inside the conversations
Three real threads, condensed. Prospect names are changed; the agent’s and prospects’ words are verbatim.
AI illustration“I am about to start an rfq for some castings. Send me email with your contact details.”
Purchasing lead, aerospace and defense components group
R&D director, oilfield equipment maker
From “we only do surface equipment” to “my engineers will test it”
- Garrett (agent)Day 0Matt, as Director of Engineering at [company], what's the biggest challenge you're currently facing with the high-performance metal components for your drilling equipment?
Role-specific diagnostic opener.
- Matt, Director of EngineeringDay 1At this time, we only have surface equipment [...] Challenges we face are typically with respect to a metallic coating that can survive high PV values in a dirty environment (drilled cuttings).
A correction, and a real pain.
- Garrett (agent)+1 h 40Thanks, Matt! That's a clear challenge for surface equipment. When those components need to stand up to high PV in dirty environments, is the focus more on the base material's integrity or achieving a consistent surface for coating application?
Accepts the correction and narrows with an either/or engineering question.
- Matt, Director of EngineeringDay 5It is both, but most importantly is the surface finish and hardness of the shaft @ the seal surface.
- Garrett (agent)Day 14Matt, just circling back on that seal surface challenge! What kind of annual volumes and alloys are you typically running for those shafts? [...]
A nudge anchored to his words, with qualifying questions.
- Matt, Director of EngineeringDay 17Apologies, hit the road for some work travel [...] I run the Engineering dept & R&D lab so anything that is a possible solution I can have my Engineers review it and do prototype/field testing
The nudge reopened a thread that had been silent for nine days.
- Garrett (agent)+9 min[...] We could share some focused technical details or relevant case studies for your R&D engineers to review. How does that sound?
A low-friction offer that suits an R&D lead, not a call.
- Matt, Director of EngineeringDay 17Sounds fine to me, but I'm sure they'll be skeptical of how investment casting is a solution [...] But, hey. Don't judge a book by the cover right! My email is [email]
Hand-raise. He talks himself past his own skepticism.
Why it worked: The agent kept asking engineering questions instead of pitching, came back after his travel with his own words, and offered an asset his team could test.
Account lead, industrial equipment dealer, found through engagement with a competitor’s page
Cold connection to a booked call, with the prospect proposing it
- Garrett (agent)Day 0Hi Cole, I saw your work at [company] with industrial automation solutions. Are your manufacturing clients running into any challenges with component reliability or sourcing for critical parts right now?
He sells to manufacturers, so the agent asks about his customers’ pain.
- Cole, account leadDay 0[...] fluctuating lead times, obsolete parts, and unpredictability. [...] Trucks sitting for 6 months in the shop waiting for parts
- Garrett (agent)+1 min[...] Are these issues primarily affecting cast or machined components for your clients?
One narrowing question, a minute later.
- Cole, account leadDay 1I'm seeing it much more on the machined components sir
- Garrett (agent)+3 min[...] With machined components, sometimes teams explore converting them to investment castings for improved lead times or performance. Is that something your clients consider for new designs or component redesigns?
Introduces the offer as a question: machined-to-cast conversion.
- Cole, account leadDay 2That would be awesome, thanks Garrett [...] Next week I'll be out of state for work. Pretty booked each day so maybe the following week we could connect over a call.
The prospect proposes the call.
- Garrett (agent)Day 39Hi Cole, following up on our last chat about exploring investment casting for machined components. Were you able to find some time to connect after your trip?
Picks the thread back up on the original topic.
- Cole, account leadDay 41Hi Garrett, next week is looking pretty open for me. [...] What works for you?
- Garrett (agent)+1 minThanks, Cole! Glad to hear next week works for you. Feel free to grab a time on my calendar here: [booking link]
Booking link within a minute.
- Cole, account lead+4 minDone, thank you!
Booked.
Controls design engineer, former casting engineer
Two months of patient nudges, then an NDA request
- Garrett (agent)Day 0Sam, saw your background developing complex investment castings at [a major aerospace caster]. Given that experience, I'm curious what the biggest challenge is when sourcing reliable cast components for the controls you're designing at [company] today.
Career-path hook: peer to peer.
- Sam, design engineerDay 11Oops, that was a fat finger send. Cost and quality control are the biggest challenges.
First reply came after the second follow-up.
- Sam, design engineerDay 21I'm not working till after the new year.
Timing objection.
- Garrett (agent)+6 minGot it, Sam! Appreciate the heads-up. I'll circle back after the new year. Enjoy your time off!
Respects the timing, keeps the door open.
- Garrett (agent)Day 33Hey Sam! Hope the new year is off to a great start. Looping back on our chat about the cost and quality challenges with your cast components. Ready to pick that conversation up?
Comes back when promised.
- Garrett (agent)Day 52Sam, is there a specific alloy or part geometry that's giving your team the most trouble with quality right now? I'm happy to do a quick DFM review if you're hitting a wall on a design [...]
Changes the ask from a call to a concrete DFM review.
- Sam, design engineerDay 55Hi Garrett, I would need to get an NDA signed by you guys to discuss specifics
A real program behind the pain.
- Garrett (agent)+1 minAbsolutely, Sam! We can definitely get an NDA in place. What's the best email for me to send that over to you?
- Sam, design engineerDay 55I'm sure that it's going to have to come from us. I'll talk to our supply chain.
Hand-raise: the prospect takes it internal.
How the agent handled the objections engineers raise
Counts are replied threads where the objection appeared. Quotes are the agent’s actual words.
| Objection | Threads | What the agent said | What happened |
|---|---|---|---|
| “Not my department” | ~23 | “Is there a supply base manager you’d recommend I connect with to start that process?” | 14 referrals to named buyers or forwarded intros |
| We already have overseas or established suppliers | ~23 | “Even with established overseas manufacturing, I often find teams value a domestic partner for critical components where tight communication and consistent quality are key.” | Mostly “keep us in mind”; one asked for a line card |
| Volumes too small, cost too high | ~14 | “The cost hurdle with lower volumes is a common challenge for many teams. We often find that upfront engineering collaboration can sometimes overcome that by optimizing the part and process for overall value.” | One engineer moved to “interested in hearing more” and got the booking link |
| “Is this a sales pitch?” | 6 | “Fair question, Mike. My intent is a genuine conversation about operational challenges, not a sales pitch.” | The prospect reacted with applause and promised a detailed answer |
| “We’d need an NDA first” | 4 | “Absolutely, Sam! We can definitely get an NDA in place. What’s the best email for me to send that over to you?” | All 4 became hand-raises (defense and firearms programs) |
| Capability questions (alloys, size, location) | 26 | “We primarily pour steels and some bronze, but not vacuum melt alloys like those often used for turbine blades.” | Answered precisely once Garrett set the capability rules |
“We actually have a current need for an investment casting [...] I can send you details of the part and we can see if it would be a good fit.”
“I'm open to new suppliers if they can support our relatively low production numbers (2500/yr) and can be responsive to expedited orders.”
“I certainly have some things you can quote. Where should I send drawings?”
“We have been putting feelers out there for potential new suppliers for some of our casting parts due to high tariffs and long lead times.”
What made it work
- 1
Send from the person the buyer wants to talk to. A president asking an engineering question earns replies a vendor pitch never gets.
- 2
Ask, don’t pitch. A 33-word diagnostic question with no link was the opener behind two thirds of the conversations that moved forward.
- 3
Keep following up, briefly. Follow-ups produced 44% of all replies and 40% of buyer outcomes, with almost no negative reactions.
- 4
Measure the right list. Intent-signal audiences reply twice as often, but named-account searches produce the buyers.
- 5
Give the agent exact guardrails. Once Garrett listed what Lamothermic does and does not pour, the agent answered capability questions precisely, and honest disqualification became a trust signal.
How we measured this
- Source: Sellinger production data for the Lamothermic workspace, read-only, snapshot 9 October 2026. LinkedIn only; a short June 2026 email test (47 leads, 0 replies) is excluded from LinkedIn figures.
- Messaged = received at least one message after connecting. Replied = wrote back after our first message; auto-replies excluded.
- Touch = the n-th message before the prospect’s first reply. Reply rate per touch = of prospects who got that message without having replied, the share who replied before the next one.
- Outcomes come from reading all 293 replied threads, not from CRM labels. The lead-source table is computed from that per-lead classification. What happened after the hand-off to email is not visible to us, so no deal values are claimed.
- Response time = prospect message to the agent’s next message; replies typed by Garrett are excluded. Since May 2026 every message carries an AI or human tag.
- Prospect names in conversations are changed and companies are described, not named. Images are AI illustrations, not photos of Lamothermic.
Want numbers like these for your pipeline?
Book a 15-minute walkthrough of the agent, the playbook and the numbers behind these case studies.
