GTM Engineering in 2026 → Nautilus
Five principles from the
best GTM engineers —
translated to selling car washes.
An 82-minute walkthrough by The Kiln, distilled into what we keep, what we change, and what we build next.
Why this video
Hundreds of GTM systems built for clients from solo founders to the Fortune 500 — and he says it all expires in 4–8 months.
Patrick Spychalski was an early Clay employee, then co-founded The Kiln, a GTM-engineering agency. This is his field report: what worked, what didn't, where it's going.
- Five tenets every system should obey
- Eleven use cases that actually produced pipeline
- Tools he trusts vs. tools he'd skip
- A clear bet: agents run GTM within a year
“If you're watching this 8 months past September 2026, don't watch it.”
The shelf life is short. That's an argument to move now, not to wait for the dust to settle.Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
The one pattern underneath everything
Every system has the same four layers. Only one of them is new.
His point: the reasoning layer is why systems are “so much more scalable and so much more effective.” Ours exists, but it sits beside the pipeline instead of inside it.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Tenet 1 of 5
Tools and models change fast. Buy and build like it.
What he says
Mid-market buyers sign 1–2 year contracts; “in the world of AI, 1 to two years is quite a bit of time.” Ask whether you'll still need the feature in a year. If a system isn't “flexible, adaptable, malleable to the new models… it probably shouldn't be built.”
What it means for Nautilus
- No annual data contracts. Apollo, Blitz, Deepline, Serper — month-to-month, swap when coverage drops.
- Monaco over MCP is the right bet — the CRM is already agent-addressable, so the model can change without the system changing.
- Build thin. Shisha's scripts call providers behind one interface; a better email finder is a one-line swap, not a rebuild.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Tenet 2 of 5
Simple and usable beats clever. Spoon-feed the output.
What he says
“Builder syndrome”: a clean single-purpose workflow grows into “this rapidly more and more complex system that nobody can adopt other than you.” The fix is always the same — “these systems should be spoon feeding the output to whoever's using it.” A rep should open the tool and find “15 different prospects that are completely researched with a message drafted.”
What it means for Nautilus
- The deliverable is a ready-to-send step in Monaco, not a Slack digest about prospects. Research, signal, draft — already in the sequence when Sam or Amayr sits down.
- One job per automation. Enrich inbound. Score accounts. Draft the first line. Separate, inspectable, replaceable.
- Adoption test: if it needs a how-to, it's too complicated.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Tenet 3 of 5
Within a year, agents build and run the systems. Humans orchestrate.
“Go to market engineers of the future… will be orchestrators of agents and visionaries for systems and that will be about it. Everything else will be completed and built and tested and run by agents.”
Point-and-click workflow building “is not going to be the case within the next year.” Knowing Claude Code is “an absolute must.”
Where Nautilus already is
- Two prospect websites (Fast5 Xpress, Carnation) built end-to-end by an agent from a URL and a spec.
- Onboarding contracts created, swapped and re-issued through the product by an agent in minutes.
- All-time contact ledgers (2,652 people) assembled from three inboxes without a human export.
The gap: these are one-off asks. The next step is agents that run on a schedule and feed Monaco without being asked.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Tenet 4 of 5
Data is still the core. Care about it as much as the system.
What he says
“Regardless of how well-built an automation is, the underlying data is what dictates the quality of the system itself.” If your email finder covers 50%, half your leads are “left on the table.” Provider quality varies massively — build waterfalls across providers to push coverage toward 100%.
What it means for Nautilus
- Our buyers are owner-operators on Gmail. ~175 of the people we've emailed use @gmail.com. B2B enrichers are weakest exactly there — so we need a waterfall, not one provider.
- Local-business list building is his named recipe: Google Maps / Serper → Exa or Parallel for the owner → Deepline / ZeroBounce to verify. That is our TAM.
- Set a coverage bar (e.g. 90% verified email on Tier-1 accounts) and measure providers against it monthly.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026. Gmail figure: Nautilus all-time email contact ledger, Oct 2026.
Tenet 5 of 5
Every system needs a direct line to closed-won. Be discerning.
What he says
Engineers build what's fun; pipeline comes from what's boring. Thousands of AI tools claim a line to revenue — “the most discerning GTM teams are going to be the most successful in the age of AI.” His cautionary tale is the AI SDR: great in theory, but ask where the data comes from and whether the copy is slop. “Every syllable, every line break, every dash matters.”
What it means for Nautilus
- Two numbers per play: demos booked, onboardings signed. Nothing else counts as a result.
- No AI SDR. Agents research and fill; a human writes the template and the first line.
- Kill rule: a play with no demo attributed in 60 days gets paused, the way the 80 already-contacted accounts were paused this month.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Use case → Nautilus
ICP from closed-won, then two scores — not one.
His method
Feed an LLM 100+ closed-won deals to define the ICP; with fewer, test messaging across candidate ICPs instead. Score accounts twice: a firmographic fit score and a separate signal score that triggers alerts. Calibrate on 10–20 known great fits and 10–20 known bad fits.
Nautilus fit score (proposed)
- Express tunnel with a membership program
- 2–50 locations (Surf Thru at 46 is our ceiling case; single sites close too)
- POS we integrate: DRB, Sonny's, ICS, NXT, Washify, Micrologic…
- Owner-led, not PE-run
- Website with no online checkout, or a Rinsed contract
Nautilus signal score (proposed)
- New site permit / “opening soon” on Maps
- Marketing or membership-sales job posting
- POS contract renewal window
- Visit to nautilus.co pricing after a demo
- Champion changes jobs
We have ~43 customer accounts in Monaco. Fewer than 100 — so per his rule, we define the ICP from them, then test it against two or three adjacent segments rather than treating it as settled.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026. Customer count: Monaco accounts with status Customer, Oct 2026.
Use case → Nautilus
“There's no alpha in using the same signal as every other company in your space.”
He weights signals by funnel stage and by relevance, and warns that ten signals piped to Slack become noise reps stop reading. His favorites: job-listing text (“the most underrated signal”), champion tracking (“90 plus% conversion… no joke”), and signals no competitor uses.
Signals only a car-wash company would track
- GMs and regional directors moving between chains. Our champions at Surf Thru, MOJO, Splash change jobs; they bring us with them.
- Job posts for “membership sales” or “marketing manager.” An operator hiring to grow memberships is our buyer, this quarter.
- Google review velocity drops at a site — retention problem we fix.
- New tunnel permits / construction — they're choosing a POS and a website now.
- Rinsed or DRB e-commerce on their site — a renewal date exists.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Use case → Nautilus
Outbound: a human writes the template, agents fill it, and deliverability is non-negotiable.
His rules
- 60–70% of companies he meets have deliverability problems
- 15–20 emails per inbox per day, no more
- Inboxes split 50/50 Google and Microsoft; no private servers
- “Don't have agents do the copywriting for you.” Template by a copywriter, agent-written variables inside it
- The offer and the first sentence decide everything: “Most people aren't reading past the sentence if it's garbage”
Nautilus check
- Our “Short Question Play” already follows the template-plus-variables shape. Keep it; audit the first sentence per persona.
- Inbox math: at 15–20/day per inbox, the current campaign volume needs dedicated sending domains, not founders' primary inboxes.
- Human-in-the-loop fits our ACV. He reserves fully automated outbound for low-ACV, huge-TAM companies. We are neither — reps approve, agents prepare.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Use case → Nautilus
Inbound: cut the demo form to name + work email, enrich the rest, route after enrichment.
His claim
Forms with 10–20 fields bleed conversion; reducing fields can “double or triple your conversion rate.” Ask for name and work email, keep “How did you hear about us,” enrich company, size, title and phone with a near-100% waterfall, then let the CRM route.
Our demo booking today
Cal.com asks for: name, email, car wash name, POS system, number of locations, phone, plus notes.
- Keep POS and location count — no enricher knows a wash's POS, and both decide who takes the call.
- Drop phone (enrich it) and make the car-wash name optional (infer from domain or Maps).
- Add “How did you hear about us.” We currently can't attribute demos to the podcast, PC&D, or referrals.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026. Form fields: Cal.com booking confirmation, Oct 2026.
Use case → Nautilus
ABM assets and event plays — we have an unfair version of both.
The asset nobody else can send
He describes generating custom landing pages and one-pagers per target account. We can build the prospect's actual new website — Fast5 Xpress and Carnation Auto Spa were each built from their URL in under two hours, with a working checkout drawer and their real pricing.
Play: for every Tier-1 account, send the live site, not a deck. The demo becomes “here's what your customers would see next week.”
Event play (ICA, Car Wash Show, SCWA)
His Dreamforce recipe: scrape LinkedIn posts mentioning the event, have Claude filter to real attendees, qualify in the CRM, message before the show. The client “couldn't even attend” all the meetings it booked.
Ours: same recipe on #TheCarWashShow and regional shows, with a pre-built Shipyard mock for the ten accounts we most want at the booth.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026. Build times: Nautilus Shipyard builder runs, Oct 2026.
Use case → Nautilus
Meetings write themselves into the CRM. Hygiene runs forever, not once.
Meetings
Pre-call prep as a Claude skill that runs automatically; post-call, an agent updates the CRM with human approval. Proposals drafted by a skill trained on past proposals.
Nautilus: Gemini and Pocket already transcribe every call; Monaco already holds the transcript. Missing piece: the approval step that turns a transcript into updated stage, next action, and a drafted follow-up.
CRM hygiene — “arguably the most important”
Four failure types: stale, incorrect, empty, non-normalized. Normalize, validate, then enrich; run continuously on new records. He is emphatic that pushes to the CRM should be deterministic, not agentic.
Nautilus: this month we tagged and paused 80 already-contacted accounts by hand. The standing version is a nightly job: dedupe, verify emails, flag already-contacted, and propose changes for a human to apply.
Source: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
His tool calls, our stack
Where we already agree, and where we'd differ.
He recommends
Nautilus
today → proposedSource: Patrick Spychalski (The Kiln), “GTM Engineering in 2026 — A Complete Walkthrough,” YouTube, Sept 2026.
Proposed · next 90 days
Five builds, each with a demo-booked number attached.
- Demo form trim — name, email, POS, locations, “how did you hear.” Measure booking rate before/after.
- Already-contacted guard — nightly job proposes DNC/pause for anyone a human has emailed.
- Fit + signal scores in Monaco, calibrated on our 43 customers and 20 known bad fits.
- Champion tracking — every past contact at a customer, alert on role change.
- Morning feed — 15 researched prospects with a drafted first line waiting in Monaco for Sam and Amayr.
- Shipyard-as-ABM — a live site for each Tier-1 account before outreach.
Owners and dates to be set by Sam and Amayr. Nothing here is live yet; this is the proposal.
“The most tasteful and discerning go-to-market operators of the future will prompt and build like an art form.”
We already build the thing most GTM teams only mock up. The work now is making it run every morning without being asked — and counting it in demos.