The ERP for the AI Revolution is here

Glasgow — The ERP for the AI Revolution is here (YC Root Access)

John Glasgow is the CEO and founder of Campfire (S23), an AI-native ERP platform built for high-growth tech companies that automates accounting and financial reporting workflows using a custom foundation model and agent platform — the company recently raised a Series B led by Accel and Ribbit Capital, has more than doubled ARR each quarter since Q4 2024, and now exceeds 100 employees after closing a $35 million Series A in June 2025 with just 12 people.

In this fireside, John sat down with Andrew Tan to talk about how he went from a Google Sheets prototype to pulling customers off NetSuite, why staying in founder-led sales until $1M ARR was critical, how narrowly targeting the needs of tech companies gave Campfire a wedge into a decades-old market, and why the shift to AI-native infrastructure flipped the narrative on what “safe” software buying actually means.

— channel description, YC Root Access

TL;DR

A ~27:38 founder fireside on the YC Root Access channel, published 2026-05-20. Andrew Tan (YC partner; ex-CTO at PagerDuty) interviews John Glasgow, CEO and founder of Campfire — a YC S23 AI-native ERP platform that has, since Q4 2024, doubled ARR every quarter while pulling enterprise customers off NetSuite.

The substantive contribution is the “AI-native safety inversion” thesis (~24:08–24:31). Until late 2024, the buyer-side calculus on enterprise software was buying the incumbent = safe. From Q4 2024 onward — call it the AI-native flip — being the incumbent meant you were not AI-native. Boards and executives began explicitly asking for AI-native, providing the air cover their accounting teams (who weren’t yet ready to embrace AI) needed to buy something new that “nobody had heard of and their C-suite or their auditor wasn’t familiar with yet.” The risk premium reversed: AI-native vendors are now the safer-feeling buy from the procurement-decision-maker’s perspective, even when the day-to-day operator is hesitant.

Companion claims anchored in Glasgow’s founder-vantage experience:

• Tech-stack-turnover argument for the “why now” (~22:44–23:18): everything in the finance tech stack — payroll, spend management with the Brexes of the world — had turned over in the last 5–10 years except the core general ledger. The contrast made a finance person logging into a modern spend-management tool and then the ERP feel “very acute,” prompting them to seek out the modern alternative.
• Wedge displacement beats feature-completeness (~5:41–6:42): the standard incumbent-disruption advice (“you can’t disrupt an enterprise category until you’re feature-complete”) is wrong in tech-company ERP. Tech-company customers were only using a narrow slice of NetSuite; Campfire owned that slice (approval workflows, multi-entity accounting, audit controls) and was pulling customers off NetSuite within nine months of building.
• Founder-led sales to $1M ARR doctrine, even in the AI era (~14:05–15:11): “I do really recommend founders to stay in the founder sales mode… offloading it to AI, offloading it to some AE — I think it feels like, oh let’s just bring in a professional whether it’s an agent or whether it’s a human… but I still recommend being as close to the customer and the prospective customer as possible until you have product market fit.” Glasgow personally was the only AE through $1M ARR.
• Product velocity as enterprise trust signal at seed stage (~7:17–8:38, ~10:20–10:39): $300M-revenue customers with four-employee vendors. “Literally we are making a venture investment in you by going with you… I’m literally getting fired if you shut down… this contract is longer than your runway.” What closed the gap was build-pace as a forward-looking signal: customers (Replit, PostHog cited as examples) explicitly told him “as we add new global subsidiaries, as we need new features… we feel you’re going to be able to stay ahead of us.”
• Own foundation model + custom agent platform as competitive moat (~17:47–17:54): in the named-AI-native-ERP cohort, Glasgow claims Campfire is “the only one with our own foundation model, our custom agent platform” — one of three customer-cited reasons for picking Campfire over both incumbents and AI-native peers (alongside public-company-grade audit/controls/approval workflows and best AI).

Caveats. The transcript is auto-generated and ASR-cleaned, not human-curated. Glasgow is the company’s founder-CEO speaking on a YC marketing channel — every substantive claim is filtered through obvious commercial incentives (talent recruiting, customer acquisition, future-round signalling). The empirical anchors (>100 customers, doubled ARR/quarter since Q4 2024) are concrete and falsifiable but unaudited in this venue. The “AI-native safety inversion” thesis is named here from a single vendor-side vantage; it pairs with Emergent’s independent observation of the same vendor-side shift in a different product category, which is what raises this beyond a single-source curiosity.

Why this matters in the corpus

The wiki holds three prior 2026 anchors on the AI-native-vendor advantage thesis, each from a different vantage:

1. HBS / HBR research vantage — Nishar & Nohria 2026 frame the firm-boundary shift under GenAI as a four-model collapse of one-size-fits-all enterprise software (build / compose / collaborate / buy outcomes). Their thesis is buyer-side and macro.
2. AI-native company-building doctrine — YC (same channel, ~4 weeks earlier) names the AI-native-from-day-one advantage as a structural constraint legacy incumbents cannot unwind on the relevant timescale. Hu’s thesis is supply-side and systemic.
3. Consumer / SMB / long-tail founder vantage — Emergent (one week earlier; ingested in the Claude-channel customer-story batch) names the same vendor-side flip at the adoption-velocity-asymmetry-favouring-the-long-tail end: $100M ARR in 8 months serving 7M users in 190 countries, 70–80% of whom never wrote a line of code.

Glasgow is the enterprise-B2B founder-vantage twin of Jha/Emergent. Where Jha names the long-tail asymmetry, Glasgow names the enterprise procurement-room flip: the same underlying vendor-side advantage observed independently at the opposite end of the customer-size spectrum. The two together — paired 2026 founder-CEO empirical anchors, opposite ends of the market — are now the wiki’s strongest pair on the AI-native-vendor advantage thesis. Glasgow also instantiates Hu’s intelligence-layer / AI-native-from-day-one claim with the Campfire worked example (own foundation model + custom agent platform) and provides the vendor-side empirical evidence for Nishar & Nohria’s macro buyer-side thesis (hard numbers on actual NetSuite-to-AI-native-vendor migrations from Q4 2024 onward).

The W&W cell-overlap with Jha/Emergent’s tagging (digital-seizing/strategic-agility, digital-seizing/balancing-digital-portfolios, contextual/external-triggers) is 3-of-3 — same digital-seizing posture, same external-trigger reading. The substantive contribution of this source against that cluster is the enterprise procurement-room observation: the same trigger, exploited via a different go-to-market and against a different incumbent class than the consumer/SMB Emergent/Replit cohort.

Substantive content

The “AI-native safety inversion” (the headline claim)

Glasgow’s strongest, most quotable claim arrives at ~24:08–24:31:

“Buying the legacy version was considered very safe. But then once AI started to take off — call it in the finance category, call it like a year and a half ago, call it, you know, end of ‘24 — it then become being the incumbent meant you were not AI-native. And so there was a flipping of the narrative that the board and the executives were saying, we want AI-native. And even if the accountant wasn’t fully ready to embrace AI, they had this blessing to go buy something new that nobody had heard of and their C-suite or their auditor wasn’t familiar with yet.”

Three load-bearing details for the wiki:

1. It is a narrative flip, not a feature flip. Campfire had AI features earlier (“we had AI pretty quickly” ~23:43); what changed between mid-2024 and mid-2026 was what the board would underwrite as a safe choice. The shift is on the procurement-decision-maker’s side, not on the product side.
2. The shift moves through the C-suite / board / auditor channel, not the day-to-day operator channel. “Even if the accountant wasn’t fully ready to embrace AI, they had this blessing” — the day-to-day operator’s hesitancy is overruled, not converted.
3. The narrative had to shift even though the core product didn’t. “Your narrative needs to continue to shift even if the core product ultimately, you know, ends up not shifting” (~24:50–24:59).

This is a different shape of buyer-side shift than the slow-organisation-adoption framing in much of enterprise-ai-adoption: it is fast at the procurement layer, slow at the operator layer, and the fast-procurement layer overrides the slow-operator layer because the board explicitly authorises it.

The tech-stack-turnover argument for “why now”

Investors at seed asked Glasgow “why now? This is a very established market, mission critical software” (~2:58–3:07). His answer (~22:44–23:18):

“Everything in the finance tech stack — like payroll, spend management with the Brexes of the world — had all turned over in the last 5 to 10 years except the core general ledger. And so there was, you know, a finance person would log in and they would go into this amazing spend management solution or accounts payable or payroll, and then the contrast with the ERP and everything else was very acute.”

The mechanism: every adjacent finance tool already got its modern-SaaS rewrite; the GL is the last domino. Finance teams experiencing the contrast daily were the ones reaching out, not the ones being sold to. The acuity-of-contrast claim is testable against the published modernisation timelines of payroll/spend-management/AP categories (which a future ingest could anchor).

Wedge displacement over feature-completeness

The classic incumbent-defence in enterprise software is “you can’t disrupt this category until you’re feature-complete.” Glasgow’s wedge claim (~5:41–6:42):

“Within our initial market of tech companies, they’re actually not using a lot of NetSuite — most of NetSuite they’re actually not using. So we very narrowly focused on a certain segment which was like tech companies that had outgrown QuickBooks, that needed something more powerful or more mature for certain specific features… approval workflows doesn’t sound sexy at all but it’s a key reason that folks need it for audit and QuickBooks doesn’t have that. Multi-entity accounting where you have a subsidiary — particularly in the AI native world folks are going multi-entity very quickly… we narrowly focus on those few features and then folks are saying we actually don’t have a lot but within nine months we were moving people off of NetSuite because we actually narrowly focused on being the best for a very specific profile of the customer base.”

The wedge: not be feature-complete; be the best for the narrow feature set that defined the migration trigger (here: approval workflows + multi-entity + audit/controls + revenue-recognition for SaaS reporting). The customer-segment claim that supports the wedge is tech companies use ~10% of NetSuite — which, if true, collapses the feature-completeness defence to the size of the actually-used slice.

Founder-led sales doctrine even in the AI era

Glasgow is the AI-native-vendor anchor of the wiki’s founder-led-sales concept (with Kolysh, Dinakaran, and Rubinstein & Onyemah).

Glasgow (~14:05–15:11):

“Even in the AI era I do really recommend founders to stay in the founder sales mode. I think offloading it to AI, offloading it to some AE — I think it feels like, oh let’s just bring in a professional whether it’s an agent or whether it’s a human. But I still recommend being as close to the customer and the prospective customer as possible until you have product market fit. I’m in every Slack channel with every customer… I’m trying to listen to as many sales meetings as I can to really hear, because then you’re the best person in the company to actually impact change on what is happening — the sales team driving engineering… I ended up getting to series A — the kind of classic million in ARR for us — on my own as the one AE.”

Two things worth filing for the wiki: (a) the explicit don’t offload to AI discipline is a counter-claim against the most aggressive agents-everywhere narratives — relevant to automation-vs-augmentation as a vendor-CEO-stated boundary; (b) the sales-team-driving-engineering feedback loop is the (un-named) operating logic behind why founder-led sales produces the product velocity customers later cited as a trust signal.

Product velocity as enterprise trust signal at seed stage

Two named anchors in the transcript — one customer-side, one peer-side.

Customer-side, on the four-person company selling to $300M-revenue customers (~7:57–8:30):

“One CFO told me, literally we are making a venture investment in you by going with you because you can only be in one ERP really at a time. And so he was like, I’m literally getting fired if you shut down. And we are so large, and this contract is longer than your runway.”

Peer-side, on what closed the trust gap (~10:20–10:36):

“A lot of YC companies obviously Replit, PostHog — a lot of what they tell us is, you guys are building so fast. As we add new global subsidiaries, as we need new features, as our usage-based revenue gets more complex, we feel confident you’re going to be able to stay ahead of us.”

The wiki-relevant pattern: forward-looking build-pace as the procurement-side substitute for present feature parity. This is consistent with the Emergent customer-story-cluster pattern of velocity as the feature, but Glasgow names it specifically as the enterprise-buyer’s trust mechanism when the vendor is seed-stage. To this day, “we’ve never had anybody outgrow Campfire” (~10:39).

Own foundation model + custom agent platform as moat

Brief but load-bearing for the agent-harness cluster (~17:47–17:54):

“We are consistently hearing we have the best AI. And so I think maybe it’s because we’re the only one with our own foundation model, our custom agent platform.”

Three substantive elements named:

1. A custom foundation model — Campfire is not just a thin wrapper on a frontier-lab API; Glasgow claims a vertical model for the finance/accounting domain.
2. A custom agent platform — Campfire’s own harness over that model, not someone else’s.
3. A competitive-moat reading of (1)+(2) — “the only one” — implicitly contra a hypothesised cohort of AI-native-ERP peers that wrap general-purpose models without a vertical foundation-model investment.

The claim is too lightly anchored in this source alone to ratify into the foundation-models or agent-harness concept page as a substantive convergence — but it is a vendor-side empirical data point that aligns with the harness-as-moat thesis already named on those concept pages by other practitioner anchors (Lopopolo, Jha, Fung, Osmani). Filed as a one-line addition to agent-harness’s enterprise-vertical-vendor cluster.

What was actually ingested

• Source format: auto-generated English captions from a 27:38 YC Root Access fireside interview (publish date 2026-05-20, view count 3,588 as of 2026-05-21 ingest).
• Transcript provenance: ASR-cleaned at acquire time. Fixed proper nouns (John Glasgow / Accel / NetSuite / Replit / PostHog / RevRec / ARR / PagerDuty / C-suite); stripped a YouTube ASR artefact where the spoken-timestamp accessibility text (“X minute, Y seconds”) leaked into the segment text from [1:02] onwards; light filler-word cleanup; no segment merging.
• No speaker labels at the raw layer. Attribution in this source page is by ear (Andrew = host, John = guest); the raw file leaves it flat.
• Reliability: founder-CEO speaking on his own company’s marketing channel. The empirical anchors (>100 customers, doubled ARR/quarter since Q4 2024, 4-person team selling to $300M-revenue customers) are concrete and falsifiable but unaudited in this venue.

Linked entities and concepts

Concepts touched (existing): enterprise-ai-adoption (substantive — new vendor-side enterprise-procurement-flip vantage), automation-vs-augmentation (modest — Glasgow’s automate manual work → focus on more strategic work framing for finance teams + his founder-led-sales boundary against offloading to AI), agent-harness (modest — vendor-vertical own-foundation-model + custom-agent-platform data point).

Entity references (existing): Y Combinator (the channel publisher, Glasgow’s accelerator).

Dangling (single-source first-mentions, deferred per the second-source promotion rule): John Glasgow (Campfire CEO/founder), Andrew Tan (YC partner, host), Campfire (S23 AI-native ERP), Accel, Ribbit Capital, NetSuite, QuickBooks, Brex, Mercury, Metronome, PagerDuty, Replit, PostHog, Dalton Caldwell (referenced as one of Glasgow’s YC interview partners).

Open questions

• Is the “AI-native safety inversion” reproducible across categories beyond finance? Glasgow names finance specifically and ties the inflection to Q4 2024 / “a year and a half ago” (relative to 2026-05-20). A future ingest could either ratify or contest this across HRIS, CRM, ATS, and other classical enterprise-SaaS categories. The Jha/Emergent pairing suggests yes on the long-tail end; the Nishar & Nohria framework suggests yes on the macro buyer-side; the missing piece is direct vendor-side founder testimony from another enterprise-B2B category.
• What is in the “custom foundation model”? Glasgow makes the claim once and does not elaborate. Is this a from-scratch pretrain (unlikely at Campfire’s scale), a domain-adapted continued-pretrain over a frontier base, or a fine-tune marketed as “own foundation model”? A future ingest naming peer AI-native-ERP vendors (Lovable for ERP-adjacent, Tabs, Numeric, etc.) could ratify or contest the “the only one” claim.
• Does the founder-led-sales-to-$1M-ARR doctrine hold under AI-agent-augmented SDR/AE workflows? Glasgow’s claim is explicit: don’t offload to AI yet. This is in productive tension with the more aggressive agents-everywhere narratives in the harness cluster. A future ingest from a 2026/2027 founder who reached $1M ARR with AI-led-sales-from-day-one would be a direct counter-anchor.