I started tracking this funding cluster in October 2025 when the DualEntry and Campfire rounds dropped within weeks of each other. Three tier-one syndicates, three portfolio companies, three variations of the same underlying read — all landing inside twelve months. When Lightspeed+Khosla, Accel+Ribbit, and a16z+ICONIQ pile into the same category simultaneously, that is not three independent views: it is consensus. What follows is my read on what they all saw, why the double edge matters more than the funding announcements admit, and what a CFO or CTO should actually do with this information.

The mechanism

ERP switching cost has always had two distinct components, and I find practitioners conflate them consistently. The first is transition risk — the operational exposure during the window when your old system is off and your new system is not fully validated. That risk is irreducibly real and AI does not touch it. The second is migration labor — the actual hours of consultant time spent mapping data, rewiring integrations, and recreating configuration. That is what AI is cutting.

The migration labor fraction is where the moat lived. A NetSuite implementation runs 6-18 months and $100-500K in implementation fees at the mid-market. Most of that cost is not platform complexity — it is pattern-matching work. Chart-of-accounts mapping. Reconciling historical transactions to a cut-over date. Verifying that every Stripe event and payroll run that hit the old system lands correctly in the new one. These are exactly the kinds of rote, rule-following tasks that LLMs are genuinely good at and that consultants have charged $150-300/hour to do for years.

McKinsey's 2025 analysis of AI-assisted ERP migration programs puts the reduction at two times cost and duration. That is not free switching — it is half-price switching. A migration that cost $200K and 12 months now models out at $100K and 6. For a CFO running a 15-day close and watching her finance team burn a third of their capacity on rote reconciliation, that calculation looks different than it did in 2022. That is the window the $313M is betting into.

The double edge

The mechanism is real. Here is what the funding announcements do not say: it cuts both ways.

The same LLM-assisted chart-of-accounts mapping, agentic integration tooling, and automated data transformation that makes it easy to migrate FROM NetSuite also makes it easy to migrate AWAY from DualEntry, Campfire, or Rillet after year two. Moat dissolution does not target incumbents specifically. Every entrant that builds better migration tooling to capture NetSuite customers is simultaneously building the infrastructure that makes those customers portable again.

This is not a fatal problem for the entrants — it is a Judo move with an expiration date. You exploit the open window to capture customers. Then you build permanent lock-in through a different mechanism. The permanent moat is not migration convenience; it is the AI model trained on your customer's specific data over time. Campfire's LAM, after processing 24 months of a company's transaction history, has learned that company's specific vendor payment patterns, revenue recognition edge cases, and intercompany elimination quirks. That trained institutional memory is meaningfully harder to replace than any generic model in month one. It is the same logic behind Bloomberg's financial models — the moat is not the data, it is the model trained on decades of it.

The entrants that build that secondary moat survive the window closing. The ones that rely on migration convenience alone do not.

The three bets

Lightspeed and Khosla backed DualEntry's speed-and-breadth thesis: 24-hour go-live, 13,000+ integrations, multi-entity consolidation, close times from 15-20 days to 8-10 days. The Altis diligence report (the most useful independent data available on DualEntry, free at altis.vc/reports/dualentry) finds DualEntry trailing Campfire and Rillet in SaaS-specific modules and notes some evaluators finding the automation depth below what was marketed. The strongest ICP is greenfield mid-market multi-entity companies that want the fastest path to any modern ERP.

Accel and Ribbit backed Campfire's AI-model-differentiation thesis. LAM — a Large Accounting Model trained exclusively on accounting data — is the first domain-specific model in the ERP category. It achieves over 95% accuracy on reconciliations and variance analysis, and Campfire's customers report five-times faster close cycles (144 days reclaimed annually). Customers include PostHog, Replit, Decagon. The architecture-native claim is the most credible in the category: LAM acts on records directly, not in response to queries.

a16z and ICONIQ backed Rillet's vertical-precision thesis. Purpose-built for SaaS companies: native ARR, MRR, and NRR calculated directly from the general ledger — not configured dashboards, not a BI layer. Two hundred native integrations to the standard SaaS finance stack (Stripe, Ramp, Brex, Rippling, Salesforce, HubSpot). Aura AI handles flux analysis and accruals automatically. Two hundred customers, including Windsurf and Bitwarden; ARR doubled in 12 weeks after the Series B. At approximately $500M valuation, it is the most expensive bet in the category, which says something about conviction.

On current evidence: Campfire on AI architecture depth, Rillet on SaaS fit and execution velocity, DualEntry on breadth for non-SaaS multi-entity implementations. The entrants are genuinely different products, not three versions of "better NetSuite."

What this means for your evaluation

If you are a CFO or CTO evaluating whether the window applies to you: the practical question is the rote-work fraction of your last close. Map the actual time allocation: how much was categorization, matching, reconciliation, and variance formatting versus judgment calls, stakeholder communication, and review? If rote work is above 60% of your close cycle, the AI-native platforms are removing a material operational cost. If your close is already judgment-heavy, you are buying a cleaner interface for work that is already at the high-value end. The case is different.

Three questions that shorten the shortlist: Is your rote-work fraction above 60%? Are you a SaaS company with a standard billing stack (Stripe, Ramp, Salesforce, Rippling)? Do you have multi-entity consolidation requirements? If the first two are yes, Rillet and Campfire are the conversation — Rillet for SaaS-first precision, Campfire for automation depth. If the first and third are yes, DualEntry's multi-entity architecture is the relevant comparison. If your close is already judgment-heavy (rote below 40%), the entrants still improve the interface but the migration economics are less compelling.

The evaluation that matters is not the vendor demo. It is a three-day parallel pilot against your last real month-end close, with your actual transaction history in the target platform. That is the test that separates what a vendor claims from what the automation actually delivers against your specific workload. The mechanism is real; whether your workload sits inside the window is a data question, not a pitch question.