AI-Native ERP in 2026: A Buyer's Guide to Rillet, Campfire, DualEntry, and Light

Venture investors put $315 million into the four AI-native ERP vendors in this guide during 2025 alone. That money bought a wave of comparison pages, launch posts, and alternatives listicles, nearly all of it written by the vendors, their investors, or directories paid per referral. Somebody should rank these systems without holding a position. That is what this guide does: Rillet, Campfire, DualEntry, and Light, scored on the same twelve-domain methodology we apply to NetSuite and Sage Intacct, with every number labeled by who is making the claim.

Citable stat

The four AI-native ERP vendors raised a combined $315 million in 2025: Rillet closed a $25 million Series A and a $70 million Series B, Campfire a $35 million Series A and a $65 million Series B, DualEntry a $90 million Series A, and Light a $30 million Series A.

The ranking is built for the buyer these products actually target: a US software or digital-services company between $5M and $200M in revenue, a lean finance team, rev-rec or consolidation pain. If that is not your company, skip ahead to the last two sections. They matter more than the order.

What does AI-native ERP actually mean?

An AI-native ERP is a general ledger built from scratch since 2021 in which AI agents do accounting work inside the system of record: drafting journal entries, matching bank transactions, proposing accruals, all under a propose-and-approve workflow where a human signs off before anything posts and every AI action carries an audit trail. Three of the four systems here document that approval gate explicitly; AI features bolted onto a suite designed twenty years ago do not qualify.

The exception is Light: it documents an immutable ledger and AP approval workflows, but no published propose-and-approve step on its agents specifically. That gap writes your demo script. Ask each vendor to show an agent preparing an accrual, the human approval step, then the log of what the AI touched. A chatbot summarizing reports next to a ledger humans still post to by hand is AI-flavored, not AI-native.

Two more things the label does not mean. None of these products is an operational ERP: no manufacturing anywhere in the category, no warehouse operations, order management nonexistent to lightly claimed. These are finance systems. And AI-native does not mean self-service; all four vendors deliver implementation themselves, because none has a partner channel yet.

Citable stat

A vendor-led AI-native ERP implementation runs roughly 4 to 8 weeks as of mid-2026 (Rillet claims 4 to 6 weeks, DualEntry claims 4 to 8 with implementation included free), versus 3 to 6 months for a typical single-entity NetSuite project.

How this ranking works

We ranked on evidence quality more than feature grids. The top three sit close together on our twelve-domain scores, so what separates them is how much of each vendor's story checks out against independent sources: marketplace pricing data, named references at real scale, a verified customer instead of a claimed one. All four vendors are young; the honest frame is a ranking of the least unproven.

Citable stat

Customer evidence as of mid-2026: Rillet's site cites more than 500 customers, Campfire had about 100 per TechCrunch at its June 2025 Series A, DualEntry had 42 as of July 2025 per its investor Contrary Research, and Light reported $1.2 million in ARR within six months of commercial launch.

1. Rillet: the deepest evidence, the narrowest lane

• ▪Funding: just over $100 million total. A $25M Sequoia-led Series A in May 2025, then a $70M Series B co-led by Andreessen Horowitz and ICONIQ roughly twelve weeks later, at a reported valuation near $500 million.
• ▪Customers: 200+ reported at the August 2025 Series B; the company's site claimed more than 500 by mid-2026. Both figures are vendor statements, but named references exist at real scale (Windsurf and Postscript, each operating around or above $100M ARR).
• ▪Pricing posture: quote-based, and the only one in the category with independent benchmarks. Vendr transaction data shows a median contract of $28,250 per year with observed deals from roughly $20,100 to $34,800; our pricing research puts year one all-in around $35K to $60K for a three-entity SaaS company.
• ▪The one thing it does best: ASC 606 revenue recognition generated directly from CRM and billing data (Salesforce, HubSpot, Stripe, Chargebee). We score revenue and billing 5 of 5, the only revenue score of 5 in this cohort.
• ▪The one reason not to buy: the lane. No inventory, no project accounting, a closed platform, thin localization. If services are a material revenue line or you ever touch physical product, the fit erodes fast.

Rillet ranks first because more of its story survives contact with independent sources than anyone else's: the category's only marketplace pricing data, the longest customer list, named references at meaningful ARR. Our Rillet profile scores all twelve domains, the Rillet pricing page covers negotiation levers including the still-active Mercury perk, and the full Rillet review draws the fit lines in detail. Cross-shopping the incumbent? See NetSuite vs Rillet.

2. Campfire: the challenger with the boldest AI claims

• ▪Funding: roughly $103.5 million total since its 2023 founding. A $35M Accel-led Series A in June 2025, then a $65M Series B co-led by Accel and Ribbit in October 2025, announced with reported 10x year-to-date revenue growth.
• ▪Customers: about 100 reported by TechCrunch at the Series A, including one on track for roughly $250M ARR. Named logos include Replit, PostHog, and Decagon; the NYSE-listed users in its materials are vendor-cited.
• ▪Pricing posture: undisclosed. No list prices, no Vendr data, no practitioner-reported deals exist. Category peers land roughly $20K to $40K per year; our year-one estimate is $30K to $55K all-in, held at low confidence.
• ▪The one thing it does best: billing-model breadth. Subscription, usage, milestone, and transactional billing live in the same system as the ledger, and the Ember assistant lets non-finance teammates query financials directly. We score revenue and billing 4 of 5.
• ▪The one reason not to buy: the proof gap on its headline claims. LAM, the vendor's accounting model, claims 95%+ accuracy with no independent validation, and headcount quadrupled from about 10 to about 40 people between June and October 2025. A team that small on your system of record cuts both ways.

Campfire is the most direct head-to-head with Rillet, and for companies with usage or milestone billing it can be the better-shaped product. What it lacks is Rillet's depth of verification: fewer named customers at scale, no marketplace pricing data, and a flagship AI story graded by the vendor itself. The Campfire profile has domain-by-domain scores, our Campfire pricing research documents exactly which cost claims have evidence, the Campfire review goes deeper, and Campfire vs Rillet settles the head-to-head directly.

3. DualEntry: the widest module list, the thinnest verification

• ▪Funding: past $100 million company-reported, almost all of it from a $90M Series A co-led by Lightspeed and Khosla with GV participating, announced October 2025 at a reported $415 million valuation as the company left stealth.
• ▪Customers: 42 as of July 2025, per Contrary Research, which is an investor. Claims of NYSE-listed users and $100 billion in journal entries processed are self-reported.
• ▪Pricing posture: three published tiers gated by entity count (3, 20, unlimited) with unlimited users, but no dollar figures anywhere. Third-party estimates run $24K to $180K per year depending on modules. Implementation is included free on every plan, the category's clearest cost claim, vendor-published. Our planning anchor is $30K to $80K all-in.
• ▪The one thing it does best: surface area. GL, AP with OCR capture, order and purchase order management, fixed assets, multi-book, budgeting, and a light inventory module. It is the only AI-native vendor we score above 1 on inventory, at 2 of 5.
• ▪The one reason not to buy: verification. Founded in June 2024, it is the youngest vendor we score, and nearly every impressive statistic in circulation (an 80% win rate, 24-hour migrations, the $100B processed figure) originates with the vendor or its investors.

DualEntry's pitch lands hardest with multi-entity companies burned by an implementation quote; free implementation and entity-based tiers attack exactly that pain. The gap between claimed capability and independent evidence is the widest on our site, which is why it sits third despite the biggest feature grid. Scores and caveats are on the DualEntry profile, cost detail on the DualEntry pricing page, and the DualEntry review works through what is proven versus claimed, module by module.

4. Light: the Europe-first outlier

• ▪Funding: $43 million total, the smallest war chest in the category. A $30M Series A led by Balderton in September 2025, on top of a $13M seed, with a New York office opening after the round.
• ▪Customers: European scale-ups (Lovable, Sana, KeyShot appear in funding coverage and case studies). The vendor reported $1.2M ARR within six months of commercial launch and 30x growth in the preceding year, both self-published figures from a small base.
• ▪Pricing posture: a genuine blank. No pricing page, no tiers, no third-party benchmarks of any kind. Our profile renders every cost anchor as undisclosed rather than inventing a number.
• ▪The one thing it does best: European jurisdictional complexity. Per-entity tax codes, Peppol e-invoicing, embedded spend management with virtual cards, and a command interface that runs from Slack or Teams. US-centric rivals mostly lack all four.
• ▪The one reason not to buy (for a US buyer): US readiness. No documented rev-rec engine, no public SOC attestation details as of mid-2026, minimal US customer evidence, and the least funding in a category where consolidation over the next few years is a live scenario.

Light ranks fourth for a US audience and would rank differently in Europe. A Copenhagen-built ledger that treats multi-jurisdiction VAT and e-invoicing mandates as day-one requirements is solving a problem Rillet and Campfire have barely started on. For a US SaaS company with rev-rec pain it is the wrong tool today, and for a multi-country scale-up it belongs on the demo list. Details on the Light profile and the Light pricing page, which mostly documents how little exists.

What has none of them proven yet?

Three things, as of July 2026: an IPO audit cycle, consolidation at incumbent scale, and inventory. No AI-native vendor has published a named public-company reference that completed an IPO on its ledger. In our July 2026 profile reviews we found no published customer reference above roughly 50 entities, against NetSuite OneWorld's documented capacity of 250 subsidiaries. And none has real inventory operations; DualEntry's light module is the category's high-water mark at 2 of 5.

Rillet's about page claims public companies with over $1B in ARR as customers; as of our July 2026 profile review, no named reference supported it. Campfire's NYSE-listed users appear in vendor materials (LimaOne, an affiliate of NYSE-listed MFA Financial, is the most concrete). DualEntry's public-company claims are self-reported. Big 4 audit familiarity, SOX ITGC maturity at scale, and a completed IPO cycle on any of these ledgers remain publicly unevidenced. If you are within two years of an IPO, that absence is your answer.

Citable stat

NetSuite OneWorld supports up to 250 subsidiaries in a single account per Oracle's documentation; the largest multi-entity reference we found in our July 2026 reviews of the AI-native vendors is a DualEntry case study citing a customer with 50-plus entities.

Multi-entity consolidation is where the category genuinely helps, and where the ceiling is unmeasured. Every vendor here does intercompany eliminations and multi-currency translation natively, the exact pain that pushes companies off QuickBooks. But the published evidence tops out at vendor case studies in the tens of entities, with localization (statutory formats, e-invoicing outside Light, local GAAP books) thin across the board. A holding structure with dozens of subsidiaries across many tax regimes is still OneWorld or Intacct territory.

Inventory is simpler: it does not exist here. Rillet, Campfire, and Light score 1 of 5 because there is nothing to score. Campfire's answer is a February 2026 partnership with DOSS, a two-vendor stack with an integration seam rather than a module. DualEntry lists inventory tracking and demand forecasting, with no public evidence on costing methods, lot and serial tracking, or warehouse workflows. Physical product means a different category of system.

Three quieter gaps run across the cohort. No proven FP&A layer: Rillet, Campfire, and Light have none, and DualEntry's top-tier budgeting module has almost no independent evidence behind it, so assume planning lives in another tool. No renewal track record, because the oldest commercial deployments are only a few years into their contracts; no data is not a clean record. And no independent partner or admin talent market, so every implementation, integration fix, and training session runs through the vendor that sold you the software.

Who should skip this category entirely

• ▪Manufacturers and distributors. No BOMs, no work orders, no warehouse operations exist anywhere in the cohort, and DualEntry's module is light inventory accounting, not operations.
• ▪Services-heavy firms that need project accounting, resource management, or timesheet billing. The best projects score among the four is 2 of 5.
• ▪IPO-imminent, PE-owned, or regulated buyers who need an attested track record and an audit ecosystem that already knows the platform.
• ▪Companies whose current ERP earns its keep through customization. All four are closed platforms we score 2 of 5; you configure within the vendor's rails or you wait for the roadmap.
• ▪Single-entity startups without rev-rec or consolidation pain. QuickBooks or Xero remain cheaper and entirely sufficient until that pain is real.

The bottom line

The AI-native category is real and heavily funded, and for one specific company shape its products are already the right answer: pure software, multi-entity or rev-rec pain, a lean team, no IPO on an eighteen-month horizon. Within that shape, Rillet carries the most independent evidence and Campfire runs closest behind. DualEntry promises the most and has proven the least of the top three. Light is a different bet for a different continent.

Practical advice from the deals we have watched: shortlist two, never one. These vendors discount against each other, and quotes flex accordingly. Demo against your own contracts and your own close checklist rather than the vendor's demo data, get data-export terms and a renewal cap in the order form, and involve your auditors before go-live, not after. Before taking a sales call, the ten-minute assessment scores all sixteen systems in our catalog, these four included, against your actual footprint.