Most shortlists for an AI-native ERP start as a Google search for "Rillet competitors" and end as a slide the controller has to defend to the audit committee. The names are not in dispute: Rillet, Campfire, DualEntry, Lensing (formerly Everest), and the operations-first platforms that pair with them, like Doss, show up on every analyst list. What is in dispute is the scoring frame. "AI-native" is a marketing word that covers everything from a chat box over a legacy GL to an agent that drafts, codes, and posts journal entries while a controller watches.
This piece gives finance-systems leads a frame that survives a procurement review: five axes a controller can score, applied to the realistic Rillet shortlist as it looks at the end of 2026.
What Rillet Actually Is, and Who It Loses To
Rillet is not a full ERP. It is an AI-native general ledger aimed at mid-market SaaS and subscription businesses, with ASC 606 revenue recognition, bank reconciliation, multi-entity consolidation, and an AI layer branded Aura. It has no inventory, no manufacturing, and no CRM. It is also the best-funded entrant in the category, with over $100M raised and more than 200 customers at the time of its Series B.
The companies Rillet loses deals to fall into three groups. The first is other AI-native finance ledgers: Campfire and DualEntry, both chasing the same SaaS controller. The second is the AI-enhanced incumbents: NetSuite, Sage Intacct, and Microsoft Dynamics 365 Finance, which carry breadth Rillet does not pretend to have. The third is the operations-native platforms, led by Doss, which do not compete at all and instead sit next to Rillet to cover inventory and procurement.
The Five Axes That Separate Them
Vendor demos are optimised to show the AI drafting a journal entry. That is the easy part. The hard part is what happens next. The scoring frame below asks the questions a controller will have to answer at go-live, not at demo day.
Agent Scope and Ledger Write-Path
The question is not whether the vendor has an agent. The question is where the agent is allowed to write. In a serious design the agent proposes a journal entry, a validation layer checks it against policy, and only then does the entry hit the ledger. The agent itself never writes to the general ledger directly. This is the architectural choice that decides whether "AI-native" means audit-ready or means a probabilistic model editing your trial balance. The pattern is covered in detail in what an agent structurally cannot do inside a ledger, and it is the first question to put to every vendor on the shortlist.
Reconciliation Evidence Produced Daily
A close is fast because reconciliation is continuous, not because the controller works harder in the last week of the month. Closing faster is mostly a reconciliation problem, and the vendor question is whether the system emits a daily reconciliation artefact a reader can actually open: bank-to-ledger ties, intercompany balance proofs, revenue-schedule reconciliations against billed amounts. Rillet, Campfire, and DualEntry all claim some version of this. Only one of them will show you the artefact in a sandbox.
Migration Terms, in Writing
This is where the shortlist thins fastest. DualEntry publishes a 24-hour NextDay Migration with zero implementation fees and a free sandbox against real data before contract. Rillet does not publish implementation fees, with third-party estimates in the mid-four to low-five figures and a 4–6 week go-live. By contrast, a standard NetSuite mid-market implementation runs 4–6 months through six phases. Migration speed is not a vanity metric. It decides whether the first close on the new system happens in the fiscal quarter the business case was built around.
Close Cadence and the Rollback Condition
The marketing claim is a three-day close. The procurement question is what happens on day four of a close that is not going well. A credible vendor has a documented rollback condition and a parallel-run period against the legacy GL. One worth adopting is in the piece on cutover rollback conditions on this blog. If the vendor cannot describe the rollback plan without consulting its CSM, that is the answer.
Governance and Chart of Accounts Authority
Who owns the chart of accounts after go-live is a governance question, not a configuration question. An AI-native platform that mutates account structures to fit its templates will create reconciliation debt by month three. The default should be that the controller owns the chart and the agent is scoped to the sub-ledgers. A chart of accounts migration is not a copy job, and the vendor's answer to who gets write access to it is a signal about everything else.
The Shortlist, Scored
Applying the frame to the realistic 2026 shortlist produces four coherent choices, not ten.
Rillet. The strongest match for a Series B through pre-IPO SaaS business where revenue recognition is the hardest problem. ASC 606 depth is real. The gap is operations: no inventory, no manufacturing, limited UK and European compliance. Independent benchmark data puts the median annual contract at around $28,300 with most deals between $20,100 and $34,800, per third-party buyer data.
Campfire. Closest head-to-head alternative, same ICP, similar AI pitch, with explicit support for 180+ currencies and real-time consolidation. Independent reviews place the sweet spot at roughly $20M–$300M ARR. The frame question that separates it from Rillet is which platform's revenue-recognition engine handles the specific billing model: usage-based pricing, hybrid contracts, multi-element arrangements.
DualEntry. The migration-led choice. Different thesis: switching costs, not product quality, are the real lock-in. If the blocker to leaving NetSuite is the implementation estimate, DualEntry's 24-hour migration and $0 implementation fee are a serious answer. The scope gap is the same as the others: no inventory, no manufacturing.
NetSuite with an AI operating layer. The honest option when the business already runs on NetSuite, has multi-entity complexity, and cannot absorb a replatform. Shipping Intelligent Close Manager, AI bank matching, and EPM agents across its 2026 releases does not make NetSuite AI-native, but it does narrow the gap for finance teams whose real blocker is manual AP or close drag. This is the "wrap, don't replace" path, covered in replace the ERP or wrap it with agents first.
Where erp.io Sits in This Frame
erp.io is built around the write-path discipline this frame scores for: role-based agents (AP, AR, Controller, CFO) propose entries, a deterministic ledger validates them, and the AI never writes to the ledger directly. Daily reconciliation proof is the output, not the aspiration. Migration, integration, and implementation are included rather than quoted. The architectural reasoning behind that choice sits in AI in ERP, and the same scoring frame can be run against erp.io in the ERP software comparisons page.
How to Run the Shortlist Without Falling for the Demo
A demo will not answer any of the five questions above. A two-week pilot against the shortlist's own sandbox will answer four of them. The frame to run the pilot inside, including the specific artefacts to require, is in how to run a two-week AI ERP pilot, and the contract clauses that make the artefacts enforceable are in what to put in the AI ERP contract before you sign.
What to Do Before the Next Vendor Call
A defensible shortlist decision is not a vendor-feature matrix. It is a scored answer to five questions a controller already knows how to ask. Write the frame down before the next demo. Mark where each vendor refuses to answer in writing. The refusals are the shortlist. For finance teams actively running this exercise, a free assessment will apply the same frame to the current ERP footprint.