AI-Native ERP [What It Changes]

TL;DR

• AI-Native ERP is accounting software whose general ledger was built so AI agents can post, reconcile, and explain entries on live books, instead of bolting a chatbot onto NetSuite, Oracle, SAP, or Intuit.

• Rillet is the current proof point: a $100 million Series C at a $1 billion valuation, closed in under 48 hours in August 2026, bringing total capital above $200 million and the customer count to more than 600.

• The product pitch is a perpetual ledger and a “zero-day close,” with Aura agents inside the GL, human-readable audit of every agent decision, and model routing so OpenAI or Anthropic does not train on the tenant.

• A 2-truck HVAC shop, a 10-person marketing agency, and a solo CPA firm should care because the money is betting that month-end grunt work is software now, and that replacing the ledger is no longer a 12-month program.

Key Takeaways

• The minted term is architectural: agents sit in the ledger, not next to it.

• Rillet’s round was Iconiq-led, with Sequoia and Andreessen Horowitz returning; the company says it was not trying to raise.

• Implementation is claimed at 4–6 weeks versus 3–12 months for traditional ERP, from founder comments in 2025 and the Series B post.

• Public-company rules, as Kopp described them, still require a human to approve every AI-agent transaction. That is the honest limit on “autonomous close.”

• Practice-management and tax stacks (Drake versus ProConnect, Fathom versus Jirav) do not disappear; they sit beside a new ledger, not inside it.

What AI-Native ERP Is

AI-Native ERP is an enterprise resource planning system whose general ledger, close, and reporting were designed so specialized accounting agents can act on live entries with a full audit trail, rather than storing transactions in a “dumb database” and doing the real work in spreadsheets.

A 2-truck HVAC shop that still exports QuickBooks to Excel every Friday, a 10-person agency that waits two weeks for a controller to say whether a retainer was recognized, and a solo clinic that closes the books after the 15th are all living in the old architecture. The constraint that broke is not “AI can write a journal entry.” It is that a well-funded vendor is now selling a replacement ledger, not a copilot, and venture capital just priced that replacement at a billion dollars.

Rillet’s own homepage calls itself “The AI-Native ERP” and sells a perpetual general ledger, Aura agents, close management branded “Zero-Day Close,” ASC 606 analysis from contracts, and an MCP connector so Aura can talk to Salesforce, Snowflake, or Claude on live books.

What Happened in August 2026

According to TechCrunch, Rillet raised $100 million at a $1 billion valuation in 48 hours, without a planned process, after a board meeting showed annualized revenue had doubled in the last quarter. Total capital is above $200 million. Customers number 600. Iconiq led; Sequoia and Andreessen Horowitz returned. Seth Pierrepont of Iconiq joined the board.

The 19 August 2026 TechCrunch brief adds that the company emerged from stealth in 2024, that Kopp said on X the round came together in less than 48 hours, and that a prior $70 million Series B was co-led by Iconiq and a16z after a Sequoia-led $25 million Series A in May 2025 (plus a $13.5 million seed and pre-seed from First Round, Creandum, and Susa).

ICONIQ’s own post dated 18 August 2026 says Rillet now serves more than 600 companies across 60 countries, names Neuralink, Mercor, Sotheby’s, LangChain, and Skild AI, and states a global alliance with EY, KPMG, and half of the top 30 U.S. accounting firms.

An EY newsroom note dated 29 April 2026 described Rillet (Gnau, Inc.) as supporting over 400 customers at that time, with a 5.0-star G2 rating in EY’s “About Rillet” blurb. The August 600-customer figure is the later count.

How the Mechanism Works in Plain Language

Legacy ERP stores the journal. Humans, or a bot sitting on top, do the thinking. Rillet’s Series B post calls those systems dumb databases: transactions in the ledger, real work in Excel and bolt-ons. An AI-native ledger inverts that. Bank, bill-pay, payroll, and CRM feeds land in a structured GL. Agents propose accruals, flux, collections, and close tasks on that live book. A governance layer, shipped about three months before the August interview, compresses agent steps into a human-readable audit of which numbers were pulled and how they were calculated.

Kopp told TechCrunch that customers can route models (OpenAI, Anthropic) and that the harness blocks training on tenant data, with no cross-customer training. Agents have memory of prior actions. For public companies, he said current rules still require a human to approve every transaction an agent makes.

The homepage splits Aura into Assistant (plain-English questions with accounting logic first), Agents (scheduled workflows written in English), Continuous AI (anomaly flags and proposed accruals), and Aura MCP (live books in the tools you already use). Close management is labeled zero-day. Revenue recognition is sold as ASC 606 from contracts. Integrations shown include Stripe, Ramp, Brex, Rippling, ADP, Bill.com, HubSpot, and Salesforce.

Why This Landed Now

Two labor facts sit under the round. According to the U.S. Bureau of Labor Statistics, employment of accountants and auditors is projected to grow 5 percent from 2025 to 2035, with about 1.6 million jobs in 2025, 79,400 added by 2035, 115,300 openings a year, and a median wage of $83,680 in May 2025. BLS also says automation of routine tasks such as data entry is expected to make advisory work more prominent, not wipe out demand.

According to Kent State University’s shortage briefing, accounting graduates have decreased 20 percent since 2010, citing AICPA, while total U.S. higher-ed enrollment was 18.6 million in spring 2025. That is the pipeline Rillet and every close-automation vendor are selling into.

The Series A story from May 2025 already had the implementation wedge: Kopp said Rillet can cut a many-month ledger conversion to about four to six weeks, with the old GL running until the new one is trusted. The Series B post repeats 4 weeks versus 12 months for traditional systems, and names Postscript closing in three days and Windsurf running finance with two people. Treat those as named-customer anecdotes, not a survey.

Kopp’s mix, in the August TechCrunch interview: 50% of customers come from Intuit, 30% from NetSuite and Sage Intacct, and 20% from Oracle, SAP, Workday, and Microsoft. That is a founder-reported mix, not a 10-K.

Capital and Customer Timeline

Sources: TechCrunch Series A; Rillet Series B post; TechCrunch Series C.

Sources: TechCrunch; ICONIQ; EY; BLS; Rillet Series B.

USTA analysis: $208.5 million on the tape, 4–6 weeks versus 3–12 months

USTA analysis. Add only the four disclosed rounds: $13.5 million seed + $25 million A + $70 million B + $100 million C = $208.5 million. That is consistent with “more than $200 million” in the August coverage. It is not a post-money cap table; option pools and unannounced checks are out of scope.

Second derivation, using only the implementation ranges already cited: Rillet 4–6 weeks (Kopp, May 2025; Series B FAQ 4–6 weeks) versus traditional 3–12 months (Series B post). Convert to weeks: Rillet midpoint 5 weeks; traditional midpoint 7.5 months × 4.3 weeks ≈ 32 weeks. 32 ÷ 5 ≈ 6.4. If both ends of the vendor’s own ranges hold, the claimed implementation is on the order of six times shorter than the traditional midpoint. That is a ratio of two marketing ranges, not a time-and-motion study. Keep it labeled as such.

What a Firm or Small Finance Team Should Actually Do

Do not rip out tax software because a unicorn raised. Best accounting practice-management tools still run the shop. Do ask three questions of any “AI ledger”:

1. Does the agent post to the book of record, or only draft a CSV?

2. Can a reviewer see the source document, the rule, and the posting in one pane?

3. For anything that will be in an audit file, is a human still the approver?

If the answers are yes, yes, and yes, you are looking at AI-native ERP as Rillet uses the phrase. If the agent lives in a browser sidebar on top of NetSuite, you are looking at a copilot.

US Tech Automations can sit on the intake side: bank files, bills, and CRM exports land in a structured handoff, and a confirm step sits in front of any posting an agent proposes. That is a workflow graph, not a new GL. Firms already using form-to-CRM automation or executive-assistant automations should not dump unstructured email into an agent that can book revenue.

Workday Financial Management remains the enterprise incumbent path; Rillet’s 20% bucket is the claim that some of those logos are already leaving. Treat that as founder-reported until a 10-K exists.

Kopp also said clients range from laundromats to a major sports franchise. That is color, not a segment mix you can budget against.

Honest Limits

VentureBeat’s matching story returned a checkpoint, so this hub does not rely on it. BusinessWire’s Series C URL denied access. Oracle’s ERP page and NetSuite’s homepage failed in this pass. EY’s April blurb said 400+ customers; August coverage says 600 — both can be true on different dates. The 50/30/20 incumbent mix is Kopp, not a census. Postscript’s three-day close and Windsurf’s two-person finance team are named anecdotes. “Zero-day close” is a product label, not a GAAP status. Public-company human-approval of agent postings, as Kopp described it, is the ceiling on autonomy. Rillet is not a 10-K filer; none of these operating figures are audited in the sources above.

The state of small-business automation still includes shops that will keep QuickBooks until a CPA firm they trust runs Rillet for them.

Signal vs Speculation

Demonstrated fact (sourced). In August 2026 Rillet closed a $100 million Series C at a $1 billion valuation in under 48 hours, with Iconiq leading and Sequoia and a16z returning, taking disclosed rounds to about $208.5 million and customers to 600 across 60 countries per ICONIQ. The product is sold as an AI-native ERP with agents in a real-time GL. EY announced an alliance in April 2026. BLS still projects accountant employment up 5% through 2035. Kent State, citing AICPA, still records a 20% drop in accounting graduates since 2010. Implementation is claimed at 4–6 weeks. Public-company agent postings still need a human, per Kopp.

Our read (12–36 months, SMB and firms). If the 4–6 week implementation holds for companies that are not already Rillet-shaped SaaS, mid-market controllers will start putting “replace NetSuite” on the 2027 budget instead of “add a close bot.” The operators who get hurt are write-up shops whose hours are month-end tickets. The operators who benefit are firms that sell review and judgment, not keying. Do not expect a two-truck HVAC company to buy Rillet this year; expect their CPA firm to. US Tech Automations stays useful as the pipe that gets clean source files into whichever ledger wins.

Our read, continued. In 12 months, “AI-native ERP” will be a procurement checkbox the way “cloud GL” was after 2015. In 36 months, the interesting fight is whether Oracle, Sage, and Intuit ship a true in-ledger agent or keep wrapping the old database. The 48-hour unicorn is a pricing signal, not proof that close is solved.

Glossary

• AI-Native ERP. An ERP whose ledger and close were built for agents on live data, not a chatbot bolted onto a batch GL.

• Perpetual / real-time general ledger. Books that stay current as feeds land, rather than waiting for a period close to become true.

• Zero-day close. Rillet’s product label for collapsing month-end into continuous posting plus review.

• Aura. Rillet’s agent suite (Assistant, Agents, Continuous AI, MCP).

• ASC 606. FASB revenue-from-contracts standard; Rillet sells automated analysis from contracts.

• Model routing. Sending a prompt to a chosen foundation model without letting that model train on the tenant.

• Human-in-the-loop posting. Kopp’s description of public-company rules: every agent transaction still needs a person.

• Incumbent mix. Kopp’s 50/30/20 split of customers coming from Intuit / NetSuite+Sage / Oracle-SAP-Workday-Microsoft.

Frequently asked questions

Is AI-Native ERP just a chatbot on NetSuite?

No. As Rillet and ICONIQ use the phrase, the agents post and reconcile inside a ledger built for them. A sidebar that drafts a journal for a human to paste into NetSuite is a copilot, not this architecture.

Did Rillet really raise in 48 hours?

That is how Kopp and TechCrunch described the Series C process in August 2026. Iconiq said it was not a cold start: they had co-led the B and watched a year of delivery.

How many customers does Rillet have?

EY’s April 2026 blurb said over 400. TechCrunch and ICONIQ in August 2026 said 600 or more than 600, in 60 countries. Use the dated figure.

How long does a cutover take?

Kopp said about four to six weeks in May 2025, with the old GL running in parallel. The Series B post contrasts that with 3–12 months for traditional ERP. Those are vendor ranges, not a third-party benchmark.

Will this replace staff accountants?

BLS still projects 5% job growth for accountants and auditors through 2035 and says automation should shift work toward advisory tasks. Kopp said he does not see people losing jobs soon. Neither source is a guarantee for a given firm’s headcount.

If you already move bills, bank files, or close checklists through US Tech Automations, keep the confirm step in front of any agent posting, and treat Rillet or a rival ledger as a system of record swap rather than a rebuild of intake. You can map that split on the agentic-workflows canvas.