---
title: "Nace for professional firms"
description: "Nace for audit and accounting firms: an AI Senior Accountant that performs bookkeeping, control and reports, end to end, with every control run on every entry. Models the firm owns, inside the firm's tenant, running in 24 hours."
canonical: "https://nace.ai/business/professional-firms"
---

# Nace for professional firms

> Nace for audit and accounting firms: an AI Senior Accountant that performs bookkeeping, control and reports, end to end, with every control run on every entry. Models the firm owns, inside the firm's tenant, running in 24 hours.

## Put a senior accountant on every client.

Nace for professional firms

An AI Senior Accountant that performs bookkeeping, controls and reports end to end, on every entry instead of a sample. Small models the firm owns, inside the firm's tenant.

Partners sign. Nace performs the procedures and shows its evidence.

- Arrange a briefing: [contact form](https://nace.ai/business/professional-firms#contact)
- Meet the AI Senior Accountant: [the role](https://nace.ai/business/professional-firms#roles)

### One engagement's population (illustrative)

Does this entry need a reviewer?

| Review threshold | Cleared, with the source attached | Routed to a reviewer |
| --- | ---: | ---: |
| 0.30 | 111 | 49 |
| 0.50 | 126 | 34 |
| 0.70 | 134 | 26 |
| 0.90 | 146 | 14 |

Every entry gets a probability. You set the line: what falls under it clears with its source attached, and the rest goes to a reviewer. Each square is one entry in a 160-entry population, drawn for the example.

### Key figures

- 100%: Of the population tested, instead of a sample
- 1 day: From a working session to the first loop running
- #1: On the Decision Index, across 38 public benchmarks
- 0: Client files kept after the engagement, or used to train anyone else's model

## Test the population, not a sample of it.

Sampling exists because nobody has the hours to look at everything. Nace puts one question to every entry and returns a probability, so your people read only what needs them.

|  | Sampling by hand (The hours left decide the sample) | General-purpose LLM (An opinion to check) | Nace (A probability on every entry) |
| --- | --- | --- | --- |
| What gets tested | A sample, sized to the hours left | Whatever fits in the prompt | Every entry in scope |
| What comes back | A tickmark and a note | A paragraph to read and interpret | A probability for each option you supplied |
| What a person reviews | Everything in the sample | Every answer, because any of them could be wrong | Only what falls past your threshold |
| Where client files go | Your file share | The vendor's service, on the vendor's terms | Your perimeter, up to air-gapped |

## Bookkeeping, control and reports, performed end to end.

Drex deciding and Nace Document Intelligence reading, wired into the work your staff already do, on small models that belong to the firm.

### The AI Senior Accountant

Performs bookkeeping, control and reports, end to end.

#### Bookkeeping

- Reads: bank feeds, invoices and receipts, the chart of accounts.
- Hands back: transactions coded and posted, uncertain lines queued for a person.

#### Controls

- Reads: entries as they post, the client's control set.
- Hands back: controls run on every entry, exceptions documented with evidence.

#### Reports

- Reads: the closed books, prior-period reports.
- Hands back: reconciliations, month-end reports drafted for sign-off.

#### Example call to Drex

```json
{
  "model": "drex-latest",
  "questions": {
    "account": {
      "criteria": {
        "marketing": "Marketing",
        "office": "Office supplies",
        "software": "Software subscriptions"
      },
      "instructions": "Which account does this bank line belong to?",
      "type": "choice"
    }
  },
  "state": "Bank line, 2 Oct: CARD 4417 DESIGNTOOLS SAAS 45.00 USD. Same vendor charged monthly since March, last coded to Software subscriptions."
}
```

probabilities.software = 0.97 (probability the line is software subscriptions). Above the threshold. Coded and posted with the bank line as evidence.

## 100% customizable to your methodology.

Your materiality, thresholds, request list and workpaper format, as rules your quality team can read.

- Written in plain language, approved by your engagement partners
- Set per engagement: materiality, thresholds and the request list
- Changes take effect without rebuilding anything
- Every rule traced to the calls it produced

### Review policy (Example, FY25 audits, approved by the engagement partner)

| Clause | Rule | Owner |
| --- | --- | --- |
| §1.2 | Entries scoring above 0.50 for review go to a reviewer with the entry and its source attached. | Engagement manager |
| §2.1 | Any entry above performance materiality is reviewed by a person, whatever its score. | Partner |
| §3.3 | Client files that match no request-list line go to the senior, never to the archive. | Senior |
| §4.1 | A figure clears only with its page, cell or span cited. No source, no tickmark. | Quality |
| §5.0 | Risk ratings are proposed by Nace and set only by the engagement partner. | Partner |

## We don't replace the engagement team. We staff alongside it.

Your engagement software, file share and channels stay where they are. Nace works one engagement beside your team first.

| Phase |  | Your team lane | Nace lane |
| --- | --- | ---: | ---: |
| 1. One engagement, two teams | We build the first loop on a closed engagement in one working session. It runs within 24 hours, and you compare its findings with the file. | 100% | 100% |
| 2. Nace performs, your seniors review | Repeating procedures move to Nace. Your people review what falls past the threshold, evidence attached, and sign. | 45% | 100% |
| 3. Your team leads, Nace performs | Nace carries the population work across the season. Your team keeps judgment, the client and the signature, and every review tunes Drex for your firm. | 10% | 100% |

- No new system. Your engagement software, file share and channels remain the record.
- No leap of faith. Results are proven on a closed engagement of yours first.
- No loss of the signature. Partners sign. Nace never issues a report.

## Client files stay inside your perimeter.

Pick the boundary your risk and independence teams sign off on. The models and the API stay the same.

| Boundary |  | What it means |
| --- | --- | --- |
| Managed | In our cloud | Our GPUs, behind our API, with strict tenant isolation. Live in days. |
| Dedicated | Single tenant, private links | Single-tenant compute in our cloud, over private links. |
| Your cloud | Your Azure or AWS account | Inside your Azure or AWS account, with your keys, logs and IAM. |
| Air-gapped | No network at all | Your datacenter and hardware, with no egress and no telemetry. |

- SOC 2 Type II: Type II audited, with the report on the trust center.
- No shared training: Client documents never train models shared with anyone else.
- Your models, your weights: Tuned weights, adapters and eval sets stay with the firm, inside its tenant.
- One API at every boundary: The same request returns the same shape at every boundary.

[Visit the trust center](https://trust.nace.ai)

## Questions firms ask first.

### Who signs off on the work?

No. Partners and managers keep professional judgment and the signature. The AI Senior Accountant performs the work and documents it; every call it makes carries a probability and the source span behind it, and anything past your threshold comes to a person with that evidence attached.

### Which parts of the work does the AI Senior Accountant take?

The work that repeats for every client: coding bank feeds, invoices and receipts to the chart of accounts, running the controls on every entry as it posts, reconciling and drafting the month-end reports. The judgment calls stay with your team.

### Is it accurate enough for client work?

Drex 1.5 is first on the public Decision Index across 38 public benchmarks, with under 10B parameters. Every answer is a probability, so you set the threshold: the confident calls clear themselves and the rest go to a reviewer. Tuned on your firm's own labelled decisions, it gets better at exactly the calls your reviewers make.

### Does it read the client's documents?

Yes. Nace Document Intelligence parses, extracts, splits, classifies and grounds any file type, from scanned PDFs and spreadsheets to email. Every value comes back tied to its page, cell or span, so a reviewer opens the evidence behind a call instead of searching for it.

### Do client files train your models?

Never into a shared model. When we tune Drex on your outcomes, the weights are trained for your firm and shipped to you. Nace is SOC 2 Type II audited; the report is on the trust center.

### Can it run inside our own environment?

Yes. Four boundaries: Managed, Dedicated, Your Cloud and Air-Gapped. Drex fits on a single accelerator, and every deployment answers the same request in the same shape.

### How do we start?

Bring one engagement to a working session and leave with the first loop running within 24 hours. Or start alone: create an API key in the Console at console.nace.ai and send your first decision in a minute.

## Let's talk about your season.

Leave your details and we'll arrange a private conversation. Bring one engagement in mind.

1. A short conversation: We learn which engagements eat the season and where.
2. A tailored view: Your kind of engagement, with a line through the population.
3. A side-by-side proof: Nace works one closed engagement beside your team, inside your perimeter.

[Arrange a briefing](https://nace.ai/contact?source=business-professional-firms)

## For agents

- [Drex API guide](https://nace.ai/drex.md): the decision model: request shape, question types, benchmarks and deployment, as markdown
- [Decision endpoint](https://console.nace.ai/v1/systemone): POST with model "drex-latest", a state string and typed questions (choice, noul, score); https://console.nace.ai/v1/models lists checkpoints
- [Document endpoints](https://console.nace.ai/v1/documents/{parse|split|classify|extract|ground}): POST a source (a public URL, an uploaded workspace file or a finished parse job) and poll GET https://console.nace.ai/v1/documents/jobs/{id}; the same key and credit as decisions
- [Console](https://console.nace.ai): API keys, the playground, usage and billing; send `Authorization: Bearer $NACE_API_KEY` on every request
- [Console docs](https://console.nace.ai/docs): quickstart, question types, the document routes, errors, limits and pricing
- [Nace Document Intelligence](https://nace.ai/ndi.md): the Perception series: parse, split, classify, extract and ground any file type with a location on every value
- [Developers](https://nace.ai/developers.md): Console: what it holds, a scripted agent session, the ways in and the four deployment boundaries, from our managed cloud to air-gapped
- [Site index](https://nace.ai/llms.txt): every page on nace.ai; append `.md` to any page URL or send `Accept: text/markdown` for the markdown version