Sub-processors and hosting.
For enterprise
Governed voice AI, scoped to your deployment.
Paid design-partner pilot at a fixed €15,000 · production: By application
Enterprise is the direct tier: a custom master agreement, a named contact, your own integrations, and commercial terms set against your real call volume rather than a published bracket: a paid design-partner pilot at a fixed €15,000, with production agreements by application. The compliance engine and the evidence chain are the same as in the agency tier. What differs is the buyer, the scope and the fact that the deployment is co-engineered rather than productised.
The obligations a regulated deployment runs into
- A person interacting with an AI system must be informed of that, unless it is obvious from the circumstances and the context of use.
- Regulation (EU) 2024/1689, Art. 50(1)
- Synthetic audio must be marked in a machine-readable format, detectable as artificially generated.
- Regulation (EU) 2024/1689, Art. 50(2)
- For SMEs and start-ups the penalty ceiling is the lower of €15 million or 3% of worldwide annual turnover, not the higher.
- Commission guidelines C(2026)5054 of 20 July 2026, point (152)
- Financial entities must manage ICT third-party risk, including contractual terms and exit strategies for critical providers.
- Regulation (EU) 2022/2554 (DORA), directly applicable since 17 January 2025
- A party that puts its own name, trade mark or other distinguishing feature on a product is treated as its manufacturer, including for software.
- Directive (EU) 2024/2853, Art. 8(1)(b), applying from 9 December 2026; transposition runs per member state
- Systems placed on the market before 2 August 2026 only have to satisfy Art. 50(2) from 2 December 2026.
- Digital Omnibus (EU) 2026/1744, in force 27 July 2026, inserting Art. 111(4)
What the platform contributes
- Every request routed through the governed workflow is evaluated against the active mandate before it is spoken: the disclosure, the approved wording, the prices and the commitments the agent is allowed to make.
- In controlled testing
- A receipt is visible in the product for governed interactions, carrying the input hash, the output hash, the link to the receipt before it and Merkle inclusion. The first three a third party can recompute today, without a key. The fourth, inclusion in the Merkle root, is checked against the published root with the inclusion proof from the call's Article 50 report.
- The governance gate is designed to sit in front of the voice stack you already run, so the proof arrives without a rip-and-replace. Control is the part being built; the design-partner pilot is where that placement is scoped with you.
- Control · in development
- The monthly evidence package is not in the offer; it is Under validation: we do not ship an evidence artefact we have not tested ourselves.
- Under validation
- Next to the receipts, a reviewer sees the disclosure events and the policy version recorded on each receipt.
- In controlled testing
- Policy differs per jurisdiction, because the national layer on top of Article 50 differs.
- Germany: national policy layer active. Netherlands: in development. Belgium and Austria: planned.
- Early Customer Program intake.
- Machine-readable marking of synthetic audio under Art. 50(2). Not in the product today.
- Planned
The scoping call
Sixty minutes, and what we go through in them.
- The data the deployment processes, and where it is processed. Primary application hosting is Frankfurt; the full subprocessor list is in the Trust Center.
- Subprocessors and retention, per data category rather than as one number.
- Role-based access and single sign-on against your identity provider.
- Fallback behaviour and the handoff to a human: when it fires, what the caller hears, what the record shows.
- The service level and the implementation timeline, against your volume rather than a generic tier.
- The exports your legal and procurement teams receive, and in which format. Production export worked when it was retested on 26 August 2026 after the deploy, measured on browser calls rather than over a telephone line: we scope what it does now, not what it is meant to do.
Where it differs
Enterprise against the agency tier.
- Who buys
- Agency Voice
- An agency reselling to end-clients
- Enterprise
- The organisation that runs the calls
- Rate
- Agency Voice
- €1,497 once, then €499 a month · founding price, first three external agencies
- Enterprise
- Paid design-partner pilot at a fixed €15,000; production agreements by application
- Contract
- Agency Voice
- Standard terms
- Enterprise
- Custom master agreement
- Brand in the call
- Agency Voice
- The end-client's, white-labelled
- Enterprise
- Yours
- Integrations
- Agency Voice
- The productised set
- Enterprise
- Built against your systems
- Compliance engine and evidence chain
- Agency Voice
- The same
- Enterprise
- The same
Alongside, not instead
The design does not require you to remove anything.
The gate is designed to sit in front of the voice stack you already run, so the proof arrives without a rip-and-replace. Control · in development
Backed by a patent-pending governance stack filed at the UK IPO.
Questions
- How is this different from the agency tier?
- The agency tier is productised: €1,497 for the implementation and €499 a month — the founding price for the first three external agencies, €2,995 and €749 after that — with the agency reselling under its own brand. Enterprise is direct: a custom master agreement, a named contact, your own integrations, a paid design-partner pilot at a fixed €15,000 and production agreements by application. The compliance engine and the evidence chain are identical; the buyer, the scope and the contract are not.
- Why is this tier gated rather than listed?
- Because it is scoped, not configured. The agency tier buys a product; enterprise buys a co-engineered deployment, and we would rather map it in an hour than sell you a size that does not fit.
- Can we migrate from a platform we already run?
- Yes, and it is part of scoping rather than a separate project: prompt conversion, webhook mapping and number porting. Webhook mapping is part of the migration; a migration engineer confirms the event mapping before go-live. Number porting depends on where the numbers are registered and who holds the account.
- What can a reviewer actually check?
- On any receipt, three fields can be recomputed by anyone today: the input hash, the output hash and the link to the receipt before it. No key, no account, no access to our systems. A fourth, inclusion in the Merkle root, is checked against the published root with the inclusion proof from the call's Article 50 report. Alongside them the record carries the disclosure events and the policy version recorded on each receipt.
Scope it before you sign it
A sixty-minute call, your real volume, and a quote against that rather than a bracket. Or write to enterprise@audact.ai.
Related
Trust Center Security Architecture Where the gate sits in the request path.
Governance Gate API DORA Patents