# Precision AI OS — AI implementation firm building data pipelines, a private firm OS and task agents for service firms.

> Source: The Agents Index — https://theagentsindex.com/precision-ai-os (structured, researched, re-verified)
> Facts last verified: 2026-10-06

Precision AI OS is a Miami AI implementation firm that builds a data pipeline, a private firm OS and single-job agents inside service firms. Precision builds, hosts and operates the system while the client's own staff review and approve what agents prepare.

| Fact | Value |
| --- | --- |
| Website | https://getprecision.ai |
| Price | If your firm becomes a client, our work together is governed by the signed Master Services Agreement, Order Form and Data Processing Addendum between us (the "Client Agreement"). |
| Free tier | No |

No free plan, trial or self-serve sign-up. Every entry point is a discovery call, and scope and fees are set in a signed Master Services Agreement and Order Form.

## What it does

**Done-for-you service** · Yes — Delivery is a service engagement, not software a buyer installs: Precision scopes, builds, hosts and operates the pipeline, firm OS and agents, and states clients need no technical staff of their own.

> No. We operate and maintain everything. Your team's role is reviewing and approving what agents prepare.
> — https://getprecision.ai/services/managed-ai · Questions section, answer to "Do we need technical staff?"

## How much it does unattended

**Runs autonomously** · Partly — Agents run preparation work on their own but are deliberately built to stop: anything client-facing or uncertain is queued for a named reviewer, so no task completes end-to-end without a person.

> They prepare work inside your firm. We build every agent so that anything going to your clients, and anything it isn't sure about, waits for a person on your team. Every approval is recorded.
> — https://getprecision.ai/how-it-works · Questions section, answer to "What do agents do without approval?"

**Multi-agent** · Yes — The engagement is described as several single-job agents coordinated around one firm — gather, prepare and check roles handing work along with shared permissions and review points.

> Specialized agents take on the work between your tools: gathering context, preparing outputs and checking details. Precision connects them around your business, with clear permissions and review points.
> — https://getprecision.ai/ · "Many agents. One way of working." section, body paragraph

**Agent permissions** · Partly — Reach is fixed per engagement rather than tuned by the buyer at will: the scope names which systems and documents an agent reads and under which permissions, and work stops at agreed points.

> Access and dependencies: which systems and documents the agent reads, under which permissions, and what must be connected first.
> — https://getprecision.ai/services/ai-agents · Section 03, "What an agreed engagement covers", second bullet

## Which models it runs on

**Claude** · Yes — Anthropic is named as a model provider used for client platform work, and only under zero-data-retention, no-training terms; the specific Claude model is not published.

> OpenRouter and Anthropic | Only for client platform work, and only under zero-data-retention, no-training terms
> — https://getprecision.ai/legal/privacy · Section 5, "Who we share it with" provider table, final row

**Model choice** · Partly — Precision, not the client, decides which models may run: code-level gating blocks any model lacking zero-retention, no-training, no prompt logging and no shared-provider fallback, and no customer-facing model choice is described.

> Before a language model can be called, it must be covered by a zero-data-retention agreement with no training on your data, with prompt logging off and fallback to shared providers off. A model that doesn't meet all four conditions can't be called.
> — https://getprecision.ai/security · "Language models only under zero-retention terms" section, body paragraph

_Not established: GPT, Open models_

## Where you use it

**In a browser** · Yes — Client teams work in a hosted per-firm web workspace with engagement, program and review views; Precision runs everything behind it, so the workspace is the only surface described.

> Your team works in a private workspace for your firm, where they review and approve what agents prepare. We build, run and maintain everything behind it.
> — https://getprecision.ai/platform · Questions section, answer to "Do we log in to use it?"

## Whose machine it runs on

**Self-hosted** · No — The platform runs only in Precision's own Google Cloud estate in the United States and Precision operates it; no bring-your-own-infrastructure or customer-tenant deployment is offered anywhere on the site.

> Hosted on Google Cloud in the United States.
> — https://getprecision.ai/security · "Infrastructure and engineering" list, first bullet

_Not established: Open source_

## What it costs to run

**How it meters** · Partly — Precision places itself in the partner-delivery column of its own comparison table, whose cost line is a build fee plus run costs and support; no rate, retainer or unit figure is published.

> Cost categories | Licences and configuration time | Engineering time, infrastructure and upkeep | Build fee, run costs and support
> — https://getprecision.ai/resources/build-buy-or-have-it-built · "Compare the three paths on the same criteria" table, Cost categories row, Partner delivery column

**Free tier** · No — There is no self-serve or trial entry: every call to action is a discovery call, and the vendor states scope is set in that call rather than assumed, so nothing can be started without talking to them.

> Scope is set in a discovery call, not assumed.
> — https://getprecision.ai/services/ai-agents · Hero paragraph, final sentence

_Not established: API access, MCP server, Bring your own key_

## Buying it for a team

**A company can buy it** · Yes — The buyer is a firm, not an individual: client work runs under a signed Master Services Agreement, Order Form and Data Processing Addendum between the two companies.

> If your firm becomes a client, our work together is governed by the signed Master Services Agreement, Order Form and Data Processing Addendum between us (the "Client Agreement").
> — https://getprecision.ai/legal/terms · Section 1, "What these Terms cover"

**Admin controls** · Partly — Constraints exist but are set in the engagement and in code rather than by a client administrator: agreed read permissions, built-in separation of duties, and an AI use policy Precision writes for the firm.

> Checks for separation of duties are built in, so the person who prepared the work isn't the one who approves it.
> — https://getprecision.ai/security · "A permanent record of every approval" section, final sentence

**Audit log** · Yes — Approvals and model calls are written to records that cannot be edited or deleted through the platform during retention, including by client or Precision administrators, and each model call records job, model and cost.

> When your team approves work, the approval is written to a record that can't be edited or deleted through the Platform during its retention period, including by your administrators or ours, and it's tied to exactly what was approved.
> — https://getprecision.ai/security · "A permanent record of every approval" section, opening sentence

_Not established: Seat model, Pooled budget, Single sign-on_

## What happens to your code

**Opt out of training** · Yes — No opt-out is needed because no-training is enforced as a precondition in code for every model call, and Precision separately states it trains no models of its own on client data.

> A language model can't be called unless it's covered by a zero-data-retention, no-training agreement with prompt logging and shared-provider fallback off, enforced in code.
> — https://getprecision.ai/security · Questions section, answer to "Is our data used to train AI models?"

**Data residency** · Yes — A single jurisdiction is published and no region choice is offered: the client platform sits on Google Cloud in the United States, with daily backups for 30 days and seven-day point-in-time recovery.

> On Google Cloud in the United States. Backups are kept daily for 30 days, with point-in-time recovery over the last seven days.
> — https://getprecision.ai/security · Questions section, answer to "Where is our data stored?"

**Getting out** · Yes — Exit is contractual rather than self-serve: an export window precedes deletion, deletion runs as a reviewed preview, legal holds are respected, and firm encryption keys are destroyed once backups expire.

> When a firm leaves, its data is deleted after it has had time to export it, and its encryption keys are destroyed once backups expire, so any encrypted copy that remains can no longer be read.
> — https://getprecision.ai/security · "Legal hold and controlled deletion" section, third sentence

**Certifications** · Partly — Controls and written policies are mapped to SOC 1 and SOC 2 criteria, but the vendor states plainly that no audit has been completed; no ISO 27001 or GDPR certification is claimed anywhere.

> Our controls are designed against SOC 1 and SOC 2 criteria, with written security policies in place. We haven't completed an audit yet, and we'll say so plainly until we have.
> — https://getprecision.ai/security · "Built toward SOC 2" section, body paragraph

## Provenance

Every fact above comes from the vendor. Nothing here is independently corroborated yet.

## Coverage elsewhere

None recorded. Every fact above comes from the vendor and is not independently corroborated yet.

## What changed

- 2026-10-06 — Capabilities · Commercial terms · Free tier · Price

## Correcting this record

Anything marked "Not established" is a question we have not answered — it is not a No. If you represent Precision AI OS, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth.

Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.
