01
Name the job and how it could go wrong.
We define the task, approved sources, expected output, people affected, unacceptable failures, and the actions that require human review.
AI for a real job. With a real person accountable for it.
We build and operate a defined AI workflow using the sources you approve. Review rules, fallback behavior, provider costs, and the things it must never do are agreed before launch.
Bring us the repetitive task
- The job comes before the model
- Humans stay in high-risk decisions
- Outputs and exceptions get measured
Start with the annoying task. The model comes later.
The brief follows one sequence: define the job and risk, build the smallest useful workflow, then measure and maintain it.
02
Build the smallest version worth using.
We configure the model, retrieval, integrations, instructions, review step, logging, and fallback needed for that job, not a general AI platform looking for a purpose.
03
Measure it. Fix it. Keep it useful.
We review the agreed quality and operating signals, correct source data, adjust the workflow, and manage provider or model changes under the service scope.
Useful AI is usually less dramatic than the sales demo.
Good. It should solve a specific problem. We only include building blocks with an approved source, owner, action, and review boundary.
Answers from approved sources
Website or internal answers grounded in the hours, services, policies, documents, and other sources your organization approves.
Drafting with review
Replies, summaries, articles, messages, or follow-up drafts routed to the person or team responsible for approval.
Extraction and organization
Turn incoming documents, forms, notes, or messages into structured fields, summaries, tags, and next-step queues.
Routing and triage
Classify routine requests, identify missing information, and send work to the right person without pretending every decision is automatic.
A short answer instead of five dashboards
Combine approved operating signals into a digest so staff can review exceptions instead of checking several systems.
A route back to a human
Source limits, approval rules, blocked topics, logs, and a clear next step when the workflow is uncertain.
Automate the boring, observable, reversible work first.
A practical workflow can save real work before anyone asks the system to make a sensitive decision or act without review.
Strong fit
The same task keeps coming back.
People regularly answer the same question, summarize the same material, or move the same information between systems.
A useful draft would already save time.
The workflow can create value before anyone trusts it to take an irreversible action.
A person can own the weird cases.
The organization can identify who approves sensitive outputs, corrects source facts, and handles uncertain cases.
Do not automate this yet
The facts are missing or unreliable.
A model cannot create a trustworthy operating fact from stale policies, contradictory records, or information no one owns.
The system would make a high-stakes decision alone.
Medical, legal, employment, credit, safety, or other consequential decisions need specialist controls beyond a general marketing promise.
The only goal is ‘add some AI.’
A feature should remove a known task or improve a measured outcome. A broad transformation label is not a sufficient scope.
AI needs rules before it needs autonomy.
The agreement identifies what the system may read, produce, store, recommend, and do, plus the person responsible for review and exceptions.
Managed by Rehost
- Workflow design for the approved job and source material
- Configuration, integration, testing, and documented review behavior
- Monitoring and routine adjustments within the agreed service
- A fallback path when the system lacks confidence or permission
Named in the agreement
- Approved data sources, processors, subprocessors, and access rules
- Human-review requirements and prohibited automated actions
- Retention, logs, evaluation criteria, and incident handling
- Usage limits, third-party fees, model changes, exports, and service-end steps
AI pricing without pretending every workflow is the same.
Rehost does not publish a stand-alone chatbot price. The Business starting price is first-party context; the signed proposal defines the AI job, usage, providers, controls, and any added costs.
Business
From $950
USD per month
Practical AI work can be included in the Business operating scope alongside the app, website, customer workflows, and ongoing support. The proposal names the actual job and limits.
Review Business pricingCheck the price, rules, and data story.
These pages show the sources behind price, ownership, review, privacy, and provider boundaries.
Decision guide
Would an AI agent help?
Check the task, source, risk, and human owner before you build.
Operating model
How Rehost works
Review ownership, evidence, account, request, and transition boundaries.
Public policy
Privacy and subprocessors
Read the public privacy policy and current subprocessor disclosure.
Questions to settle before AI touches real work.
Straight answers about models, review, data, pricing, and measurement.
Is Rehost selling access to a chatbot?
No. Rehost scopes a practical workflow around a defined task, approved information, integrations, review rules, and an operating owner. A model may be part of that workflow, but access to a general chat interface is not the product.
Which AI model does Rehost use?
The model depends on the task, data, quality needs, latency, cost, and contractual constraints. Providers and models can change. The scope should describe the job and controls so the workflow is not defined by one model name.
Can AI publish or message customers without review?
Only when the signed scope explicitly permits that action and defines the source, guardrails, monitoring, and fallback. Human review is the safer default for sensitive or customer-facing output until the workflow has evidence that supports a narrower rule.
How is our data handled?
Data handling follows the signed customer agreement and Rehost's public privacy and subprocessor disclosures. The scope should name approved sources, access, retention, processors, logs, exports, and prohibited uses for the workflow.
Who owns our data and the AI workflow?
You retain the content and data you provide. Under the public Terms, Rehost owns its service software, code, and designs unless a signed customer agreement grants a different license or handoff. The agreement should distinguish customer data, configured workflow behavior, vendor accounts, exports, and service software.
Does practical AI cost extra?
The public Business plan starts at $950 per month and can include defined AI work with the app, website, and operating service. Unusual model usage, data processing, compliance, or third-party costs are scoped before work begins.
How do we know whether the workflow works?
The scope identifies observable signals for the job, such as review acceptance, response coverage, routing accuracy, time to resolution, or the number and type of exceptions. Rehost reports measured results without turning them into a universal guarantee.
Bring one repetitive job, not an AI transformation brief.
Tell us what repeats, where the source facts live, who reviews exceptions, and what the workflow must never do.