Choose the scope that fits
The AI Agent package covers one role and channel, with defined permissions, evaluations and a practical handover. A predictable sequence may fit AI Workflow instead.
AI AGENTS
An agent with a clear job to do.
Build an agent that retrieves approved context, selects permitted tools and routes consequential actions for human approval.
The AI Agent package covers one role and channel, with defined permissions, evaluations and a practical handover. A predictable sequence may fit AI Workflow instead.
One optional knowledge collection is limited to 50 text-readable documents and 500 pages in one language. Multi-agent systems, unrestricted autonomy and custom portals need separate scope.
An agent is useful when the next action depends on context: finding relevant information, choosing an approved tool, or deciding what needs review. A fixed sequence of steps is often better served by a simpler workflow. We make that choice during discovery.
Name the process owner, information sources, permitted tools and actions that require approval. Agree what the agent must decline or hand back to a person.
Use representative tasks to check tool selection, missing information and failure handling. Record evaluation results against the agreed acceptance criteria before activation.
Receive the agreed prompts, configuration, project exports, documentation and training. You control your accounts and can stop the agent or return the process to manual handling.
ONE OUTCOME. AN AGREED SCOPE.
Fixed implementation pricing. We confirm fit and scope before payment. Delivery estimates start after scope approval, access and sample data are ready.
Automate one repetitive task.
One-time implementation
Applicable GST and running costs extra
A clear process with accessible tools: lead capture, task creation, record updates or a scheduled report.
Process mapping, supported connectors and field mapping, duplicate protection where the destination supports it, bounded retries and failure notification. Deployment into one existing customer-owned automation environment is included.
A simple AI step follows a fixed instruction and produces an agreed structured result. This package does not include an autonomous agent.
Tests cover missing data, duplicates and a failed destination where applicable. Handover includes the workflow export, configuration notes and a short operating guide.
Outside scope: Document OCR, knowledge retrieval, custom APIs, browser automation, approval interfaces and agents that choose their tools are outside this package.
Turn incoming information into business steps.
One-time implementation
Applicable GST and running costs extra
Document processing, email triage, lead qualification, onboarding or operational reporting. Choose one primary use case.
Structured AI output, field validation, exception routing, duplicate handling, bounded retries and failure alerts. A review queue, basic run log and usage summary use existing supported tools.
If document processing is selected: one document category, one language, up to 3 agreed layout families, up to 20 extracted fields and 10 pages per file. Inputs must be legible printed text or scans, without complex line-item tables. These limits describe the selected use case, not an extra deliverable.
Low-quality or unsupported input is flagged for review; required values are not silently invented. Handover includes workflow exports, prompts, configuration documentation and an operating guide.
Outside scope: Handwriting, arbitrary document layouts, line-item reconciliation, tax decisions, accounting filing, open-ended agents, custom portals and new applications are outside the standard scope.
An agent for one defined business role.
One-time implementation
Applicable GST and running costs extra
Requests that need context, selection among permitted tools, and approved actions within a clear operating boundary.
Choose one supported form, existing web interface, email or Telegram bot. An application used as the channel counts toward the application limit. Each tool has a documented purpose, input, permission boundary and output; an unrestricted API executor is not one tool.
Optional knowledge collection: up to 50 text-readable documents, 500 pages total, one language and one existing folder or upload source. Includes initial ingestion and one scheduled refresh. Complex OCR, website crawling and document-level access control are separate.
Defined instructions, structured output, tool allowlists, maximum steps per run, timeouts and supported usage limits. Includes budget alerts; hard spending ceilings are promised only where the provider and architecture can enforce them. Approval expiry, rejection handling and a stop procedure are defined.
Tests include unauthorized requests, misleading instructions in inputs, tool failures, repeated requests and human escalation. Handover includes configuration, prompts, project-specific code, workflow exports, an evaluation report and runbook.
Outside scope: Multi-agent systems, multiple departments, unrestricted autonomy, custom portals or mobile apps, WhatsApp onboarding and enterprise access requirements need a separate proposal.
Hosting gives you a running application that you configure. Development builds a specific business process with agreed tools, permissions and tests. Hosting alone does not include designing an agent for your team.
The AI Agent package can include one knowledge collection of up to 50 text-readable documents and 500 pages in one language. Access rights, document suitability and updates are agreed in scope; larger collections need a separate proposal.
Consequential actions require human approval by default. An agreed scope may allow specific routine low-risk actions after testing. Unrestricted autonomy and guaranteed error-free decisions are not included.
Applicable GST, hosting, model usage and platform fees are extra. We agree running costs, acceptance criteria and human approvals before purchase. Free discovery is one 60-minute initial session per business.