Skip to content

Technical Overview

AI agents and workflows that reason across your documents

Backed by deterministic workflows for the steps that need guaranteed consistency. Full audit trails. Managed cloud, private cloud, or on-premises.

How It Works

Two modes of operation. One platform.

A conversational AI agent handles the unpredictable work: interpreting documents, answering questions, reasoning across sources. When precision matters, the agent invokes a deterministic workflow that executes identically every time. The user sees one interaction. The platform chooses the right tool for each part of the job.

Deterministic Workflows

Deterministic workflow: linear flow from Download File through Extract Clauses, Classify Risks, Human Review, to Send Email

Every step defined. Every outcome predictable. Auditable end to end.

Agentic AI with Tools

AI Agent reasons about the task and calls tools

AI reasons about the task and picks the right tools. Adapts to novel requests.

Why not both

An agent can call a workflow as a tool. The workflow runs deterministically and returns the result to the agent. Reliability where you need it, intelligence where it counts.

Example

"When was Helen Lane appointed director of Morrison Holdings, and did that happen before or after the share transfer?"

Agent calls the chronology workflow on the disclosure bundle, reads the output, answers with source references.

Workflow Engine

Built for reliability, not just vibes

Workflows are idempotent: rerunnable without duplication or inconsistency. The engine handles scheduling, retries, parallel execution, and state recovery automatically. When a workflow hits a human review gate, it suspends and resumes after approval.

Expressive, readable logic

Business rules expressed as readable code, not sprawling node graphs. A legal professional can read it and verify it's correct.

score = 0
score = score + 3 if contract_value > 500000
score = score + 2 if days_until(renewal) < 90
score = score + 2 if governing_law != "England and Wales"
score = score + 3 if liability_cap = "uncapped"

Auditable, version-controlled, diffable. We also provide a no-code visual builder to generate these expressions.

Production-grade orchestration

  • Idempotent execution: rerun any workflow without duplication
  • Automatic retries with exponential backoff on failures
  • Parallel fan-out across branches with result aggregation
  • Human-in-the-loop gates: suspend, wait for approval, resume
  • Scheduled and event-driven triggers (email, calendar, file changes)
  • Full observability: run status, duration, cost, and reasoning traces

Workflow Builder

Visual or code. Your choice.

Build workflows visually with drag-and-drop, or write expressions for complex logic. Configure extraction schemas, approval gates, and output formats. Connect workflows to agents as tools so the AI can invoke them when the task demands precision.

CompleteFlow Workflow Builder: visual editor showing a contract review workflow with document ingestion, clause extraction, risk classification, and human review nodes

State machine execution

Workflows modelled as state machines with validated transitions. Every state change logged. Version-controlled definitions with rollback.

Agents invoke workflows

An agent can call any workflow as a tool. The workflow runs deterministically and returns the result. AI flexibility with guaranteed consistency.

Triggered or scheduled

Run workflows manually, on a schedule, or in response to events: new emails, calendar invites, file uploads, or API calls.

Agent Builder

From idea to production agent in minutes

Define agents with code or the guided builder UI. Connect data sources, set governance rules, test against real data in a sandbox, and deploy to production with full audit logging from the start.

01

Define the task

Describe what your agent should do. Connect data sources (SharePoint, email, APIs) and set guardrails. Choose your model tier and provider.

02

Test and refine

Run your agent against real data in a sandboxed environment. Review outputs, check reasoning traces, and tune behavior before going live.

03

Deploy with governance

Push to production with full audit logging, access control policies, human-in-the-loop controls, and cost monitoring.

Governance

Every AI decision. Logged. Explainable. Reviewable.

Built to meet the governance requirements of regulated industries, from financial services to legal to government. Not bolt-on compliance. Governance is built into the agent execution pipeline.

Immutable audit trail

Two configurable levels: minimal (summary, model, tokens, cost, policy decisions) and maximal (full prompt and response). Every record timestamped with user attribution and agent version. Default 7-year retention.

Access control

Workgroup-based role system with default-deny policies. Users access only the documents, workflows, and tools their role permits. Every access decision is logged.

Human-in-the-loop

Agents escalate to humans when confidence drops below configurable thresholds. Review queues surface low-confidence outputs for approval, rejection, or correction, and corrections feed back into agent improvement.

Workgroup permissions

Organize users into workgroups (matters, teams, projects) with granular roles: owner, contributor, reviewer, viewer. AI agents can only access resources the user is authorized to see.

Cost tracking & attribution

Per-agent, per-user LLM cost attribution with token-level granularity. Track spend by model tier, provider, and department. Set usage quotas and budget alerts per team.

Reasoning traces

Every agent output includes the chain of tool calls, data sources consulted, confidence scores, and the decision path that led to the result. Exportable for regulatory review.

Enterprise Features

Built for production

Multi-provider LLM

Swap between Anthropic, OpenAI, Azure OpenAI, and open-weight models without changing agent code. All commercial API tiers. Your data is never used for model training. Model registry maps tiers to providers.

Channel abstraction

Agents are channel-neutral. Deploy the same agent to Teams, Copilot Chat, web UI, or API. The channel adapter handles formatting and auth.

Delegated OAuth

Agents inherit the requesting user's Microsoft 365 permissions via delegated tokens. No separate credential store. No over-provisioned service accounts.

Workflow orchestration

Multi-step workflows with parallel fan-out, conditional routing, human gates, and automatic retry. Idempotent execution means workflows are rerunnable without duplication or inconsistency across any environment.

Workgroup-based access control

Organise resources into workgroups with granular roles. Control exactly what each user can access per workgroup. Single-tenant deployment with database-level isolation.

Natural language interface

Users invoke workflows through conversation in Teams, Copilot, or the web UI. Agents auto-discover registered workflows and expose them as conversational tools.

Review queue

Batch approval dashboard for human oversight. Review low-confidence outputs, approve or reject with notes, make field-level corrections. Workflows suspend and resume automatically around human decisions.

Vector search

Built-in similarity search for retrieval-augmented generation across your internal document corpus.

Container-native deployment

Docker Compose for development, Azure Container Apps for production. Hub-spoke VNet with private endpoints. IaC with Bicep.

Agent versioning

Version-controlled agent configurations with rollback capability. Promote agents through dev, staging, and production environments with full traceability.

Webhook notifications

Notify external systems (Slack, email, SIEM) when agents complete tasks, escalate to humans, or trigger policy violations.

SSO & identity

Native Microsoft Entra ID integration. Extensible to Okta, Google Workspace, and SAML 2.0 identity providers for broader enterprise deployment.

Deployment

Your deployment. Your rules.

Recommended

Private Cloud

Deploy on your own Azure, AWS, or GCP tenancy. Container-native with Azure Container Apps or ECS. Hub-spoke VNet with private endpoints. Data never leaves your environment.

Fastest setup

CompleteFlow Cloud

Hosted and managed by us on private cloud infrastructure. We handle ops, updates, and monitoring. You get the fastest path to production with data residency in your chosen region.

Maximum isolation

On-Premises

Full air-gapped deployment on your own hardware. Docker Compose or Kubernetes. Open-weight models only. Maximum isolation for the most sensitive workloads.

Integrations

Connects to your existing systems via MCP

CompleteFlow agents connect to external systems through the Model Context Protocol (MCP), the open standard for tool integrations. Native Microsoft 365 support via Graph API with delegated user permissions, plus any system with an MCP server.

Microsoft 365
SharePoint
Outlook
Google Workspace
Salesforce
HubSpot
Slack
Atlassian (Jira)
Confluence
GitHub
Snowflake
PostgreSQL
MongoDB
BigQuery
Notion
Dropbox
Stripe
Xero
AWS
Azure

Plus anything with an MCP server. The list grows every week. Custom integrations scoped during pilot.

FAQ

Technical questions

What models does CompleteFlow support? +
CompleteFlow supports multi-provider LLM access: Anthropic Claude, OpenAI GPT, Azure OpenAI, and open-weight models (Llama, Mistral) running in your own tenancy. All cloud models are accessed through commercial API tiers that do not use your data for training. A model registry maps tiers (budget, standard, premium) to provider/model combinations, so you can swap models without changing agent code.
How does access control work? +
CompleteFlow uses workgroup-based role access control with default-deny policies. Users are assigned roles (owner, contributor, reviewer, viewer) per workgroup, and permissions are enforced on every resource access and tool invocation. Every access decision is captured in the audit trail.
How does the audit trail work? +
Every agent action is recorded at two configurable levels. Minimal captures: agent ID, user, action summary, model used, tool calls, token usage, cost, confidence score, duration, and policy evaluations. Maximal adds the full prompt and response. All records are timestamped and retained for 7 years by default.
Where does my data go? +
Your data stays in your chosen deployment environment. Models run in your cloud tenancy or on-premises. Agents access documents via delegated OAuth. They inherit the requesting user's Microsoft 365 permissions, so there's no parallel credential store or over-provisioned service account.
How do agents access our documents? +
Through the Microsoft Graph API with delegated user tokens. When a user asks an agent to research something, the agent searches SharePoint and OneDrive using that user's existing permissions. No admin consent for broad access required.
What happens when an agent isn't confident? +
Configurable confidence thresholds trigger human-in-the-loop escalation. Low-confidence outputs are routed to a review queue where team members can approve, reject, or correct them. Corrections feed back into agent improvement.
How does the workflow engine work? +
Workflows are idempotent and rerunnable without duplication or inconsistency. The engine handles scheduling, retries, parallel execution, and state recovery. When a workflow hits a human gate (e.g., low confidence score), it suspends and resumes after review. No polling, no timeouts.
Can we run this without Microsoft 365? +
Yes. The channel layer is abstracted, and agents are channel-neutral. The web channel and API endpoints work independently of M365. Microsoft integration is the primary channel for enterprise deployments but not a hard dependency.

See the platform in action

Book a 30-minute technical walkthrough tailored to your industry and infrastructure requirements.

Book a Technical Demo