Foreman
The workflow engine — runs steps in order, handles retries, and keeps an audit trail.
AI agents that run your operations without losing control of them
Speed and control shouldn't be a tradeoff. Sotura runs your operations workflows in the background — with approvals, audit, and budget control built in from day one. Start with us running it. Take over when you're ready.
Most managed workflows go live in 2–3 weeks.
console.sotura.com/tenants/acme
Sotura ConsoleOperator overview
Tenantacme
Approval queue
Gate decisions before typed tools execute side effects
Pending approvals
3
Active workflow runs
2
Audit events (24h)
148
AI budget used
62%
Scheduled agent workflow
Policy gate · typed tool execution
Full audit trail attached
Batch classification run
Agent slot · structured output review
Full audit trail attached
Weekly operations report
Report step · no side effects
Full audit trail attached
Recent workflow runs
| Workflow | Status | Created |
|---|---|---|
| tenant-sync-flow | completed | Mar 10, 9:14 AM |
| approval-policy-check | running | Mar 10, 8:42 AM |
console.sotura.com — one shared dashboard for approvals, workflow runs, audit, and usage across every client.
See it in practice
Every client runs on the same system. Our first engagement is in active development — TrustPoll's workflow, with Sotura handling approvals and audit behind the scenes.
Current engagement
TrustPoll.io — first engagement, in active development
Governed polling workflows with human approval before anything goes public.
Topic intake, verification, and draft release — each step audited, nothing published without sign-off.
trustpoll.io/workflows
TrustPoll.ioCurrent engagement
Live Sotura buildWorkflow steps
Derived from tenant Prefect flows
D1Prefect flow
Topic intake
completedD2Prefect flow
Verification
completedD3Prefect flow
Draft release
runningD4Prefect flow
Approval queue
pendingWorkflow activity
Topics prepared
86
72 complete · 14 pending
Draft releases
24
6 awaiting approval
Votes captured
12
This week
Pending approvals
6
42 qualified issues
Approval queue
Operator review before release
TrustPoll.io
Workflow review
ApproveReject
Approve the release path for the current workflow surface and keep the audit trail intact.
TrustPoll's day-to-day workflow dashboard.
console.sotura.com/tenants/acme/audit
Sotura ConsoleAudit log
Tenantacme
Audit log
Durable record of policy checks, approvals, and side effects
All eventsApprovalsPolicySide effects
| Time | Actor | Action | Resource |
|---|---|---|---|
| 09:14:02 | policy.gate | approval.required | scheduled-agent-workflow Typed tool blocked pending operator sign-off |
| 09:12:48 | operator@acme | approval.granted | batch-classification-run Agent slot output reviewed and released |
| 08:55:11 | workflow.engine | run.completed | tenant-sync-flow 12 steps · 0 side effects · audit attached |
| 08:42:03 | budget.guard | usage.recorded | acme-tenant AI spend 62% of monthly cap · 148 events (24h) |
console.sotura.com/audit — durable record of policy checks, approvals, and side effects.
Here's how it works under the hood
One platform, many tenant experiences
Sotura owns orchestration, governance, and audit. Client products plug in through adapters and custom operator dashboards — without re-architecting the core.
Who does what
Four roles, one system
Each role has a clear job — and none of them bypass approvals or audit.
Worker
AI inside a step — research, draft, classify. Flexible thinking, but every action is gated.
Human operator
Your team in the admin console — approves or rejects work before anything ships.
Front-desk
Chat on WhatsApp, Slack, and similar — same rules as admin, not a second brain.
Worker ≠ front-desk.Chat interfaces plug into the same backend — they don't replace the workflow engine.
Why Sotura
Governance. Visibility. Consistency.
Built for teams that need autonomous operations with the compliance and oversight their stakeholders require — not another chatbot wrapper.
Governance stays with leadership
Define what agents may do before anything runs in production. Policy, approval gates, and risk checks sit between AI output and real-world side effects — not buried in prompts.
Every decision is visible
Workflow runs, tool calls, and approvals roll into a durable audit trail. Operators see what happened, why it happened, and who signed off — without digging through logs.
Consistency at scale
The same step runtime — gate, agent, typed tools — repeats across every workflow. Swap models or agents without rewriting how your organization stays in control.
Risk controls
No side effect ships without a control point.
Each workflow step follows the same runtime: policy checks happen first, risky actions can pause for approval, agent output stays structured, and typed tools are the only way to touch your systems.
1
Policy gate
Disallowed actions stop before they can execute
2
Approval gate
Higher-risk steps wait for human or policy sign-off
3
Typed execution
Email, social, CMS — only through audited tools
4
Audit + budget
Every run is tied to tenant, workflow, and spend
What to expect
From first call to production
A typical Sotura Managed engagement moves quickly without skipping governance. Here is what the first few weeks look like.
1Week 1
Discovery & scoping
We map the workflow, approval points, and integrations. You leave with a written scope and a staging tenant to review against.
2Week 2
Workflow build
We implement agents, typed tools, and approval gates against your process. You test runs in staging and tune policy before production.
3Week 3
Live with approvals
The workflow runs on Sotura infrastructure with your team in the loop. Side effects only execute after policy checks or operator sign-off.
Book a call.
Tell us what you're building — managed deployment, platform, or product updates. We'll get back to you within one business day.
Fast response
We usually reply within one business day.
Clear next step
We review your note and suggest managed or platform if there is a fit.
Direct conversation
You'll talk with the Sotura team handling the work.