Costly exceptions look ordinary
A customs hold, last-free-day warning or schedule change arrives beside routine status chases. The inbox gives both the same visual weight.
AutonomeX · governed intelligence for operations
InboxOS · email operations for freight teams
The freight inbox,
already understood.
InboxOS turns bookings, quotes, customs holds, shipment updates and invoice queries into structured tickets with priority, risk and a ready-to-review draft. Your team approves every reply.
Logistics pack shipped
Human approval stays on
Every step recorded
One email. Four controlled stages.
Fast enough for the desk. Clear enough for review.
- IntakeEmail + documents
- UnderstoodIntent + fields
- RoutedPriority + risk
- ApprovedA person sends
Built now. Built next.
Ambition, with the boundary visible.
InboxOS is the first product from AutonomeX, a company building governed intelligence and execution systems for AI-native enterprises.
Shipped in the current build
Understand and prepare the work
- Ingest and classify inbound email
- Extract required fields and flag gaps
- Assess priority and operational risk
- Route to a queue and prepare a draft
- Keep human review before outbound send
Product direction
Govern and execute the full workflow
- Authoritative validation and policy evaluation
- Configurable, risk-based approvals
- Downstream operational actions
- Production integrations and enterprise identity
- Complete lifecycle evidence
The operational gap
Email starts the work.
It should not hide it.
Freight teams do not lack information. They lack a reliable way to turn unstructured messages into visible, owned work before the clock runs out.
Context lives across threads
References, routes, documents and prior replies stay buried instead of becoming one coherent ticket.
Skilled people become routers
Operators spend judgment on sorting and re-keying before the real work even begins.
Product, in context
From message to
review-ready work.
Customers keep emailing as usual. InboxOS makes the operational structure visible behind the scenes.
See what needs attention
One queue, already prioritised.
Bookings, quotes, holds and status requests arrive with intent, confidence, priority and risk visible at a glance.
- High-risk work rises above routine mail
- Each item has a clear review state
- Operators start with context, not triage
Move from email to action
Details extracted. Gaps made obvious.
The ticket brings the message, required fields, risk signals and a draft together, ready for an operator to check.
- Required details stay attached to the request
- Missing information is flagged, not guessed
- The outbound draft remains under human control
Reconstruct the decision
A record of how the work moved.
InboxOS records processing steps and review activity so an operator can follow what happened without reverse-engineering a black box.
- Processing steps stay visible
- Prompt versions are recorded in current runs
- Review and send events remain attributable
One real workflow
A customs hold,
made impossible to miss.
“Container MSKU7601234 is on customs hold at Rotterdam. They need the commercial invoice and packing list today.”
InboxOS recognises the exception, extracts the reference and required action, then raises its risk and priority for an operator.
✓A person checks the documents, edits the draft and sends.
-
01
Intake
Email and attachments arrive
The customer uses the same shared inbox as before.
-
02
Understanding
Customs hold identified
Shipment reference, hold type and required documents become structured fields.
-
03
Risk-aware routing
Time-critical work rises
Demurrage exposure moves the ticket ahead of routine status requests.
-
04
Human authority
An operator approves the reply
The system prepares the work. The person remains responsible for the send.
Platform with industry packs
Freight is the wedge.
Governed operations is the company.
InboxOS keeps the operational core stable while each pack defines the industry's request types, required fields, risk and priority.
InboxOS core
Understand · route · draft · record
IngestionOrchestrationReviewAudit
Logistics &
freight forwarding
Seven request families across bookings, status, quotes, documents, exceptions, claims and invoicing.
Wealth &
private banking
A second domain on the same product core, proving the pack model can travel.
Insurance · health
legal · more
Configured with a real taxonomy and workflow evidence before it is presented as shipped.
Register interestControlled autonomy
Useful AI, held inside
clear authority.
AutonomeX calls this Pragraha: autonomy is valuable when its boundaries are designed, visible and proportional to risk.
Human approval before send
InboxOS prepares and routes. A person reviews every outbound reply in the current product.
Traceable by design
Processing, review and send events are recorded so the path can be reconstructed.
Single-tenant deployment
The current deployment model uses one Docker-based instance per organisation.
Role-based access
Current product permissions limit who may view or perform sensitive actions.
About the founder
Built from inside the systems
this product is meant to change.
Shubhankar Mittal is the founder of AutonomeX and an enterprise AI systems builder focused on turning fragmented communication, data and policy into governed operational work.
Across banking, compliance, consulting and AI product architecture, he kept encountering the same failure: intelligence was available, but trustworthy execution across people and systems remained difficult. InboxOS is where those lessons converge.
01
Kotak Mahindra Bank
Learning to turn data into decisions.
Worked with more than 60 million transactions across customer analytics, merchant classification and behavioural insight. The lesson was foundational: data matters only when it changes a decision.
02
UBS
Seeing why enterprise automation needs governance.
Worked inside compliance and operational-risk analytics, translating fragmented processes into controlled automation. Across UBS and Deloitte, he later trained more than 1,000 professionals in practical Python automation.
03
Deloitte
Moving from analysis into operational execution.
Helped build operational-risk automation across 11 risk taxonomies. A process that took approximately 20 working days was reduced to approximately two minutes, while human review remained where judgment was required.
04
Accenture Strategy & Consulting
Designing governed AI as a product system.
Led AI product architecture and enterprise solution strategy across financial services, including relationship-manager orchestration and a governed agentic data-product platform.
05
AutonomeX · full-time since 2025
Building the governed execution layer.
After leaving consulting, Shubhankar built knowledge systems, interactive agents and financial-assistance products to test where AI creates durable value. The conclusion became AutonomeX: autonomy should be bounded by explicit authority, evidence and human judgment proportional to risk.
The operating beliefs behind AutonomeX
Systems over isolated intelligence
A strong model is only one component. Authority, context, tools, people and evidence determine whether it can create real operational value.
Governance belongs in the architecture
Trust is not a claim added after launch. It must appear as explicit boundaries, review states, permissions and records.
Human judgment scales with risk
The goal is not maximum autonomy or permanent manual review. It is the right authority for the ambiguity, impact and risk of each action.
Request demo access
Bring us the inbox
your team knows by heart.
We will walk through InboxOS using the operational requests that matter to your desk. Freight and logistics teams are first in line for design-partner access.
- 01
See the current product clearly. Queue, ticket, draft and trace, with no roadmap features presented as live.
- 02
Keep the human boundary on. Every outbound reply remains under operator review.
- 03
Shape the pack around real work. Bring the request types and exceptions your team handles every day.
Demo access request
Tell us about your desk.
Your request is ready.
We copied the details to your clipboard. Send them to the AutonomeX team, or open your email app below. We review every request before granting access.