Resolve
L0 Agents answer on published workflows, approved knowledge, and authorized tools.
Why OpenOS
The operation, with a defensible record on every eligible interaction.
OpenOS is an AI-native customer operations company for India's banks, insurers, and NBFCs. We take over inbound support on WhatsApp, email, and voice, resolve the routine majority with AI, and hand exceptions to a small human team with the transcript, summary, and recommended action.
Quality audits every eligible closed interaction instead of a sample. Compliance gets a record that can be inspected against RBI and IRDAI clocks. The client keeps their telephony, WhatsApp provider, helpdesk, and CRM.
What we run
Four moves. One operating record.
A bot, a queue, a scorecard, or a dashboard can each help. The operation is the four of them on the same interaction.
Escalate
Workspace gives a person the transcript, summary, customer context, and a recommended action.
Audit
Quality scores configured eligible closes on a versioned scorecard and shows coverage against that population.
Decide
Trinetra turns falling containment, SLA risk, and repeat contact into a Decision Card with an owner and a deadline.
Follow through
A selected recommendation becomes a managed task. The card itself is not the assignment.
Claim boundaries
Two alternatives already sit in the account.
A traditional BPO sells headcount. A US chat-bot vendor sells English-first deflection. Neither produces an audited human exception desk for regulated India.
AlternativeHow it is soldWhat OpenOS runs
Indian BPOSeats. Every automated contact cuts their revenue.Per resolved-and-audited interaction. AI on the routine majority, people on exceptions.
US bot vendorEnglish-first chat deflection, priced on containment.Indic speech via Sarvam, a human exception desk, and an audit of every eligible close.
The client's stackTelephony, WhatsApp BSP, helpdesk, and CRM stay authoritative.We do not rip and replace those systems. Production connections are scoped per client.
The inspectable object
Leaders get a Decision Card, not another chart.
Trinetra generates the issue, evidence, impact, recommendation, proposed owner role, deadline, and confidence. A person selects the recommendation before it becomes managed work.
- It is not an alert: it frames a decision.
- It is not a task: a selected recommendation can become one.
- It is not proof of outcome: the affected measure is reviewed after action.
Decision required
High confidence Contain the return-status escalation spike
Twelve unassigned return-status cases are driving repeat contact while L0 containment for the same intent is falling.
Business impactSupport capacity and repeat-contact risk
Proposed owner roleHead of Customer Support
DeadlineWithin 72 hours
RecommendationPublish the corrected workflow and route unassigned cases
Evidence
- Containment declined for return-status intent
- Twelve cases have no current assignee
- Repeat contacts increased in the demonstration window
What is live
Production on AWS Mumbai. Voice is live.
Email and WhatsApp
Inbound L0 resolution, escalate to Workspace, Quality on eligible close, and Trinetra Decision Cards.
AWS ap-south-1
The production platform is hosted in Mumbai. Models in use are Amazon Bedrock (Qwen), OpenAI where a workflow requires it, and Sarvam for Indic speech.
Voice
Voice is live on Exotel and Sarvam. 11 Indian languages today, 22 targeted. The provider still carries the call.
Connectors
Adapters exist. A production integration is scoped to the client's APIs, credentials, and acceptance. A logo is not a live connection.
Tailored walkthrough
Bring one regulated workflow.
We will map the channel you already own, the exception path, the eligible audit population, and the baseline the 60-day guarantee is measured against.
- 01Your current loopChannels, systems, volume, and the operational gap.
- 02The product proofResolve, escalate, audit, and decide in one scenario.
- 03A scoped first deploymentOne workflow with explicit evidence and success measures.
Prefer email? sales@openost.com