Coru
Book a demo

ADI AGENTS · SUPPORT

Resolves on first contact, on any channel

One agent with the same logic on voice, WhatsApp and web, with context crossing all three and reaching the human operator intact.

Flows, knowledge base and guardrails defined once, executed on all three channels
Hybrid RAG over a versioned base: a question in the customer's words finds the right document
Strict topical focus — outside enabled subjects the agent routes
Escalation with the complete conversation and the data already collected
ISO 27001 PCI SOC 2
14:32 ▮▮▮ ⌁ 84%
ADI · Example Bank online
my statement has a fee i don't recognise 14:30
Let me check. It is the out-of-network ATM withdrawal fee, charged on the twelfth of August, of nine reais and ninety centavos. 14:33
and how do i avoid it? 14:36
Message

The operational jump

From overflow to resolution

Both models get the same contact. What changes is how many times the customer tells the story.

Three bots on separate stacks

× Each channel has its own knowledge base and its own script
× The customer who switches channel starts from scratch
× At the hand-off to a human, context is lost again
× Imprecise RAG: the bot answers, but answers wrongly or slowly
× First-contact resolution does not rise and overflow keeps growing

ADI support agent

One logic defined once and executed on voice, WhatsApp and web
Live context: starts on WhatsApp, continues on voice, ends on the web
Hybrid RAG, vector search with BM25 over a base versioned per agent
Strict topical focus: out of scope, it routes instead of offering an opinion
Escalation with the full thread and the data already collected

Results in production

Impact that is measured

Aggregated platform results, measured in production. Not attributed to any specific account.

70% of volume automated, with human relief on demand
53% OpEx reduction in contact operations
<800 ms time to first byte of a voice response

How it runs

From contact to resolution

01 · ENTRY

Any channel

Inbound voice, WhatsApp, web widget, hosted chat page or app webview.

02 · CONTEXT

History retrieved

Previous conversations and customer data load at the start, without asking them to repeat.

03 · RETRIEVAL

Hybrid RAG

Vector search combined with BM25 over a versioned base, to find the document in force.

04 · ANSWER

Inside scope

Topical focus limits the agent to enabled subjects; outside them, it routes.

05 · ACTION

Resolves, not just informs

When the case needs an operation, the agent executes it within the permitted catalogue.

06 · ESCALATION

Human with the full thread

By the rule you define, the operator receives the complete conversation and the collected data.

Capabilities

Resolving without repetition

Live cross-channel context

The conversation continues on voice, WhatsApp and web without restarting, and the same holds at escalation.

Knowledge base versioned per agent

Every publish creates a version, with retrieval proof before it goes live.

Consolidated inbox

Conversations from every channel, with filters by state and assignment, labels and priority.

Our CRM or yours

If you already have a CRM the platform integrates with yours; if you do not, it brings its own.

Metrics in one place

Resolution, satisfaction and cost per contact, broken down by channel, by agent and by period.

Strict topical focus

Outside enabled subjects the agent does not offer opinions: it routes to an operator.

Feasibility

Connects to what you already have

Channels, knowledge base and CRM connected through a documented interface. No core changes.

Inbound PSTN voice WhatsApp Business Web widget Hosted chat page App webview HTTP Tools MCP +1000 integrations
Timeline

Onboarding in 1 to 3 weeks. With an existing SIP trunk, the voice channel goes live in 2 to 5 days.

Inside the rules

Supporting without leaving policy

Guardrails in code

Deterministic filters before and after the model, three of them mandatory on any agent.

Crisis signals

Crisis indicators trigger an operational alert and routing, with no configuration required.

Per-turn traceability

What the agent said, on which prompt and knowledge base version, and which filter acted.

Tokenised PII

Personal data is tokenised before any call to a model provider.

FAQ

Frequently asked questions

Does the customer have to repeat the case when switching channel?

No. Flows, knowledge base and guardrails are defined once and executed on all three channels, and context travels with the conversation between them.

How do we guarantee the answer comes from the document in force?

The knowledge base is versioned per agent, with retrieval proof before publishing, and search is hybrid: vector combined with BM25.

Can the agent comment on subjects outside scope?

No. Topical focus limits the agent to enabled subjects. Outside them it routes to an operator instead of answering.

Do we need to change CRM?

No. If you already have a CRM the platform integrates with yours. If you do not, it brings its own, with a consolidated inbox per channel.

How do we measure the result?

Resolution, satisfaction and cost per contact sit in one place, broken down by channel, by agent and by period.

Request a demo

Bring your monthly volume per channel.

With your current resolution rates and your cost per contact, the projection is yours.

What you bring

Monthly volume per channel, current first-contact resolution rate and cost per contact.

What we bring

Benchmarks measured in production, the knowledge base built on your documents and a dedicated compliance liaison.

What comes out

A resolution and cost projection, a proposed pilot KPI and the basis of the rollout plan.

Shall we start?

Please use your work email.

By submitting you accept our privacy policy. We do not share your information with third parties.

Newsletter

Insights on AI in regulated operations, once a month.

Governance, architecture and what happens after go-live. No spam.

By subscribing you agree to receive communications from Coru.