Skip to Content

AI FOR CUSTOMER SERVICE

AI agents for customer service.

Answer order and delivery questions with your real ERP data, and hand the rest to your team. Part of the AI agent platform.

Start freeSee what it resolves

What can an AI agent resolve in customer service, and what can't it?

It resolves the routine questions your records can answer, such as order status or deliveries, and hands everything else to a person with the context ready. Here is exactly where that line is.

THE SUPPORT BOTTLENECK

Most questions already have an answer in your system.

Your team still has to open the order, read the history and type the reply.

The same questions, every day

"Where is my order?" takes a minute to answer, and it arrives all day.

Context lives in several places

Orders, deliveries, tickets and past conversations sit in different screens.

Answers depend on who replies

Quality and tone change with experience and with how busy the day is.

THE CATEGORY

An AI agent for customer service, not another chatbot.

An AI agent for customer service reads the customer's real records, answers what they support and routes the rest to a person.

AI agent inside your ERP Typical chatbot
Where answers come from Orders and history in your ERP A script or FAQ someone maintains
"Where is my order?" Status of that specific order Generic reply or tracking link
When it doesn't know Hands over with the context Loops or gives a vague answer
Permissions Same roles as your team Separate from your system
Traceability What it read, answered and escalated Chat transcript only

New to the concept? Read what an AI agent is and how it differs from a chatbot.

HOW IT WORKS

Read, answer, hand over.

1 Reads the context

Orders, deliveries, tickets and past messages, based on permissions.

2 Answers from your records

Routine questions get a reply in your tone and within your policies.

3 Hands over and logs

What needs a person goes to your team with the context, and it's logged.

WHAT IT RESOLVES, AND WHAT IT DOESN'T

Resolved alone, with a person, or never.

The limits are part of the design. You set the escalation rules; the agent follows them.

Resolved by the agent

Handed over to a person

Never decided by the agent

Order status  Status from the order record  Conflicting or missing data  Whether it was mishandled
Delivery  Shipping status and tracking  Delays, lost parcels, repeated failures  Compensation
Product or service question  Answers from your knowledge base  Questions the documentation doesn't cover  Safety or legal claims
Returns and refunds  Explaining the standard process  Any refund request  Exceptional refunds or discounts
Complaint  Acknowledging it and collecting context  The complaint itself  Whether the complaint is valid
Support ticket  Reviewing history, suggesting a reply  Urgent or high-impact incidents  Closing a serious incident

For the reasoning behind each column, read what AI can and cannot solve in customer service.

DATA AND CONTROL

What the agent sees, and what happens when it doesn't know.

  •  Sees: orders, deliveries, tickets, past messages and your knowledge base
  •  Doesn't see: anything your team's role can't access
  •  When it doesn't know: no guessing, the case goes to your team
  •  You configure: tone, common phrases and escalation rules
AI agent replying to a customer with order and delivery context
Support team reviewing a customer conversation with order context

LIVE CHAT

Reply with real customer context.

Before replying, the agent checks order history, past chats and account details.

  •  Order and delivery context
  •  Previous conversation history
  •  Tone aligned with your brand

TICKETS AND HANDOVERS

Faster tickets, cleaner handovers.

The agent reviews the order and past tickets before suggesting a reply, and summarises long threads for whoever takes over.

  •  Order and ticket history
  •  Key points and action items extracted
  •  Escalation when needed
AI agent summarising a long email thread for the support team

CHANNELS

Website chat, email and tickets. And WhatsApp?

The agent works where your support already happens: website chat, email and helpdesk tickets, with the same rules in each. WhatsApp has its own limits, such as the 24-hour reply window.

How an AI agent on WhatsApp works

PRICING

What it costs.

Priced per workspace, not per user. Every plan includes all apps and AI credits.

Free Platform

€0/mo
  • 1 user
  • Access to all apps
  • 1 AI credit /mo
Most popular

Professional platform

from €59/mo
  • 1 to 19 users
  • Access to all apps
  • 5 AI credits /mo

Growth platform

from €159/mo
  • 20 to unlimited users
  • Access to all apps
  • 15 AI credits /mo
OTHER AREAS

The same agents, across your business.

Sales app icon

The agent in the CRM

Follow-ups, summaries and next steps for your sales team.

Accounting app icon

Agents for accounting

Invoices read and entries drafted for approval.

Helpdesk app icon

Process automation

Repetitive routines run with approval where it matters.

RELATED GUIDES

Before you automate your support.

Limits of AI in customer service

What AI can and cannot solve in customer service

What to automate, what to escalate and how to measure it.

Read the guide

AI agent on WhatsApp

How it works, what it needs and the limits of WhatsApp Business.

Read the guide

Virtual assistant use cases for businesses

AI virtual assistants: real use cases

Use cases by department, and how an assistant differs from an agent.

Read the guide

George, the Binhex AI agent for customer conversations
CUSTOMER SERVICE AI · FAQ

Questions you're probably asking

Routine requests that can be answered from your records: order status, delivery information, purchase history and standard product or service questions. Anything that involves a refund, a discount, a complaint or unclear data goes to a person.

When the records do not answer the question, when the data conflicts, when the customer asks for money, an exception or a legal decision, or when the conversation needs judgement. The handover includes the context, so the customer does not have to start again, and it is logged.

Only what the user or team it works for is allowed to see in the ERP: orders, deliveries, purchase history, previous conversations, tickets and your knowledge base. Support sees support data, not finance.

It should not guess. It collects the context of the request and passes it to your team, following the escalation rules you configure.

No. A typical chatbot answers from a fixed script or FAQ. An AI agent reads the customer's real records in your ERP, so it can answer about a specific order, and it knows which requests it must hand over to a person.

Yes. You can configure tone, common phrases and escalation rules, using your brand guidelines, templates and knowledge base.

Binhex Cloud plans are priced per workspace, not per user, and every plan includes all apps and AI credits. The Free plan costs €0 per month with 1 user and 1 AI credit. Professional starts from €59 per month billed annually, for 1 to 19 users and 5 AI credits. Growth starts from €159 per month billed annually, for 20 or more users and 15 AI credits.

No. Binhex Copilot adds Emma and George to your existing Odoo database without migrating data. Setup depends on your current Odoo, so we review it and scope the work before quoting.

START WITH THE ROUTINE

Automate the routine. Keep the important conversations human.

Create a free workspace, connect your orders and see what the agent can answer on its own.

Create a free workspaceTalk to a specialist

No credit card. Last updated: September 2026.