# Enterprise Chatbot: What It Needs and How to Roll One Out

> What an enterprise AI chatbot needs: single sign-on, roles, audit trails, data residency, approvals and system integrations, plus a phased rollout plan.

Leonard Loo · CEO & Co-Founder, Volty

Updated Nov 21, 2026

7 min read

An enterprise chatbot is an AI chatbot built for a large organisation: it answers customers or employees at volume across several channels, works inside the company's own systems, and runs under the controls a security team expects, such as single sign-on, permissions, audit trails and clear rules on where data lives. Governance, integrations and the handover to people are what separate it from a small website bot.

This guide covers what an enterprise chatbot has to do, the questions procurement and security teams will ask, how these tools are priced, and a rollout plan that keeps risk low. It uses Volty, the AI helpdesk we build, as the worked example, with its real settings and published security posture.

## Small-business bot versus enterprise chatbot

The same AI model can sit behind both. The difference is the surrounding system, and it shows up in six places.

| Area | Typical small-business bot | Enterprise chatbot |
| --- | --- | --- |
| Channels | One website widget | Website, WhatsApp, Instagram, Messenger and email, with one record per customer |
| Knowledge | A list of FAQs | Policies, product data and documents from several teams, kept current |
| Actions | Answers only | Books, updates orders and records, with approval steps for sensitive actions |
| People | An email notification | Routing to the right team or branch, queues, reply targets and escalation |
| Access | One shared login | Single sign-on, directory provisioning, roles and row-level permissions |
| Evidence | Chat history | Audit trail, exportable logs and a written data-protection position |

## What the chatbot has to do

### Answer from company knowledge

Large organisations rarely have one tidy FAQ. Prices sit with finance, policies with operations, and branch details with each location. The chatbot should answer in plain language from your website and documents, including questions nobody wrote an answer for, and it should say so and pass the question on when the answer is missing. In Volty you point the agent at your website and documents, and when a teammate corrects a reply, the lesson is kept for everyone, with each lesson visible, attributed to whoever approved it, and reversible.

### Act in your systems, with limits

An answer is half the job if the customer then has to fill in a form. An enterprise chatbot should book the slot, update the order or create the lead in the system that owns it. Volty can take any action a connected system allows, and HubSpot, Google Calendar, Notion, Shopify and Slack are built in, with other systems connected through their APIs. Staff decide which actions happen on their own and which come to a person first: when a customer asks for a refund, Volty asks the right person, who answers in one tap.

### Hand over cleanly

Some conversations need a person, such as an upset customer, a sensitive question or a large order. The bot should route to the right team by topic, tell the customer who is taking over, and give that person the whole history. Volty labels each conversation as it arrives, routes it by your rules, shares out the rest in turn with a limit per person, and alerts managers when a chat waits past its first-reply target.

### Serve staff as well as customers

Many enterprise chatbots face inward. At Yong Kang TCM Clinic, part of a group whose brands run 19 outlets across Singapore, Malaysia and Indonesia, Volty routes each patient inquiry to the right branch, physician and slot, and an internal portal in Chinese gives physicians quick answers on policies, prices and more. At EtonHouse, Volty drafts lesson plans from learning objectives, topic, materials, duration and location, and planning that took teachers a full day in Excel now takes an hour.

EtonHouse. Read the case study

Yong Kang TCM Clinic. Read the case study

## The questions security and procurement will ask

Expect a questionnaire before any contract. These are the questions that come up most, with the answers Volty publishes, so you can see what a complete answer looks like.

| Question | What a good answer covers | Volty's published answer |
| --- | --- | --- |
| Where is our data stored? | Region for the database, files and backups | Database, files and backups are stored and backed up in Singapore on AWS |
| Where does AI processing happen? | Which model hosts see conversation content, and their retention | The content a reply needs goes through OpenRouter to a model host, mostly in the United States, under zero data retention |
| Is our data used to train models? | A written commitment | No. Customer data is not used to train or fine-tune AI models |
| How is data encrypted? | In transit and at rest | TLS 1.2 and 1.3 in transit, AES-256 at rest |
| How do our people sign in? | Single sign-on and provisioning | Single sign-on from Volty Max; user provisioning from the company directory on Enterprise |
| Who can see what? | Roles and record-level rules | Ready-made roles on every plan, custom roles from Volty Pro, row rules so each person sees only their rows from Volty Max |
| Can we prove what happened? | Audit trail and export | Every agent run is recorded in an event sequence; audit export from Volty Max |
| What certifications do you hold? | Named certificates and who issued them | CASA TAC Security Tier 2 and the Cyber Essentials Mark from CSA |
| What uptime do you commit to? | A target and a remedy | A 99.5% monthly uptime target, with a 10% service credit if missed |

The full detail is on the Trust Centre and the AI governance page. Whatever vendor you assess, ask for the same items in writing, and treat a vague answer on AI processing or training as a red flag.

## How enterprise chatbots are priced

There are four common models, and each one rewards a different usage pattern.

- Per seat. You pay for each person who works in the tool. Predictable, but the bill grows with the team even when the AI does most of the work.
- Per resolution or outcome. You pay each time the AI closes a conversation. Cheap at low volume, harder to forecast once the AI handles most traffic.
- Per plan with allowances. A monthly price includes seats, contacts and AI usage, with pay-as-you-go capacity on top. This is how Volty prices.
- Custom contract. Pooled usage, invoicing and terms negotiated for the organisation.

In Volty's US list, Volty Max is US$799 a month with 40 seats, 5,000 monthly active contacts and 60,000 AI credits, and it adds row rules, an approve right, audit export, single sign-on and testing tools for developers. Volty's pricing page puts that at about 3,000 to 7,500 AI agent replies a month, and extra capacity is pay as you go. The Enterprise plan adds user provisioning, two-person approval, one pool of usage across workspaces, your own AI gateway with zero-retention AI providers, custom contract and invoicing, and a dedicated account team. Current figures are on the pricing page, and they can change.

## A rollout plan that keeps risk low

- Pick one channel and one team. The busiest WhatsApp line or the website chat for one product is enough. Write down today's median first reply, after-hours backlog and share of conversations that end in a booking or sale.
- Load the knowledge and connect the systems the first use case needs, and nothing more. Fewer connections mean fewer permissions to review.
- Start in draft mode. The AI writes every reply and staff send, edit or reject it. Corrections become lessons for the whole team.
- Allow automatic replies where the drafts are reliably right, inbox by inbox. Volty can also be set to answer on its own at night and draft during office hours.
- Put approval steps on sensitive actions such as refunds, discounts and record changes, and decide who approves.
- Review after four weeks against the baseline, then add the next channel or team.

Cutis Medical Laser Clinics runs this mix of automatic and approved replies today: about half of its patient messages go out automatically and half are approved by staff in one click, and nearly all are sent as the agent wrote them. Its average first reply is 24 seconds. For more on the safety side of the same rollout, read our guide to rolling out AI customer service, and for the difference between a scripted bot and an agent that acts, see AI agent vs chatbot.

Cutis Medical Laser Clinics. Read the case study

## Running an enterprise chatbot on Volty

Volty gives you one AI agent across WhatsApp, Instagram, Messenger, email and web chat, with every conversation in a shared inbox. On your website, you paste one script tag, list the domains allowed to show the chat, and the same agent answers there with cards for products, time slots and reply buttons. The chat can also report events such as a started conversation to your website and to Google Analytics. Multilingual traffic works the same way: at Yat Guan, Volty answers buyers in English, Mandarin and Malay, 24/7, and its director, Ryan Lim, said: "Inquiry to quotation now only takes one working day - 60% faster than before. Operating at over 95% accuracy in English, Mandarin, and Malay."

See how the website agent works on the Volty Website & App AI page, then start a free trial on one channel to test it against your own traffic.

## Frequently asked questions

What is an enterprise chatbot?

An enterprise chatbot is an AI chatbot built for a large organisation. It answers customers or employees at volume across several channels, works with the company's own systems, and runs under controls such as single sign-on, role-based permissions, audit trails and rules on where data is stored.

How is an enterprise AI chatbot different from a regular chatbot?

A regular chatbot usually answers questions on one website from a fixed set of answers. An enterprise AI chatbot answers in natural language from company knowledge, takes actions in connected systems, hands conversations to the right team, and gives administrators control over who can see and approve what.

What security features should an enterprise chatbot have?

Look for single sign-on, user provisioning from your company directory, role-based access with row-level rules, an exportable audit trail, approval steps before sensitive actions, encryption in transit and at rest, a written answer on where data is stored, and a commitment that your conversations are not used to train AI models.

How much does an enterprise chatbot cost?

Enterprise chatbots are priced per seat, per resolved conversation, per plan with usage allowances, or by custom contract. In Volty's US price list, Volty Max is US$799 a month and includes single sign-on, audit export and row rules, while the Enterprise plan is priced on custom terms.

How long does it take to roll out an enterprise chatbot?

A focused first channel can go live in weeks when the knowledge and the system connections are ready. Most large organisations run a pilot with the AI drafting replies for staff to approve, then allow automatic replies topic by topic once the drafts are reliably right.

## Recommended for you

GoHighLevel Alternatives: 6 CRM and Messaging Options · Gorgias Alternatives: 6 Help Desks Compared on Price and AI · Instagram Automation Tools: 7 Options Compared for 2026 · Interakt Alternatives: 6 WhatsApp Platforms Compared · Interakt Pricing 2026: Plans, Add-ons and Message Fees · Intercom Alternatives and Fin Alternatives: 7 Options · Cliniko WhatsApp integration: what a physio or allied health clinic can automate · Facebook Messenger Auto Reply: Set It Up and Do It Well · How to Get WhatsApp API Access: Step-by-Step (2026)

## Start your Sandbox Trial

Email us questions · Start your Sandbox Trial

Sign up for our 7-day trial. No credit card needed.

Future-proof your business with AI. Hear from our CEO

The Leonard and Volty Show · October 8, 2026

Meta now charges for WhatsApp replies

The Leonard and Volty Show · October 1, 2026

Too busy to hire?

The Leonard and Volty Show · September 21, 2026

From $500 chatbot jobs to agentic AI

The Leonard and Volty Show · September 10, 2026

Your AI is a new employee

The Leonard and Volty Show · September 4, 2026

Trust in the AI world

The Leonard and Volty Show · August 29, 2026

Does the AI model matter?

The Leonard and Volty Show · August 19, 2026

What are AI agents?

Meta now charges for WhatsApp replies. Here's what it means for your business, October 8, 2026

## Links

- [Yong Kang TCM Clinic](https://voltade.com/case-studies/yong-kang-tcm)
- [EtonHouse](https://voltade.com/case-studies/etonhouse)
- [Trust Centre](https://voltade.com/security)
- [AI governance](https://voltade.com/ai-governance)
- [pricing page](https://voltade.com/pricing)
- [Cutis Medical Laser Clinics](https://voltade.com/case-studies/cutis)
- [rolling out AI customer service](https://voltade.com/blog/ai-customer-service)
- [AI agent vs chatbot](https://voltade.com/blog/ai-agent-vs-chatbot)
- [Yat Guan](https://voltade.com/case-studies/yat-guan)
- [Volty Website & App AI](https://voltade.com/product/website-app-ai)
- [GoHighLevel Alternatives: 6 CRM and Messaging Options](https://voltade.com/blog/gohighlevel-alternatives)
- [Gorgias Alternatives: 6 Help Desks Compared on Price and AI](https://voltade.com/blog/gorgias-alternatives)
- [Instagram Automation Tools: 7 Options Compared for 2026](https://voltade.com/blog/instagram-automation-tools)
- [Interakt Alternatives: 6 WhatsApp Platforms Compared](https://voltade.com/blog/interakt-alternatives)
- [Interakt Pricing 2026: Plans, Add-ons and Message Fees](https://voltade.com/blog/interakt-pricing)
- [Intercom Alternatives and Fin Alternatives: 7 Options](https://voltade.com/blog/intercom-alternatives)
- [Cliniko WhatsApp integration: what a physio or allied health clinic can automate](https://voltade.com/blog/cliniko-whatsapp-integration)
- [Facebook Messenger Auto Reply: Set It Up and Do It Well](https://voltade.com/blog/facebook-messenger-auto-reply)
- [How to Get WhatsApp API Access: Step-by-Step (2026)](https://voltade.com/blog/how-to-get-whatsapp-api)
