Get Started

Custom AI Agent and Helpdesk

Build your own AI agents designed around your business data, industry rules, existing systems, and customer channels.

$0
seat fees, for your whole team
$0
per-answer fees, at any volume
100%
Ownership of Data and System
Request sent. We are on it.

Free, no commitmentFrom your public siteReply in 1 business day

Built on the models and tools you already use

OpenAIAnthropicGeminiLlamaMistralAWSAzureGoogle CloudWhatsAppSlackShopifyHubSpotSalesforceZendeskStripe

It answers, acts and hands over.

Standard in every AI agent

Set your own limits and rules
  • Answers only from approved sources. Hands over when they do not cover it.
  • Quotes prices, dates and policies from your systems, not from memory.
  • Keeps card numbers and passwords out of every reply.
  • Hands over whenever a customer asks for a person.

Sample conversation, workflow, inbox and dashboard. Every action, limit and approval route is defined with you during the build.

One agent, on every channel your customers and your team already use.

One knowledge base, one set of rules and one inbox for every conversation. Already on Zendesk, Intercom, Gorgias or Freshdesk? Keep it: the agent files what it should not answer as a normal ticket.

Sample conversations, each in the channel’s own layout.

Owned by you. Built to your compliance.

You choose where it runs. Your data stays under your control, and the agent is built around the security rules your team already follows.

You choose where it runs

Wherever it runs, the model account, the keys and the conversation history are yours.

  • Our cloudLive fastest, nothing for IT to set up
    Hosting
    BrillMark infrastructure
    Operations
    Managed by BrillMark
  • Your cloudInside your AWS, Azure or Google Cloud
    Hosting
    Your cloud account
    Operations
    Managed by BrillMark, inside your account
  • On-premise Scoped on requestYour own servers, behind your firewall
    Hosting
    Your own servers
    Operations
    Your IT, with BrillMark support
Frontier
ClaudeGPTGeminiMistralGrokCohere
Open-weight
LlamaDeepSeekQwenKimiMiniMaxGemmagpt-oss

Built to the controls you already follow

We build inside your environment, to your security team’s rules.

  • SOC 2
  • HIPAA
  • ISO 27001
  • GDPR

Certification needs such as SOC 2, HIPAA or ISO 27001 are scoped with you during discovery. We sign a DPA, and your data is never used to train models.

SSO and role-based access

Your team signs in through your identity provider. Roles decide who can view, approve and change what.

  • AdminRules and limits
  • ApproverActions above a limit
  • AgentReplies and handovers

An audit trail of every answer, action and approval

Each entry records what happened, who or what did it, and the sources it used.

Sample log

AnswerCited the returns policy, section 2AI agent
ActionRefunded $86, inside the $250 limitAI agent
HandoverExpress shipping above the $15 limitAI agent
ApprovalExpress shipping approvedSam K.

Real businesses. Measured on real outcomes.

Most AI vendors came from support. We came from conversion: the same team has shipped 45,000+ Experimentation for 200+ agencies and global brands, so every agent is measured on a completed journey.

Figures are each client’s own reported results for the journey the agent was built for.

Ecommerce · travel gear
+32%assisted conversion
-18%cart drop-off

An AI shopping assistant on their catalogue, shipping rules and approved knowledge.

Health and wellness
+61%FAQs resolved on time
+24%repeat purchase rate

An AI helpdesk answering order and product questions at any hour.

Free Prototype, Priced Once.

Start with a free prototype on your own site. Then a fixed quote, not a subscription. Model tokens bill to your own account at cost.

  1. Free prototype

    $0

    Usually within 1 business day

    Send your website. We build a working prototype agent on your public pages. No access, no install, no commitment.

    Build my prototype

  2. Discovery & solution design

    $999

    Two weeks, credited to the build

    We review your conversations and systems, then deliver a written architecture, workflow plan, scope and a fixed build quote.

  3. Custom build

    Quoted once

    Fixed after discovery

    We design, build, connect and test your agent and helpdesk in your chosen environment, in short review cycles.

  4. Launch

    6 to 10 weeks

    From approval, behind a switch you control

    Optional managed operations (monitoring, uptime, incidents) are quoted with the build. Improvements after launch are quoted when you ask.

Timings are typical. Discovery sets the real plan and the build quote.

What buyers ask us before they start.

Something else on your mind? Email info@brillmark.com or contact us.

What exactly is a custom AI agent?

An AI system built for one business that handles its customer support and sales conversations, from that business’s own knowledge, rules, tone and connected systems, under its own model keys. Unlike an off-the-shelf chatbot, its behaviour is defined during the build, not configured inside a vendor’s product.

Can you build around our exact workflow?

Yes, that is how every build starts. You describe the outcome, we scope what it takes in writing, and you get one fixed quote against that scope before anything is built.

Which models and providers can it use?

Closed-weight models such as Claude, GPT, Gemini and Mistral, and open-weight models such as DeepSeek, Kimi, MiniMax, Qwen and Llama. They can run on AWS Bedrock, Google Vertex AI, Azure AI Foundry, OpenRouter, direct provider APIs, or your own GPUs. You hold the account and keys, and models can be swapped without rebuilding the agent. We pick one by testing candidates on your own questions, the way we tested Jev on 100 support tickets.

Can it meet our security and compliance requirements?

Yes. It runs where you choose: our cloud, your own AWS, Azure or Google Cloud account, or, scoped on request, your own servers. The helpdesk has SSO, role-based access and a full audit trail. We build to the controls your team already follows, scope certification needs such as SOC 2, HIPAA or ISO 27001 during discovery, and sign a DPA.

How is this different from Intercom Fin or Zendesk AI?

Fin and Zendesk AI are products you configure inside their own helpdesk, usually billed per resolution. We build one agent around your business instead: your systems, rules, channels and model keys, running where you choose, with no per-answer fee. Already on Zendesk or Intercom? Keep it, and the agent hands over into it as a normal ticket.

Which channels does it work on?

Customers reach it on your website or app, WhatsApp, Telegram, Microsoft Teams, Slack and Discord. Messenger, Instagram DM, email, SMS and voice are scoped on request. One agent, one knowledge base, no per-channel licence.

What does it cost, and how long does it take?

Scope drives cost. Discovery and solution design is $999 flat and produces a written design plus a fixed build quote, credited toward the build. Scoping takes two weeks, most builds go live 6 to 10 weeks after that, and the prototype is free.

Will it make things up?

It answers only from knowledge you approve and systems you connect. When it cannot, it asks one of your people rather than guessing, and the customer sees a reply a human signed off.

Do I own what you build?

Yes. The build, prompts, workflows, integrations and data are yours, documented and handed over. It can also honour login, membership and entitlement rules, so gated knowledge stays gated.

When are you the wrong choice?

If you need it live next week, if you want the cheapest option, if your knowledge is one FAQ page, if nobody on your side will take the pings or approve what the agent may say, or if you want AI to replace your support team outright.

See it working on your own site.

Send your website. We build a working prototype agent on your public content and send it back, usually within 1 business day.

Request sent. We are on it.

Ask AI for a summary of this page