Custom AI agents. Built for your business.
Assistants for support, sales and internal knowledge that answer from your own sources, take action in your systems and know when to hand over to a person.
- Typical timeline
- 6–16 weeks
- Starts with
- Discovery Sprint or Pilot
- Channels
- Web, email, WhatsApp, Teams
- Languages
- English, Turkish and more
The same questions, answered a thousand times.
Queues grow faster than teams
Most requests are repetitive, yet each one waits for a person. Response times slip and good people burn out.
Answers live in ten places
Policies, product sheets, past tickets and a colleague's memory. Finding the right answer takes longer than giving it.
Chatbots that only chat
A bot that cannot look up an order or update a record just adds a step before the human.
Agents that answer, act and escalate.
We design the agent around a job, not a model. Every agent ships with evaluation, guardrails and a human handover path.
- 01
Customer support agent
Handles order, billing and returns requests across chat and email, and passes the rest to your team with the full context.
- 02
Knowledge assistant (RAG)
Answers from your documents and data with the source shown, for customers or for employees.
- 03
Sales assistant
Qualifies inbound leads, drafts follow-ups and keeps CRM records up to date.
- 04
Voice assistant
Takes calls for routine requests such as bookings and order status, and transfers to a person when needed.
- 05
Connected to your systems
Reads and writes to SAP, Salesforce, HubSpot and your own APIs with scoped permissions.
- 06
Tested and safe by design
An evaluation set built from your real conversations runs on every change. Guardrails, audit logs and a clear handover come as standard.
Three agents to start with.
Support agent
Resolves order, billing and returns requests across email and chat, in the customer's language.
- Order lookup
- Refunds
- Handover
Sales assistant
Qualifies inbound leads, drafts follow-ups and keeps the CRM up to date after every meeting.
- Lead scoring
- CRM notes
Internal knowledge assistant
Answers employee questions from policies, manuals and wikis, with a link to the source.
- Policies
- Onboarding
From first call to live agent.
- 01 · 2 weeks
Discovery
Map the conversations, data sources and the actions the agent must take. Fixed price.
- 02 · 4–6 weeks
Pilot
A working agent on real data with a small group of users, measured against an agreed target. Fixed price.
- 03 · 8–16 weeks
Build
Full integrations, evaluation suite, security review and a staged rollout.
- 04 · Ongoing
Run
Monitoring, monthly tuning and SLA-backed support.
Model-agnostic. Your cloud or ours.
We choose the model for the job and keep the option to switch. Deployments can run in an EU region or inside your own cloud account.
Models
- OpenAI
- Anthropic
- Azure OpenAI
- Google Gemini
- Open-weight models
Systems
- SAP
- Salesforce
- HubSpot
- Zendesk
- Microsoft Teams
- WhatsApp Business
Data
- PostgreSQL
- Vector search
- SharePoint
- Confluence
About custom agents.
How do you stop the agent from making things up?
Answers are grounded in your approved sources and shown with citations. When the agent is not confident, it says so and hands over to a person. We measure accuracy on a test set before every release.
Which channels can the agent work in?
Web chat, email, WhatsApp, Microsoft Teams and Slack. Voice can be added as a separate workstream.
Who owns the agent and the data?
You do. The configuration, prompts, evaluation sets and data belong to you and can run in your own cloud account.
What does it cost to run?
Running costs depend on volume and model choice. We estimate them during the Pilot and report them to you every month.
Will it replace our support team?
No. It takes the repetitive requests so your team can spend its time on the cases that need judgement. Every agent has a handover path to a person.
Have a queue that keeps growing? Let's fix it.
Tell us about the requests your team handles. We will show you what an agent could take on.
Or write to us: hello@nbmcore.com