AI Forwarding

AI integration for companies: hosted models or your own, built into the systems you already run.

AI Forwarding is how we bring AI into a company: we take the model that fits your data and put it to work inside your ERP, ticketing, documents and databases. That can be a hosted model over an API, an open-weight model on your own servers, or both. Underneath sits the engineering we have done for 20 years in cloud and cyber security: identity, access boundaries, monitoring and cost control.

Talk to us about your AI integration

When companies call us

  • Staff already paste company data into public chatbots. You want a sanctioned way to use AI, with clear rules for what may leave the network. (Why shared AI chats are a risk.)
  • A pilot worked in a demo and never reached production. Nobody owned identity, permissions, monitoring and cost.
  • The data may not leave your network. Regulation, contracts or internal policy rule out an external API, so the model has to run where the data is.
  • People spend hours asking systems questions. ERP, quality management, tickets and documentation hold the answers, but getting them out takes a specialist.

What we build

Public AI Integrations

Anthropic Claude, OpenAI GPT and Google Gemini, over their own APIs or through Amazon Bedrock, Azure OpenAI and Vertex AI. Claude and OpenAI integration →

Private AI Integrations

Open-weight models such as Qwen, Llama and Mistral on your own servers, on-premises or in your own cloud account. Private AI and local LLMs →

Assistants and agents

Persistent agents on AI Central, our own multi-agent platform, or on your stack: in internal portals, Microsoft Teams, Slack, WhatsApp and Telegram. They answer from live ERP and production data, read your documentation and triage tickets. AI Consulting & Agents →

Identity and access

Everything sits behind your SSO, with read-only access and least-privilege credentials.

Data boundaries

We define what may leave your network and what may not, personal data included, and enforce it in the integration itself.

Operations and cost

Monitoring, usage limits that keep cost predictable, updates and capacity. We run the result afterwards with the same operations discipline as everything else we do.

Public or private model?

There are two ways to bring a model into your company, and many companies use both. The same engineering sits underneath either way.

How a project runs

1 · Workshop

A structured discovery with your business and IT leads: use cases ranked by value and feasibility, data and system readiness, model and hosting options, and a cost model for development, tokens, infrastructure and operations.

2 · Roadmap and offer

You leave with a prioritized roadmap and a fixed-price offer for the first agent.

3 · Build

We connect the model to the systems the use case needs, behind your SSO and inside the data boundaries agreed in the workshop.

4 · Run

We operate what we build: monitoring, updates, capacity and cost.

Customer projects

  • Paysafe: AI Forwarder projects and an Anthropic integration for a global payments platform (reference)
  • Presimsan: local AI for a manufacturer (reference)
  • Harput Holding: a family of assistants over ERP, quality-management and business data (reference)

Frequently asked questions

What does AI integration mean in practice?

It means connecting a language model to the systems and data you already have, so that people can ask questions, get drafts and hand over routine tasks inside the tools they already use. An integration is the model plus the connections to ERP, ticketing, documents and databases, plus identity, permissions, monitoring and cost limits.

Which AI models do you integrate?

Hosted models: Anthropic Claude, OpenAI GPT and Google Gemini, over their own APIs or through Amazon Bedrock, Azure OpenAI and Vertex AI. Local models: open-weight models such as Qwen, Llama and Mistral, running on your own servers.

Does our data leave our network?

That depends on the route. With a hosted model, prompts are processed by a cloud provider under contract, and we define what may be sent. With a private model, prompts, documents and answers stay inside your network.

Can we use hosted and local models together?

Yes, and many companies do: a hosted model where capability matters most, a local model where the data must stay inside. Both are connected to your systems the same way, with the same identity and monitoring.

How do we start?

With a workshop with your business and IT leads. You leave with a prioritized roadmap and a fixed-price offer for the first agent.

Further reading

Talk to us about your AI integration