# What is forward-deployed AI engineering?

Forward-deployed AI engineering usually means engineers work close to a customer’s operation to shape AI around real data, systems, permissions, and workflows. The work is more specific than handing over a generic model. It starts with how a team receives an invoice, handles an order, dispatches a technician, follows up an open item, or approves a report, then connects that process to the right sources and review boundary.

## What the term normally covers

A forward-deployed team might map the current process, identify the records and owners involved, configure a workflow, test it against real exceptions, and help the customer decide what can be prepared versus what must remain human-owned. The exact staffing model, deliverables, duration, and responsibility vary by company. The term alone does not prove that an engineer is embedded on site, that a service includes ongoing operations, or that the vendor owns the customer’s decisions.

## How Queli works with customers

Queli uses the same practical principle: start close to the operation and configure Makra around the customer’s systems, permissions, data boundaries, sources, owners, and workflow rules. The work begins with a recurring process and follows it from arrival to decision, including the awkward exceptions that generic software usually leaves to people.

That customer-specific approach can begin with the work surface already in use. A process may start with email, WhatsApp, a PDF, an Excel file, an ERP, a CRM, an accounting system, or a shared folder. Makra can prepare a structured record, task, draft, classification, reminder, report, or exception queue, then leave consequential approval with the authorized person. Connection method, model provider, region, retention, and customer-hosted options are confirmed per workload before production.

## What a useful engagement produces

The result should be more than a workshop or model demo. A useful engagement produces a mapped process, agreed sources and permissions, a working queue or application, clear review roles, an exception path, and a measurable baseline. It also establishes which system remains the official record and what Makra may read, prepare, or write.

This approach is visible in Queli’s invoice and Intrastat work. Supplier-specific rules, multilingual documents, ERP fields, classification decisions, and finance review were shaped into one operating flow. The value came from understanding the real process and building around it, not from dropping a generic assistant into the company.

For a new customer, the first step is a focused scope conversation: which process consumes time, where the work arrives, who decides, which systems matter, and what result would prove the workflow is better. Queli can then define the roles, systems, permissions, deliverables, and ownership for that implementation.

## Sources

- https://www.queli.ai/
- https://www.queli.ai/llms.txt

## Other languages

- [English](https://www.queli.ai/ai-agents/answers/forward-deployed-ai-engineering)
- [Hrvatski](https://www.queli.ai/hr/ai-agenti/answers/inzenjer-za-uvodenje-ai-rjesenja)
- [Deutsch](https://www.queli.ai/de/ki-agenten/answers/kundennahe-ki-implementierung)