Expertise · Agentic AI

AI agents that get tasks done reliably

An agent doesn't just answer questions; it takes actions in your systems. That raises the upside, and the risk. The developers in our network build agents with clear boundaries, so you can trust them.

How we approach agents

  • Start narrow. One well-defined process with clear rules and a measurable outcome. A workflow with a few smart steps is often more reliable than a fully autonomous agent.
  • Least privilege. An agent gets only the tools and data it needs, and irreversible actions go past a human.
  • Everything traceable. Every step, tool call and decision is logged, so you can explain what happened.
  • Evaluate the whole trajectory. Not just the final answer, but whether the agent took the right route at a reasonable cost.

Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, due to escalating costs, unclear business value and inadequate risk controls. Source: SD Times on Gartner, 30 June 2025 (opens in a new tab)

Those are exactly the areas where a good engineer makes the difference.

What that looks like in code
agent/refunds.pyPython 3.12
from decimal import Decimal
from typing import Literal
from pydantic import BaseModel

REFUND_LIMIT = Decimal("250.00")

# the model picks from fixed actions, with a typed answer
class Decision(BaseModel):
    action: Literal["answer", "refund", "escalate"]
    amount: Decimal | None = None
    sources: list[str]

async def handle(ticket: Ticket) -> Decision:
    context = await retriever.search(ticket.text, k=5)
    decision = await llm.parse(
        build_prompt(ticket, context), output=Decision
    )
    if decision.amount and decision.amount > REFUND_LIMIT:
        return await escalate(ticket, decision)  # above the limit, a person decides
    await audit_log.write(ticket.id, decision)  # every decision traceable, with its sources
    return decision

Request profiles

Profiles of suitable AI agent developers within 48 hours

  • No obligation: you only pay for hours worked
  • Dutch-speaking, from our own network and assessed by us
  • On-site, hybrid or remote

Indicative € 106–132 per hour, all-in, excl. VAT

Rather call? +31 35 7111 373

What we select for

  1. 01

    Thinks in risks

    Starts by asking what happens when the agent gets it wrong, and designs the controls around that.

  2. 02

    Software engineer first

    An agent is software. Tests, error handling, versioning and observability come with it.

  3. 03

    Measures value

    Shows with data how much time or money the agent saves, and what it costs.

Technology we often see

Tools change fast. We select on principles, but this is what the developers in our network use in practice.

PythonLangGraphOpenAI Agents SDKModel Context ProtocolPydantic AIFastAPILangfusePostgreSQL

Profiles we are looking for

Need a Python developer for this?

Send us your vacancy or request. Within 48 hours you hear which developers in our network fit and what it costs.