Agent systems
Agents that take real work off your team's plate, with the judgement calls routed back to people.
- +Tool and API integration
- +Multi-step orchestration
- +Escalation and handoff
We design and build agentic systems, automations and the products around them.
hello@neunek.aiSee the work →UGC Desk, Hermes, Datum.ai, Majlis and Barjeel are ours. The last one is a placeholder.
Turns a product or an idea into a story, a set of hooks and a shot list, then generates the video variants and keeps a record of which ones actually performed.
A self-hosted chief-of-staff agent with layered memory, so it still has the context weeks later. Reachable from chat, and it asks before it acts on anything outbound.
A common operating picture built on open sensor feeds — aircraft transponders, ship beacons, orbital tracks and seismic data — fused into one live map with alerting on top.
A members' dining club where the seating is computed, not guessed: a weighted engine scores every candidate table on who has already met, gender balance, age spread and professional mix, and returns an arrangement reproducible from a seed.
A UAE property-intelligence dashboard. It benchmarks live listings against registered government sales, estimates the rent, yield and payback on every unit, and flags price drops between scans.
A nightly pipeline that reconciles records across three systems and flags only what needs a human.
Defense, security and critical operations are where agentic systems get interesting — and where a confident wrong answer is the whole problem. We design for the operator: every call traceable, every automated action reversible, and a human on the trigger.
Reference build: Datum.ai — see the work above.
Six capabilities, mixed to fit the job. We plug into your team rather than replacing it.
Agents that take real work off your team's plate, with the judgement calls routed back to people.
The unglamorous pipelines that run every night and quietly save a department a day a week.
The interface around the model: where it helps, where it defers, how it admits it is unsure.
Agents get graded before they ship and watched after, so quality is a number, not a vibe.
A working thing in days, so the argument is about something real instead of a slide.
Playbooks, prompts and components handed over so your team runs it without us.
Every engagement runs the same five beats, on a loop. Each pass sharpens the playbooks the next one starts from.