Dittology/studio

> cat ./services/ai-agents.md / hong kong

AI agents that do real work.

Dittology designs and builds custom AI agents for Hong Kong businesses: assistants that answer from your own data, automations that move work between your tools, and the console to run them. AI agent development from a design and AI engineering studio in Hong Kong.

What we build.

01 / ls ./agents
/01Customer-facing assistantsAnswer from your documents and policies, and hand off to a human when unsure.
/02Internal copilotsDrafting, triage, and retrieval inside the tools your team already lives in.
/03Workflow automationsAgents that move work between systems and log every step they take.
/04Agent consolesA dashboard to watch, correct, and measure what the agent does.

Kept honest.

02 / cat ./guardrails
/01GroundedAnswers come from your data with sources attached, not from the model's imagination.
/02ScopedAn agent gets a job description and hard limits, not general intelligence claims.
/03MeasuredLogs and evals from day one, so you can see what it did and why.

Built this way.

03 / ls ./work | grep ai
AI Creative Engine previewAI Creative Engine AIGIS · 2026→Hairstyle Booking Marketplace previewHairstyle Booking Marketplace Good Looks · 2026→

Questions, answered.

04 / cat ./faq
/01

What can an AI agent actually do for a business in Hong Kong?

The reliable wins are narrow and boring on purpose: answering repeat customer questions from your own documents, triaging and drafting replies, moving data between systems that never talked to each other, and first-pass work a person then approves.

/02

How do you stop it from making things up?

Grounding and limits. The agent answers from your data with sources attached, declines what it cannot support, and escalates to a person instead of guessing. We build the eval set alongside the agent so you can measure it, not take our word.

/03

Can it plug into our existing systems?

Yes. Agents earn their keep by fitting the stack you already run, whether that is a proper API, a CRM, spreadsheets, or email. Integration scope is settled in the first call.

/04

What does an AI agent cost to build and run?

Two numbers, quoted separately: a fixed build cost sized to scope, and a monthly running cost for models and hosting that we estimate before anything is built. No surprise meters.

/05

How long until it's live?

A working pilot on real data comes first, usually within weeks; production hardening depends on how many systems it touches. We stage it so you see value before the full build.

Scope an agent →