dConcept/ hk

Practical notes on using AI at work

Short, specific writing for people trying to make AI useful in real workflows.

Have a task in mind?

These notes are about real workflows. If something here maps to your work and you want help applying it, a session is the sensible next step.

Apply for a session

Filter by topic

Newest first. Pick a topic to narrow down.

9 of 9

Systems13 min read
Diagram showing orchestrator, worker, validators and structured handoffs in a long-running multi-agent architecture

The multi-agent architecture that actually ships

Everyone is building multi-agent systems. Almost nobody agrees on how. This is the production breakdown that matters: the roles, the contracts and the discipline that let agent teams run for days without drifting.

IndustrySystems8 min read
Diagram mapping AWS Bedrock AgentCore components (Managed Harness, Memory, Gateway, Skills) to the four fundamental building blocks of AI workflows (Agent, Knowledge, Connector, Skill)

Behind the AWS Bedrock AgentCore launch

Tech giants are selling heavy, complex infrastructure, while business operators just need simple, reliable workflows.

SystemsWorkflows11 min read
Workflow map showing Orchestrator, Source Ladder, Specialists, and Evidence Log in dConcept retro Riso style

How to build a defensively scoped proposal agent

The true value of AI in a sales process is not speed or persuasion. It is commercial safety.

Teams12 min read
Workflow map showing how a team moves from process mapping to shared Codex agents and skills

How we onboarded a team to Codex

The useful part of Codex onboarding is not showing people a chat box. It is turning repeated work into shared, reviewable workflows.

TeamsIndustry10 min read
Diagram showing an embedded AI accelerator working with a cohort, building tools and moving people toward self-sufficiency

The rise of embedded AI roles

Stripe is not hiring another AI trainer. It is hiring embedded operators to change how marketing teams work.

Workflows14 min read
Workflow map showing trigger, inputs, decisions, outputs and drag before choosing an AI tool

Start with workflows, not AI tools

The fastest way to make AI useful at work is to look at what you already do every week, then pick the smallest tool that fits.

Safe use15 min read
Sensitive notes being redacted into a safe AI prompt before use

What not to paste into AI tools

Treat the prompt box like a shared surface. If you would not put it in a generic web form, do not paste it into a public model without safeguards.

Workflows11 min read
Two-path decision map comparing clarity sessions and build-with-me sessions

AI clarity session or build-with-me session?

If you cannot yet name the work, start with clarity. If you can point at the artefact and the friction, start with building.

Teams13 min read
Small team AI adoption loop moving from shared language to one pilot, review and reuse

How small teams should start with AI

Teams win when language, norms, and one visible success line up. Before procurement adds another dozen seats.