TC Tanvir Chowdhury / CTO Advisor · AI Engineering Leader
Melbourne, AUS · Big 4 Director, EY Australia

I design the operating systems for AI-native engineering.

AI-native SDLCs, agent harnesses, developer digital twins, multi-agent councils, token discipline. I partner with CTOs, CIOs and AI leaders to turn board-level AI ambition into governed delivery systems that hold up in production.

01 Run the AI maturity diagnostic 10 questions. Under 3 minutes. Get your maturity band and article recommendations to move from AI-enabled toward AI-native. Begin › 02 AI-Native Series A practitioner series on moving from AI-enabled teams to governed continuous execution. 03 CTO AI Playbook Executive operating patterns for harnesses, governance, org memory, and AI economics. 04 Engineer AI Playbook Field patterns for compiled context, specialist profiles, sandboxes, and handoffs.

// visual brief

AI agents field glossary

Full glossary ›

01

Agents

Like a teammate with a job, tools, and guardrails.

An agent is AI that works through a task in steps. It decides what to do next, uses approved tools, checks what happened, and stops when it can answer or needs human input.

The difference from chat is action: an agent can keep state and work through a loop.

Agents diagram An agent receives a goal, uses context and tools, observes results, and continues until it can answer or hand off. Goal what to achieve Agent reads context chooses action observes result Context Tools Answer Goal plus context plus action loop

02

Tools

Like the approved apps and instruments a worker can use.

Tools are actions an agent is allowed to call: search, files, browser, code, database, calendar, or an internal API.

Tools create real effects, so they need permissions, limits, and logs.

Tools diagram An agent requests permissioned actions from tools, and tools return observations the agent can use. Agent asks to act Tool permissioned action Search Browser Files API Tools let the agent do things outside chat

03

Skills

Like a recipe card for repeatable work.

Skills are packaged instructions an agent loads for a job. They can include steps, examples, scripts, templates, and done checks.

They turn good practice into a reusable method, not a one-off prompt.

Skills diagram An agent loads a skill containing instructions, examples, scripts, and checks, then performs repeatable work. Agent loads Skill pack instructions examples scripts done checks Task A Task B Task C A reusable way to do a job

04

Model Context Protocol (MCP)

Like USB-C: one bus can connect to many different devices.

MCP gives an AI app one standard connector for outside systems: files, databases, browsers, business tools, and services.

It is the bus, not the safety policy. Access still needs rules and approval.

Model Context Protocol (MCP) diagram An AI app connects to one MCP bus, which can connect to many different systems like tools, files, and data. AI app one port MCP bus USB-C style standard Files device Tools device Data device One bus, many different devices

05

Single-Agent Architect

Like one accountable operator handling the whole job.

One agent plans, chooses tools, executes steps, checks results, and writes the answer. It is both planner and worker.

Strong for focused work. Weaker when the job needs separate experts or review.

Single-agent architect diagram A user request goes to one agent, which plans, uses tools, checks results, and replies. User request One agent plan act check Tools Memory Answer One agent designs and executes the work

06

Multi-Agent Architecture

Like a delivery team with specialists and a lead.

Multiple agents share the work. One may plan, another researches, another builds, another reviews, and another reports.

The hard part is coordination: ownership, handoffs, decisions, and quality gates.

Multi-agent architecture diagram A coordinator splits work across specialist agents, then gathers results for review and final output. Goal larger task Lead coordinates Researcher Builder Reviewer Report Specialists work together under coordination

07

Agent Councils

Like a review panel where each member has a defined lens.

Several agents inspect the same evidence from different angles: risk, architecture, cost, delivery, or quality.

A council should produce a decision, dissent if needed, and a next action.

Agent councils diagram An evidence pack is reviewed by specialist agents, then a council records verdict, dissent, and next action. Evidence same facts Risk agent Tech agent Cost agent Council verdict dissent Different lenses, one recorded decision

08

Memory

Like a governed team notebook and archive.

Memory stores useful context from past work: decisions, preferences, facts, artifacts, and lessons.

It needs rules for what gets saved, corrected, retrieved, and forgotten.

Memory diagram Work produces decisions and facts, selected information is written to governed memory, and later agents retrieve it. User intent Agent does work Decision fact or artifact Governed memory write rules - corrections - retrieval - forgetting Only useful context is saved for later

09

Retrieval Augmented Generation (RAG)

Like answering with an open browser and a marked-up reference pack.

RAG makes the AI look up fresh or private information first, then use those sources to answer.

Good RAG separates search from answer writing and shows which sources shaped the result.

Retrieval Augmented Generation (RAG) diagram A question triggers retrieval from the web and internal documents. The AI uses the returned evidence to write a grounded answer. Question what is asked Retrieve search first Web current pages Docs private sources Answer grounded Search the web and docs before writing
// Latest live writing · 1 showing live articles only