# The agentic manual. Agentic PDLC tutorial

> A free, method-agnostic tutorial for running software projects where an AI model does the work: the four-phase agentic PDLC (P0 Frame, P1 Design & Spec, P2 Build & Prove, P3 Run & Learn), how AWS AI-DLC, AIDD, the BMAD Method and spec-driven development fit onto it, and what each role does. By Akash Das. MIT licence.

Each lesson is also available as markdown at the same URL with index.md appended. The full text of every lesson is at https://akash-coded.github.io/aws-bedrock-agentcore-strands/llms-full.txt. Every diagram, with its caption, is listed at https://akash-coded.github.io/aws-bedrock-agentcore-strands/pictures/.

## Getting started

- [What Is the Agentic PDLC? The P0 to P3 Framework Explained](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-the-agentic-pdlc/): The agentic PDLC is a four-phase lifecycle (Frame, Design & Spec, Build & Prove, Run & Learn) for software in which an AI model does part of the work.
- [How This Tutorial Works, and Why Every Lesson Looks the Same](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/how-this-tutorial-works/): Every lesson opens with the answer, shows it as a picture, lets you try it, and credits its sources. Here is why, with the research behind each choice.

## Agentic PDLC fundamentals

- [Evolution of the PDLC: From Waterfall to Agentic Delivery](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/evolution-of-the-pdlc/): How the product lifecycle evolved from waterfall to agile, DevOps and agentic delivery, and what genuinely changes when a model does part of the work.
- [Why Agentic AI Projects Fail: 7 Failure Modes and Fixes](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/why-agentic-ai-projects-fail/): Gartner expects over 40% of agentic AI projects to be cancelled by 2027. The seven failure modes behind that, why normal checks miss them, and the fix for each.
- [P0 Frame: How to Decide If an AI Agent Is Worth Building](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/p0-frame/): P0 Frame is the agentic PDLC's first phase: turn a request into a measured pain, test whether it needs AI at all, size the value and set autonomy per action.
- [P1 Design & Spec: Write a Spec an AI Agent Can Build From](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/p1-design-and-spec/): P1 Design & Spec turns a framed problem into an eight-field spec with EARS acceptance, a bar per slice and an authority budget an AI agent can build from.
- [The Hard Gate: The Hand-off You Can't Skip in AI Delivery](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/the-hard-gate/): Only one hand-off in the agentic PDLC halts the build. Which decisions are hard, which run behind a placeholder, and the four questions that tell them apart.
- [P2 Build & Prove: How to Build an AI Agent in Proven Slices](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/p2-build-and-prove/): P2 Build & Prove builds an AI agent in bolts of hours or days, gates every merge on a per-slice bar and ends with a shadow run beside the people doing the job.
- [P3 Run & Learn: How to Run an AI Agent in Production Safely](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/p3-run-and-learn/): P3 Run & Learn covers cut-over, rollback, drift and cost for AI agents in production, and the two-number report that decides whether the programme survives.
- [The 8 Feedback Loops of AI Delivery and the 3 Teams Forget](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/the-eight-loops/): Eight loops turn the agentic PDLC from a line into a ring. Five close themselves; three (cost, incident, governance) close only if a named person makes them.
- [The Evidence Pack: What Must Exist Before Each AI Hand-off](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/the-evidence-pack/): The minimum artefacts owed at each hand-off of an agentic AI project, and the one column that separates a real gate from a formality: enforced, or only written?

## Methods decoded

- [AI-DLC vs AIDD vs Agentic SDLC vs PDLC: What Each Term Means](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-dlc-vs-aidd-vs-agentic-sdlc/): AI-DLC, AIDLC, AIDDLC, AIDD, agentic SDLC, spec-driven development and BMAD: who coined each term, what it names, and the question it answers.
- [What Is AI-DLC? AWS's AI-Driven Development Lifecycle](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-ai-dlc/): AI-DLC is AWS's AI-Driven Development Life Cycle: AI proposes and people decide, across Inception, Construction and Operations, in bolts of hours or days.
- [What Is AI-Driven Development (AIDD)? The Daily Craft](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-aidd/): AI-driven development (AIDD) is the everyday craft of building software with coding agents: context files, story files, exact code first, and review by risk.
- [What Is the BMAD Method? Five Agents, One Build Loop](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-the-bmad-method/): The BMAD Method as it installs today: five named agents, skills in your coding tool, one Build loop that sizes itself. When its trail pays, and when not.
- [What Is Spec-Driven Development? Kiro and Spec Kit Explained](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-spec-driven-development/): Spec-driven development means writing the spec before the code and building from it with AI. How Kiro and GitHub Spec Kit do it, and the honest critiques.
- [How AI-DLC, BMAD, Spec Kit and Scrum Fit One Lifecycle](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/one-lifecycle-for-every-method/): Where AWS AI-DLC, Kiro, Spec Kit, BMAD, Scrum, Shape Up and stage-gate sit on the agentic PDLC's four phases, and the decisions none of them makes.
- [How Much Process Does a Change Need? Sizing Agentic Work](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/how-much-process-does-a-change-need/): Size process by a change's risk, not its size: four questions choose shallow, standard or deep work, and one trigger raises the depth mid-flight.

## Running delivery

- [How to Run an Agentic AI Project: A Step-by-Step Playbook](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/how-to-run-an-agentic-ai-project/): Twelve steps from a vague request to an AI agent in production that someone can defend: who owns each step, what it produces, and what is never delegated.
- [Bolts vs Sprints: Planning Work When AI Writes the Code](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/bolts-vs-sprints/): When AI builds a story in hours, a two-week sprint leaves it idle. How to plan in bolts: one unknown each, cut by dependency, integrated the same day.
- [How to Set Up a Kanban Board for an Agentic AI Project](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-kanban-board/): A project board for AI agent delivery: evidence as the exit rule for every column, risk bands as swimlanes, and WIP limits set by review capacity.
- [Cut Delivery from Months to Weeks: What Shrinks, What Won't](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/cut-delivery-time/): AI makes building fast; delivery is more than building. What compresses, what compresses only by design, what never compresses, with the arithmetic for each.
- [How to Review AI-Generated Code: By Risk, Not by Diff Size](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/review-ai-generated-code/): When agents write most of the code, review is the bottleneck. How to band changes by what they touch, route readers by band, and halve a review queue.
- [How Accurate Does an AI Agent Need to Be? The 3-Step Answer](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/how-accurate-must-an-ai-agent-be/): An AI agent's accuracy bar comes from money, not a round number: divide the damage of a wrong answer by the saving of a right one, per slice.
- [Is 82% Good Enough? How to Prove an AI Agent Meets Its Bar](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/prove-ai-accuracy/): A score from a test set is an estimate with a width. How to report the lower bound, count the cases a slice still owes, and give a verdict a sceptic accepts.
- [Shadow Mode and Canary Releases for AI Agents, Step by Step](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/shadow-mode-and-cutover/): How to launch an AI agent safely: run it in shadow beside the people doing the job, cut over at 5% per action, widen on evidence, and rehearse the way back.
- [Why Your AI Agent Costs 4× the Estimate, and How to Fix It](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-agent-costs/): An AI agent's bill rarely has one cause: context, model tier, cache and retries multiply. How to find each in the per-call log and fix them in order.
- [AI Agent Guardrails That Hold: Why a Prompt Is Not a Control](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-guardrails-that-hold/): A limit in a prompt can be talked past; a limit in a tool's signature cannot. Six controls that make an AI agent's boundaries real, with a test for each.
- [AI Governance Gates That Don't Slow Delivery: The Five Gates](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-governance-gates/): Five gates govern an AI agent from idea to wide use: intent, plan, behaviour, release and expansion. Who owns each, and the evidence each one needs.
- [AI Drift: How to Catch the Defect With No Error Message](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-drift-monitoring/): AI drift is behaviour changing with no deploy and no error. Why accuracy alerts arrive late, what to watch instead, and the two thresholds that catch it.
- [Postmortems for AI Incidents: Find the Missing Control](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-incident-postmortem/): An AI incident postmortem asks which enforced control would have made it impossible, not who was careless. Five moves, from the layer table to the brief.

## By role

- [The Agentic PDLC for Product Managers: What Changes Monday](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-product-managers/): What an AI product manager does in the agentic PDLC: measure the pain, decide if it is AI at all, set autonomy per action, derive the bar, report two numbers.
- [Agentic PDLC for Program Managers: Boards, Bolts, Reporting](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-program-managers/): How a program or delivery manager runs agentic AI delivery: the board, the cadence, the decision log, the review queue, gate conditions and two numbers.
- [Agentic PDLC for Solution Architects: Authority and Topology](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-solution-architects/): What a solution architect decides in agentic AI: which steps may be probabilistic, how many agents, the authority budget, where caps live, and the records.
- [The Agentic PDLC for Software Engineers: Build With Agents](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-engineers/): How software engineers work in the agentic PDLC: a context file, story files, exact code first, gated tools, a harness that decides merges, and a bolt a day.
- [AI-DLC and AIDD for Forward-Deployed Engineers: Field Guide](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-dlc-for-forward-deployed-engineers/): How a forward-deployed engineer runs AI-DLC and AIDD at a customer: measure their pain, have their risk owner sign the limits, build in their stack, hand over.
- [The Agentic PDLC for QA: How to Test Probabilistic Software](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-qa/): How QA works in the agentic PDLC: the proof each step owes, golden sets by slice, lower bounds instead of scores, injection suites, shadow runs and drift.
- [The Agentic PDLC for DevOps and Platform Teams](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-devops/): What DevOps and platform teams own in agentic AI: the landing zone, one model gateway, pinned versions, the harness as a required check, flags and rollbacks.
- [The Agentic PDLC for Business Sponsors: Ask for Two Numbers](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-for-business-sponsors/): What a business sponsor owns in an agentic AI programme: funding past cycle one, the autonomy ceiling, what counts as evidence, and three reports to insist on.
- [Agentic AI for Executives: Four Decisions Only You Can Make](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-ai-for-executives/): The four agentic AI decisions nobody below an executive can make, four questions to ask every cycle, seven red flags, and a first thirty days with no budget.

## Teams and organisation

- [How to Structure Teams for Agentic AI: Roles, Pods and Seams](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/team-structure-for-agentic-ai/): No new roles, none disappears, but two boundaries move. How to organise product teams, a platform team and governance for agentic AI, and where the seams go.
- [How to Measure AI Productivity Without Fooling Yourself](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/measure-ai-productivity/): Measure AI productivity against a baseline taken before the pilot, and report the saving beside the spend, review hours and re-runs, with a worked example.
- [AI Maturity Model: Six Controls You Can Test in Ten Minutes](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-delivery-maturity-model/): An AI maturity model that counts controls, not tools: six controls, each proven by a test you run in minutes, and how to pick the next one to build.
- [How to Roll Out Agentic AI Delivery: A 90-Day Plan](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/rolling-out-agentic-delivery/): A 90-day plan for rolling out agentic AI delivery: pick for provability, write the artefacts, build in slices, shadow beside the humans, report two numbers.

## Practice

- [Agentic AI Case Study: SkyWays, 90 Days from Pain to Proof](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/skyways-case-study/): A worked agentic AI case study: a fictional airline's rebooking assistant in thirteen episodes, from 31 requirements on day 1 to two numbers on day 90.
- [Agentic Delivery Cadence: What Runs Daily, Weekly, Quarterly](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-delivery-cadence/): The operating rhythm of an agentic AI project: what runs every day, on every change, every week, every cycle and every quarter, and who owns each check.
- [Agentic AI Workbench: Practise 90 Days of Delivery Decisions](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-delivery-simulator/): A free, browser-only agentic AI workbench: 13 dated episodes, 9 decision simulations, 17 prefilled tools and an evidence pack that downloads as one file.
- [Agentic PDLC Exercises: 12 Problems with Worked Answers](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-pdlc-exercises/): Twelve agentic AI delivery exercises with worked answers, three per phase: AI-fit, value, gates, bars, chains, bounds, cases, queues, bills, drift, controls.

## Interviews and careers

- [What Is a Forward Deployed Engineer? The FDE Role, Explained](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/what-is-a-forward-deployed-engineer/): What a forward deployed engineer (FDE) does, where the role came from, how it differs from a solutions architect, and why AI labs now hire FDEs.
- [How to Answer AI Interview Questions: Six Frameworks](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/how-to-answer-ai-interview-questions/): Six frameworks for AI interview questions: design, good enough or not, wrong answers, cost, incidents and behavioural stories, each with a worked example.
- [AI Product Manager Interview Questions, with Frameworks](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/ai-product-manager-interview-questions/): Ten deep AI product manager interview questions on AI judgement, design, metrics, evaluation, strategy, cost and leadership, each with a framework and answer.
- [Forward Deployed Engineer Interview Questions and Answers](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/forward-deployed-engineer-interview-questions/): Ten forward deployed engineer (FDE) interview questions (discovery, demos, evaluation, debugging, design, handover and pushback) with strong answers.
- [GenAI Engineer Interview Questions: RAG, Evals and Cost](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/genai-engineer-interview-questions/): Ten deep GenAI engineer interview questions on RAG failures, evaluation, fine-tuning, latency, cost, prompt injection, structured output and memory.
- [Agentic AI Engineer Interview Questions and Answers](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/agentic-ai-engineer-interview-questions/): Ten agentic AI engineer interview questions: the agent loop, enforced limits, tool design, MCP security, multi-agent systems, evaluation and runaways.
- [AWS Generative AI Interview Questions: Bedrock, AgentCore](https://akash-coded.github.io/aws-bedrock-agentcore-strands/learn/aws-generative-ai-interview-questions/): Ten AWS generative AI interview questions on Bedrock, AgentCore, Strands, Knowledge Bases, Guardrails, IAM, cost and resilience, with architecture answers.

## The manual, by role

- [Product manager](https://akash-coded.github.io/aws-bedrock-agentcore-strands/product-manager/): eight steps from a vibe to a number you can defend (markdown: https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki/Journey-Product-Manager)
- [Solution architect](https://akash-coded.github.io/aws-bedrock-agentcore-strands/solution-architect/): from requirements to a system that holds (markdown: https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki/Journey-Solution-Architect)
- [Engineering lead](https://akash-coded.github.io/aws-bedrock-agentcore-strands/engineering/): from a written task to code that ships (markdown: https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki/Journey-Engineering-Lead)
- [QA lead](https://akash-coded.github.io/aws-bedrock-agentcore-strands/qa/): from 'it works' to proof that it works (markdown: https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki/Journey-QA-Lead)
- [DevOps and platform](https://akash-coded.github.io/aws-bedrock-agentcore-strands/devops/): from a laptop to production, repeatably (markdown: https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki/Journey-DevOps)
- [Forward-deployed engineer](https://akash-coded.github.io/aws-bedrock-agentcore-strands/forward-deployed-engineer/): twelve steps in three stages, Frame, Deliver and Evolve, from a customer's pain to a system they run after you leave
- [Frame the engagement](https://akash-coded.github.io/aws-bedrock-agentcore-strands/forward-deployed-engineer/frame/): stage 1 of the forward-deployed engineer's guide, Qualify, Scope, Prove and Decide. Should we do this, and what exactly will we prove?
- [Deliver the system](https://akash-coded.github.io/aws-bedrock-agentcore-strands/forward-deployed-engineer/deliver/): stage 2 of the forward-deployed engineer's guide, Mobilise, Sign, Build and Hand over. Does it work in their world, and can they run it without you?
- [Evolve the relationship](https://akash-coded.github.io/aws-bedrock-agentcore-strands/forward-deployed-engineer/evolve/): stage 3 of the forward-deployed engineer's guide, Reframe, Codify, Reuse and Review. What did it teach them and us, and what comes next?

## Reference

- [The operating protocol](https://akash-coded.github.io/aws-bedrock-agentcore-strands/protocol/): for whoever funds the work; the four decisions only leadership can make
- [Twelve mental models](https://akash-coded.github.io/aws-bedrock-agentcore-strands/models/): what each predicts, the mistake it prevents, and a test for whether it landed
- [Frameworks, acronyms and the pictures](https://akash-coded.github.io/aws-bedrock-agentcore-strands/frameworks/): AI-DLC, AIDD, BMAD and SDD placed on one lifecycle
- [Templates](https://akash-coded.github.io/aws-bedrock-agentcore-strands/templates/) and [prompts](https://akash-coded.github.io/aws-bedrock-agentcore-strands/prompts/): the role journeys' artefact skeletons and prompts, copyable, by role; the forward-deployed engineer's are on the guide's stage pages
- [The wiki](https://github.com/akash-coded/aws-bedrock-agentcore-strands/wiki): the method written down, with decision trees, formulas, scenarios and exercises

## Optional

- [The method on one page](https://akash-coded.github.io/aws-bedrock-agentcore-strands/method/): the four phases, the eight loops, each role across the phases, and what a model may draft, as four boards
- [Home](https://akash-coded.github.io/aws-bedrock-agentcore-strands/): what the manual is, the four methods on one line, and a way in by role
- [Ninety Days, the simulator](https://akash-coded.github.io/aws-bedrock-agentcore-strands/simulator/): ninety days of one airline's agentic build, as a game
- [Source repository](https://github.com/akash-coded/aws-bedrock-agentcore-strands): curriculum, labs and this tutorial's source