Develop a Workplan

Agentic Engineering is a mindset where you act as a conductor coordinating an orchestra of tools rather than a solo instrument player. Modern engineering increasingly involves guiding AI-assisted workflows, maintaining context, and making design decisions while multiple tools contribute to the final solution. This lecture introduces the first and most important step in the process: developing a strong workplan.

The purpose of developing a planning mindset is to improve technical quality, reduce wasted effort, and create a clear path from concept to implementation.

Agentic Engineering Mindset

Traditional engineering workflows often focus on writing code as quickly as possible. In contrast, agentic engineering emphasizes planning and orchestration. The engineer defines objectives, breaks work into manageable components, and monitors progress while intelligent tools assist with execution. The AI does not replace engineering judgment; instead, it accelerates implementation when guided by clear direction.

A successful agentic workflow includes structured prompts, modular tasks, and deliberate context management. When these pieces are combined effectively, AI systems become reliable collaborators rather than unpredictable assistants.

Instructor Perspective

One of the most common patterns I see with engineering students and professionals is the tendency to start coding too early. Writing code feels productive because it produces immediate output, but without a clear workplan it often leads to rework, frustration, and projects that become difficult to maintain. In many cases, the real engineering challenge is not coding itself but deciding what should be built and why.

Experienced engineers spend more time thinking before implementing. They ask what data is available, what decisions the model needs to support, how success will be measured, and how the project will evolve over time. A strong plan reduces uncertainty and creates momentum because each step has a clear purpose.

Agentic tools amplify this reality. When AI can generate code quickly, it becomes even more important to slow down and think carefully about structure and direction. The limiting factor is no longer typing speed but engineering judgment. Your value as an engineer comes from defining the problem, guiding the workflow, and verifying that results make sense.

A well-developed workplan is not extra work. It is the work. Everything else becomes easier when the plan is clear.

Planning Before Coding

Planning is the phase that determines whether a project stays organized or becomes chaotic. Before writing code, the engineer should understand the system, identify required files, and define how each task contributes to the final goal. Many project failures come from skipping this step and beginning implementation without a clear roadmap.

A good workplan normally includes:

clear goals and success criteria
decomposition of the problem into smaller tasks
identification of required files, data, and tools
checkpoints for progress and review

The important principle is that you remain responsible for approving or refining the plan before execution begins. The planning stage is where engineering decisions happen.

Planning with Plan Mode

Many modern agentic tools provide a Plan Mode that helps generate structured execution plans. In this mode, the agent explores the repository, identifies relevant files, asks clarifying questions, and proposes a step-by-step plan that includes task order and file paths.

The engineer’s role is to evaluate that plan critically. You should review the proposed workflow, edit steps that are unclear, and ensure that dependencies are correct before allowing execution. Plan Mode is helpful because it encourages discipline and prevents early mistakes, but it works best when the engineer actively guides the process.

Context management is equally important. Only include files that are known to be relevant. Adding unnecessary files increases noise and can confuse the agent, reducing reliability and consistency.

Multi-Agent Orchestration

As projects grow in complexity, a single agent may not be sufficient. Multi-agent orchestration allows specialized agents to work in parallel under the supervision of a higher-level coordinator. In this structure, a meta-agent creates tasks and assigns them to worker agents, while monitoring systems track progress and results.

This approach can significantly improve productivity, but it also introduces challenges such as file conflicts or duplicated work. Effective orchestration depends on the engineer clearly defining boundaries and dependencies between tasks. The engineer remains responsible for integrating outputs into a coherent final solution.

Key responsibilities during orchestration include:

break requirements into discrete tasks
define dependencies and execution order
validate results before integration

When coordinated properly, multi-agent workflows can produce results faster than traditional approaches while maintaining engineering rigor.

Integrating Agentic Planning into the Course Project

The Machine Learning for Engineers course project provides an ideal setting to practice agentic planning. Rather than jumping directly into model development, students should begin by defining the engineering problem and outlining the full workflow from data acquisition to final reporting.

A strong project workplan typically starts with identifying a case study and drawing a system diagram that clearly shows features (inputs) and labels (outputs). This is followed by defining whether the problem involves classification, regression, or clustering, and then reviewing related literature and datasets. Data preparation, uncertainty analysis, and timelines should be considered early so that implementation proceeds smoothly.

Planning early allows students to recognize risks and avoid common pitfalls such as missing data, unclear evaluation metrics, or poorly defined objectives.

Project Deliverables and Planning Alignment

The final project report reflects the quality of the workplan developed at the beginning of the project. Students should think ahead to the final structure while planning their approach. The report includes:

cover letter introducing the context and contributions
highlights summarizing major outcomes
manuscript sections from introduction through conclusions
validation, deployment, and uncertainty analysis
project timeline and progress updates

When these deliverables are considered early, the project evolves more naturally and the final documentation becomes easier to assemble.

Key Takeaways

Agentic engineering begins with planning. The goal is not simply to use AI tools, but to organize and direct them effectively. Engineers who take time to refine their workplan produce higher-quality results with less rework. Plan Mode and multi-agent workflows are powerful tools, but they require thoughtful oversight and structured decision-making.

The workplan you create now becomes the foundation for coding, coordination, visualization, and communication in the remaining lectures.