4.1.4. Planner

Planning steps are incorporated in many modern LLM applications.

The explicit inclusion of a planning step improves overall performance by having the system consider the full scope of the problem before acting.

In this implementation, the plan remains static. In longer-horizon agentic pipelines, however, it's common to include mechanisms for replanning or updating the plan as the system progresses.

Broadly, planners take two forms:

<ol style="margin-left: 1em; line-height: 1.6; padding-left: 0.5em;"> <li style="margin-bottom: 1.2em;"> <strong>Task-orientated (used in this cookbook)</strong><br /> <p style="margin-top: 0.5em; margin-bottom: 0.5em;"> The planner outlines the concrete subtasks the downstream agentic blocks should execute. The tasks are phrased in an action-orientated sense such as "1. Extract information on R&D activities of Company IJK between 2018–2020." These planners are typically preferred when the goal is mostly deterministic and the primary risk is skipping or duplicating work. </p> <p style="margin-top: 0.5em; margin-bottom: 0.5em;"> Example tasks where this approach is useful: </p> <ul style="margin-top: 0.5em; margin-bottom: 0.5em; padding-left: 1.2em;"> <li><strong>Law</strong>: <em>"Extract and tabulate termination-notice periods from every master service agreement executed in FY24"</em></li> <li><strong>Finance</strong>: <em>"Fetch every 10-K filed by S&P 500 banks for FY24, extract tier-1 capital and liquidity coverage ratios, and output a ranked table of institutions by capital adequacy"</em></li> <li><strong>Automotive</strong>: <em>"Compile warranty-claim counts by component for Model XYZ vehicles sold in Europe since the new emissions regulation came into force"</em></li> <li><strong>Manufacturing</strong>: <em>"Analyse downtime logs from each CNC machine for Q1 2025, classify the root-cause codes, and generate a Pareto chart of the top five failure drivers"</em></li> </ul> </li> <li style="margin-bottom: 1.2em;"> <strong>Hypothesis-orientated</strong><br /> <p style="margin-top: 0.5em; margin-bottom: 0.5em;"> The plan is framed as a set of hypotheses the system can confirm, reject, or refine in response to the user's question. Each step represents a testable claim, optionally paired with suggested actions. This approach excels in open-ended research tasks where new information can significantly reshape the solution space. </p> <p style="margin-top: 0.5em; margin-bottom: 0.5em;"> Example tasks where this approach is useful: </p> <ul style="margin-top: 0.5em; margin-bottom: 0.5em; padding-left: 1.2em;"> <li><strong>Law</strong>: <em>"Does the supplied evidence satisfy all four prongs of the fair-use doctrine? Evaluate each prong against relevant case law"</em></li> <li><strong>Pharmaceuticals</strong>: <em>"What emerging mRNA delivery methods could be used to target the IRS1 gene to treat obesity?"</em></li> <li><strong>Finance</strong>: <em>"Is Bank Alpha facing a liquidity risk? Compare its LCR trend, interbank borrowing costs, and deposit-outflow and anything else you find that is interesting"</em></li> </ul> </li> </ol>

Prompting our planner

We will define two prompts (one system and one user) for the initial planner.

The most notable characteristic of our system prompt below is the use of 'persona-based' prompting. We prompt the LLM giving it a persona of an internal company expert. This helps to frame the tone of the model's response to the behaviour that we want - a direct, action-orientated task list that is fit for the financial industry.

This is then extended in the user prompt, where we prepend the user_question with information on this specific situation and how the planner should handle it.

In production settings you can super-charge this template by dynamically enriching the prompt before each call. You can inject information on the user's profile —sector, role, preferred writing style, prior conversation context—so the planner tailors its actions to their environment. You can also perform a quick “question-building” loop: have the assistant propose clarifying questions, gather the answers, and merge them back into the prompt so the planner starts with a well-scoped, information-rich request rather than a vague one.

Another flow that can work well is to allow users to view the plan and optionally edit it before it is executed. This is particularly effective when your AI system is acting in more of an assistant role. Giving domain experts such as lawyers or pharmaceutical researchers the flexibility to steer and incorporate their ideas and research directions deeper into the system often has the dual benefit of improving both system performance and end user satisfaction.

async def initial_planner(user_question: str) -> str:
    """Return an initial plan for answering the user's question."""
    initial_planner_system_prompt = (
        "You work for the leading financial firm, ABC Incorporated, one of the largest financial firms in the world. "
        "Due to your long and esteemed tenure at the firm, various equity research teams will often come to you "
        "for guidance on research tasks they are performing. Your expertise is particularly strong in the area of "
        "ABC Incorporated's proprietary knowledge base of earnings call transcripts. This contains details that have been "
        "extracted from the earnings call transcripts of various companies with labelling for when these statements are, or "
        "were, valid. You are an expert at providing instructions to teams on how to use this knowledge graph to answer "
        "their research queries. \n"
        "The teams will have access to the following tools to help them retrieve information from the knowledge graph: \n"
        "1. `factual_qa`: Queries the knowledge graph for time-bounded factual relationships involving a given entity and predicate. \n"
        "2. `trend_analysis`: Wraps the factual_qa tool with a specialised agent to perform in-depth trend analysis \n"
        "It shoudld also be noted that the trend_analysis tool can accept multiple predicate arguments as a list. \n "
        "You may recommend that multiple calls are made to the tools with different e.g., predicates if this is useful. \n "
        "Your recommendation should explain to the team how to retrieve the information from the database through these "
        "tools only. "
    )

    initial_planner_user_prompt = (
        "Your top equity research team has came to you with a research question they are trying to find the answer to. "
        "You should use your deep financial expertise to succinctly detail a step-by-step plan for retrieving "
        "this information from the the company's knowledge base of earnings call transcripts extracts. "
        "You should produce a concise set of individual research tasks required to thoroughly address the team's query. "
        "These tasks should cover all of the key points of the team's research task without overcomplicating it. \n\n"
        "The question the team has is: \n\n"
        f"{user_question} \n\n"
        "Return your answer under a heading 'Research tasks' with no filler language, only the plan."
    )

    input_messages = [
        {"role":"system", "content": initial_planner_system_prompt},
        {"role":"user", "content": initial_planner_user_prompt}
    ]

    initial_plan = await client.responses.create(
        model="gpt-4.1",
        input=input_messages
    )

    return initial_plan.output_text

plan = await initial_planner("How can we find out how AMD's research priorties have changed in the last 4 years?")

print(plan)