Skip to main content
    Use-case library
    Record-to-Report

    Flux Analysis & Variance Narratives

    Draft traceable explanations for every material P&L and balance-sheet swing at close.

    LLM workflowAutonomy: DraftSpeedQualityMedium effortOrg maturity: AugmentedCFO
    What changed

    Rules have always been able to compute every variance and rank it against a materiality threshold. What changed is the writing: a language model can now draft the explanation for each material swing from the driver data behind it, cite the rows it used, and say when the data does not explain the move. Analysts edit commentary instead of assembling it.

    How it's built

    LLM workflow (rung 4 of 6). A coded control path: chain, router, parallel fan-out, evaluator-optimizer.

    Not an agent: the sequence of steps is fixed in advance, so the whole path can be tested end to end.

    A language model handles step 4. Rules: steps 1, 2, 3 and 5 · Human: step 6.

    Autonomy: T1 · Draft
    Produces a complete output for a human to review and release. Human role: reviews every item and releases it.
    How it is governed
    The path is testable end to end. Each step is independently evaluable.
    What the certification exams call this
    • Anthropic (Claude Certified Architect – Foundations (CCAR-F))Workflow: models and tools on predefined code paths (chaining, routing, parallelization, evaluator-optimizer). source
    • OpenAI (OpenAI agent guidance (Agents SDK, Codex, OpenAI Academy))Orchestrating via code, which the Agents SDK calls more deterministic and predictable. source
    • Google Cloud (Generative AI Leader)Workflow agents (sequential, parallel, loop) that run steps in a fixed order. source
    • Microsoft (AI-901 (Azure AI Fundamentals) and AB-100 (Agentic AI Business Solutions Architect))Sequential and concurrent orchestration patterns, with steps in a set order. source

    How Koko classifies AI · the full ladder is taught in AI & Agentic Delivery Foundations.

    Modeled value on KokoAI

    Synthetic KokoAI figures — a demo fixture, not audited or a guarantee

    Variance commentary per close
    40 → 12 hrs
    lower is better
    Recurring impact
    $1.2M/yr
    run-rate benefit
    Payback
    4 months
    ≈ 5,000 hrs/yr reclaimed

    Drafting month-end variance commentary drops 40→12 analyst-hours per close across the entity set; ≈5,000 hours/yr and ≈$1.2M/yr of recovered analyst capacity, every figure traced to source.

    Run this in Claude on KokoAI's data

    Opens your own Claude with the prompt and KokoAI's p&l for variance / flux analysis pre-loaded as text — no upload, no setup. Tailor it live to your own numbers.

    Try it

    Illustrative demo

    Worked example of "Flux Analysis & Variance Narratives" on synthetic data — edit the inputs to tailor it. Edit the inputs to tailor the output — it's baked synthetic data, so nothing leaves your browser until you run it live.

    Synthetic demo company (a ~$5B high-growth US AI SaaS) — swap in any name to tailor the output.

    What this run should concentrate on.

    Run live in Claude
    Sample data

    P&L for Variance / Flux Analysis (synthetic) — grab the one-pager PDF (the ROI, the data, and the prompt on one sheet) to share or keep, or the raw CSV / Markdown to drag into Claude or Microsoft Copilot. The prompt hand-off above already includes the same data as text.

    Synthetic KokoAI demo data — not audited

    Run it on your own document

    Drop a PDF, Excel, Word, CSV, or text file (e.g. a 10-Q, an AR aging, an invoice register) and we'll run this use case on it. Your file is read in your browser and never uploaded — only the extracted text goes into your own Claude or Copilot.

    Choose a file or drop it herePDF · XLSX · DOCX · CSV · TXT · up to 15 MB

    Read in-browser · not uploaded · 10-Q/10-K trimmed to financials + MD&A

    Workflow

    1. 1.RulesPull current actuals, prior period, budget/forecast and the GL detail behind each account.
    2. 2.RulesCompute variances and rank them by materiality against the documented threshold.
    3. 3.RulesFor each material swing, query the driver data behind it (volume, rate, mix, one-offs).
    4. 4.Generative AIDraft a 1–2 sentence explanation from that driver data only, and flag anything it cannot explain.
    5. 5.RulesAttach evidence links to the journals and sub-ledger lines behind each number.
    6. 6.HumanRoute the draft commentary to the controller for review and edit.

    Prompt recipe

    You are a financial analyst. Given the attached actuals, prior-period, budget and GL detail, identify every account whose variance exceeds [$X / Y%]. For each, write a 1–2 sentence plain-English explanation grounded in the driver data, cite the source rows, and flag anything you cannot explain from the data provided. Do not speculate. Run every calculation in code (a code-execution or data-analysis tool, or a spreadsheet), never in prose, and show the table it produced.
    Try in Ask KokoAI
    Controls required
    • Materiality threshold set and documented before the run
    • Every figure traceable to a GL/sub-ledger source row
    • Controller reviews and approves narrative before it leaves finance
    • Model output marked AI-drafted in the working papers
    HomeNewsLearnIntelBuildAsk Koko