Skip to main content
    Use-case library
    Risk, Controls & Compliance

    Finance Policy Assistant

    Answer 'what's allowed?' from approved policies with citations.

    Retrieval-grounded assistantAutonomy: SuggestSpeedRisk & ControlLow effortOrg maturity: AI-EnabledCFOCIO / CTO
    What changed

    RAG over approved policy, chart of accounts, delegation-of-authority and the close calendar now answers staff 'what is allowed?' questions with citations — replacing inbox queries to the controller with grounded, traceable answers.

    How it's built

    Retrieval-grounded assistant (rung 3 of 6). Fetch your evidence, then answer. Fixed path, read-only.

    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 steps 2, 3 and 4. Rules: step 1 · Human: step 5.

    Autonomy: T0 · Suggest
    Surfaces information, ranks, or recommends. Changes nothing. Human role: does all the work; may ignore the system entirely.
    How it is governed
    Cite-or-abstain. Source freshness and access controls inherited from the corpus.
    What the certification exams call this
    • Anthropic (Claude Certified Architect – Foundations (CCAR-F))The augmented LLM: a model with retrieval. source
    • OpenAI (OpenAI agent guidance (Agents SDK, Codex, OpenAI Academy))Data tools that retrieve the context the answer needs. source
    • Google Cloud (Generative AI Leader)Grounding in first-party enterprise data; retrieval-augmented generation (exam §3.3). source
    • Microsoft (AI-901 (Azure AI Fundamentals) and AB-100 (Agentic AI Business Solutions Architect))Knowledge sources and grounding data for Copilot Studio agents (AB-100). source

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

    Try it

    Illustrative demo

    Worked example of "Finance Policy Assistant" 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

    Workflow

    1. 1.RulesConnect the approved policy corpus (policies, CoA, DoA, close calendar).
    2. 2.RetrievalRetrieve the governing policy for the staff question.
    3. 3.Generative AIAnswer with the rule and a citation to the source clause.
    4. 4.Generative AIState approval or escalation requirements where relevant.
    5. 5.HumanEscalate genuinely ambiguous cases to the policy owner.

    Prompt recipe

    Answer the policy question using only the connected approved policies, chart of accounts and delegation of authority. Quote the governing clause, state who must approve, and if the policies do not cover the case, say so and route to the policy owner. Never invent a policy.
    Try in Ask KokoAI
    Controls required
    • Answers grounded only in the approved, current policy corpus
    • Every answer cites the source clause
    • Ambiguous cases escalated to the policy owner
    • Access-controlled; usage logged
    HomeNewsLearnIntelBuildAsk Koko