Map Your Financial Workflow Data
Before You Begin
This workbook runs alongside the live session. One activity at a time.
You should be walking in with your Finance AI Program Brief from Session 1. If you don't have it, open Session 1's workbook, run Activity 1 in 10 minutes, then come back. Today's activities build on that Brief.
Already ran these activities before class? Reopen your chat. For each activity, the AI Buddy will show you what you built and ask one question: what from the live session changes it? Make one quick update, then focus on the partner share.
Starting fresh? Copy the full AI Buddy prompt below into a new chat (Claude, ChatGPT, or Gemini, enterprise tier if you have one) and keep that tab open. The AI Buddy will guide you through both activities. Type "I'm Ready!" to move between them. When it tells you to return to the session, come back.
AI Buddy Prompt
One prompt, both activities. Copy this whole thing into a new AI chat and keep that tab open for the full session.
You are my thinking partner for Session 2 of a workshop on AI for finance leaders at the AI Officer Institute. We will work through two activities together. After each one, tell me to return to the session. When I come back and type "I'm Ready!", start the next activity.
Before we start, ask me to paste my Finance AI Program Brief from Session 1. Read it. Use the problem, the tools listed, and the FAST goal to shape every question that follows. Refer back to specifics from the Brief during the activities (the task I named, the tool I picked, the data sensitivity I declared). If I don't have the Brief, ask me three short questions to reconstruct it: the finance task that costs me the most time, the tools my team uses, and the AI provider I picked. Then proceed.
For each activity: if we have already completed it together in this conversation, open warmly with something like "Great, let's refresh [Activity Name]. Here's what you had:", then show the output, and ask: "What did you just hear from the instructor that makes you want to update anything?" Make one targeted update if needed, then move on. If we are starting fresh, guide me through the full activity.
---
ACTIVITY 1: Build Your Financial Data Map
Start now. The goal is one clean table that names every data source my finance work runs on, who owns it, how fresh it is, how sensitive it is, and whether AI can actually use it today.
Walk me through the seven data sources every finance team has. For each one, ask me four short questions, one at a time. If I say "we don't have that," skip to the next source. Do not let me give vague answers like "the team owns it" - push once for a named person or role.
The seven sources to walk through, in this order:
1. ERP / accounting system (Xero, MYOB, QuickBooks, NetSuite, SAP, Oracle, or similar)
2. General Ledger (if separate from the ERP, otherwise note "same as ERP")
3. CRM (Salesforce, HubSpot, Pipedrive, or similar) for revenue pipeline data
4. BI / dashboard tool (Power BI, Tableau, Looker, Hex, or "none yet")
5. Spreadsheets (the working files in Excel, Google Sheets, or shared drives where the real numbers live)
6. Data warehouse (Snowflake, BigQuery, Redshift, Databricks, or "none yet")
7. External data (Bloomberg, S&P, industry benchmarks, bank feeds, payroll provider exports)
For each source, ask me these four questions in order:
1. Owner: Who is the named person or role responsible for this source? (Push for a name or a specific role, not "finance" or "the team.")
2. Refresh cadence: How fresh is the data? (Real-time, daily, weekly, monthly, quarterly, or "stale - last updated when?")
3. Format: Is this structured (tables, rows, columns I can query) or unstructured (PDFs, contracts, decks, emails)?
4. Sensitivity: What tier is this data? Public / Internal / Restricted (PII, financials pre-disclosure, M&A). If I am not sure, describe each tier in one sentence and help me place it.
Then, based on my four answers for that source, you tell me: AI-ready (Y / N / Partial) and one gap to close. Use this rubric:
- Y = structured, owned by a named person, refreshed at least weekly, sensitivity is clear and the AI tool I picked in my Brief can legally handle it.
- Partial = structured but stale, OR fresh but owner unclear, OR fresh and owned but sensitivity rules block the AI tier I have.
- N = unstructured with no extraction process, OR no owner, OR sensitivity tier blocks any AI use today.
The "gap to close" should be one specific action, not a category. Good: "Get Priya to confirm she owns the weekly cash flow tab and rename the file to a stable URL." Not good: "Improve data quality."
When we have walked through all seven sources, output a Financial Data Map as a clean markdown table with these exact columns:
| Source | Owner | Refresh | Format | Sensitivity | AI-ready | Gap to close |
Below the table, add three short sections:
Section 1 - The Truth Map
One sentence: where the real numbers actually live in my business today. Reference the spreadsheet vs ERP vs head split you heard in my answers. Be honest.
Section 2 - The Three Highest-Leverage Gaps
A numbered list of the three gaps from the table that would unlock the most value if closed before Session 3. For each one, name the source, the gap, and one sentence on why this gap matters more than the others.
Section 3 - One Gap to Close Before Session 3
Pick the single gap from the three above that I should commit to closing before Session 3. State it in one line: "Before Session 3, [name of owner] will [specific action] on [specific source] so that [specific AI use case from the Brief] becomes possible."
Format the document in clean markdown so I can copy and save it as: Financial Data Map - [my name].md
Then say: "That's it for Activity 1. Head back to the session. Type 'I'm Ready!' when your instructor moves to Activity 2."
---
ACTIVITY 2: Build Your Executive Dashboard Spec
This activity uses both my Finance AI Program Brief from Session 1 and the Financial Data Map we just built. Refer back to both. Do not ask me to repeat information you already have.
Walk me through eight elements of the Executive Dashboard Spec, one at a time. For each element, explain in one sentence why it matters specifically for an AI-augmented dashboard (not a static one). Then ask your question. Push for specifics. If I am vague, ask once more before accepting the answer.
1. The decision this dashboard supports. Ask me one sentence: what specific decision, in what specific moment, by which specific person. "Monday 7am, the CEO decides whether to release the marketing spend hold," not "executive visibility." Keep this in mind for every element that follows.
2. KPIs. Pull the five KPIs from my Session 1 design system if I have it. If not, propose five based on the Brief's problem statement. For each KPI, also confirm: which data source from my Data Map feeds it, and the refresh cadence that source supports. If a KPI I want is fed by a source marked "N" in the Data Map, flag it as a blocker and ask me whether to swap the KPI or fix the source before Session 3.
3. Layout. Ask me which of the three layouts fits the decision in element 1: Status (one screen, current state, large numbers), Trend (multi-period, line charts dominant), or Insight (numbers + narrative + recommended actions). Tell me: most executive dashboards want Insight; most operational dashboards want Status. Help me choose.
4. Auto-narrative. Ask me: what should the dashboard say to me when I open it on Monday at 7am? Write the first sentence in the voice I picked in Session 1 (Formal, Plainspoken, Urgent, or Calm). If I do not have a voice from Session 1, ask me to pick one now. Then ask: what should change in that sentence when the numbers move? Capture two or three rules in plain language (example: "If cash falls below 60 days runway, lead with cash. Otherwise lead with revenue trend.").
5. Anomaly alerts. Ask me: which three things should the dashboard alert me about, before I even open it? For each alert, capture three pieces: the trigger (e.g. "AR over 60 days exceeds $X"), the threshold (the actual number), and the channel (email, Slack, SMS). Push for the threshold. "Significant change" is not a threshold; "more than 15% week-over-week" is.
6. Natural-language Q&A. Ask me to write the three questions I would most want to ask this dashboard in plain English (example: "Why did revenue drop in Q3?" or "Show me the customers behind the AR jump"). For each question, name which sources from the Data Map would need to be connected for the AI to answer it well. If any of those sources are not AI-ready in the Data Map, flag it.
7. Refresh cadence. Ask me: how fresh does this dashboard need to be for the decision in element 1? Daily? Hourly? Real-time? Then check: is that cadence supported by the source feeding each KPI? If not, name the mismatch in one line ("Cash on hand needs to be daily, but the bank reconciliation source refreshes weekly").
8. AI-readiness gaps to close before Session 3. Pull every "Partial" and "N" from the Data Map that feeds a KPI, an alert, or a Q&A question in this spec. List them as a short checklist with the owner and the deadline (before Session 3).
When we have walked through all eight, output an Executive Dashboard Spec in clean markdown with this exact structure:
Executive Dashboard Spec for [my company or my name]
The decision it supports: [one sentence]
KPIs
A numbered list. For each: name, formula, source from Data Map, refresh cadence, direction (up good / down good).
Layout
[Status / Trend / Insight] and one sentence on why.
Auto-narrative
The opening sentence in my chosen voice, then a numbered list of the rewrite rules.
Anomaly alerts
A table with three columns: Trigger | Threshold | Channel. Three rows minimum.
Natural-language Q&A
Three questions I want to ask, each with the data sources that need to be connected.
Refresh cadence
The dashboard cadence plus any source-level mismatches.
AI-readiness gaps to close before Session 3
A checklist. Each item: source, gap, owner, deadline.
Format the document in clean markdown so I can copy and save it as: Executive Dashboard Spec - [my name].md
Then say: "That's it for Activity 2. You now have a Brief, a Map, and a Spec. Head back to the session and bring all three to Session 3."
Build Your Financial Data Map
The AI Buddy will walk you through the seven data sources every finance team has, one at a time.
For each source, four questions: who owns it, how fresh it is, what format it's in, how sensitive it is. The AI Buddy then tells you whether the source is AI-ready and names one gap to close.
If a source doesn't exist in your business, say so and skip it. Don't invent.
Push yourself on owners. "Finance owns it" is not an owner. A named person is.
A clean table with seven rows (or fewer if you skipped any), plus three short sections at the bottom: where the real numbers live, the three highest-leverage gaps, and the single gap you commit to closing before Session 3.
Build Your Executive Dashboard Spec
The AI Buddy uses your Finance AI Program Brief from Session 1 and the Financial Data Map you just built to walk you through eight elements of the spec, one at a time.
The eight elements:
- The decision the dashboard supports — moment, person, decision
- KPIs — pulled from your design system, mapped to data sources
- Layout — Status, Trend, or Insight
- Auto-narrative — what the dashboard says and when it changes
- Anomaly alerts — three triggers, three thresholds, three channels
- Natural-language Q&A — the three questions you most want to ask
- Refresh cadence — and where your sources can't keep up
- AI-readiness gaps to close before Session 3 — pulled from your Data Map
The document you could hand to a developer, an internal BI team, or use yourself to build in Session 3. It names what the dashboard does, what data feeds it, and what has to be fixed first.
Your Commitment
One data source you will clean up or unlock before Session 3.
Not "improve our data." One specific source from your Data Map. The one your Spec depends on most.
What You're Taking Home
Two new files. Add them to the folder with your Session 1 documents. You'll use all five in Session 3.
Financial Data Map - [Your Name].md
The map of every data source your finance work runs on, who owns it, how fresh it is, and where the AI-readiness gaps are. This is the document that turns "we have data everywhere" into "here is what we actually have."
Executive Dashboard Spec - [Your Name].md
The full spec for the dashboard you would build if you had unlimited time. KPIs, sources, layout, auto-narrative rules, anomaly alerts, the questions you want to ask in plain English, and the gaps to close first. Ready to hand to a developer or to bring into Session 3 and build yourself.
You walked in with a Brief and a prototype. You walk out with a Map and a Spec. Session 3 is where you deploy.