A complete fictional sample. This analyst reports frequent data work, repetitive tasks and limited AI use. Budget reporting is an illustrative task. Your purchased plan uses your own answers; it will not reuse this example.
A fictional example. Your plan will use your own answers.
Spend less effort assembling reports. Get better at the decisions behind them.
You reported frequent data work, repetitive tasks and limited AI use, alongside regular judgment and responsibility. Start by improving one reporting workflow. Then use the results to explore deeper analysis or finance business partnering. Budget reporting is a worked example, not a task you told us you do.
A practical path forward90 days
Start smallBuild skillApply itLook ahead
Useful actions, prompts and learning resources. Take them at your own pace.
Your best place to start
Build a first variance report in ChatGPT
A six-row variance table, a profit bridge and a short draft commentary you can check against the source data.
Start with the first action. The rest is here when you need it.
Weeks 1–2
Start with one useful task
01Build a first variance report in ChatGPTA six-row variance table, a profit bridge and a short draft commentary you can check against the source data.35 min
Preparing a first draft is a useful starting point when recurring reporting takes time; you still own the numbers and explanation.
Here’s how
1
Open ChatGPT and start a new chat. Use the fictional numbers in the prompt below; no company files or customer information are needed.
2
Copy the complete prompt below. It includes the CSV, calculation rules and required output, so you can run it as written without a file upload.
3
Save the returned table and commentary in a practice spreadsheet or document titled January variance — draft. Keep the original CSV beside it.
4
Highlight every claimed cause, such as hiring or lost customers. Delete claims not supported by the supplied data and keep the remaining questions for investigation.
Your ready-to-use prompt
You are helping prepare a fictional monthly management report. Use only this CSV; all amounts are USD. This is training data, not a real business.
account,type,budget_usd,actual_usd
Revenue,revenue,1250000,1180000
Payroll,expense,420000,438000
Cloud,expense,96000,112000
Marketing,expense,140000,128000
Contractors,expense,85000,103000
Travel,expense,32000,29000
Calculate actual minus budget for every account and variance percentage as (actual-budget)/budget. For revenue, a negative variance is unfavorable; for expenses, a positive variance is unfavorable. Do not add revenue and expenses together. Calculate budget and actual total expenses, operating profit (revenue minus expenses), and the operating-profit change. Show the arithmetic or calculation code so I can verify it independently.
Return: (1) a six-row table with budget, actual, dollar variance, percentage variance and favorable/unfavorable label; (2) a bridge from budget profit to actual profit showing each account's contribution; (3) no more than 100 words of commentary separating observed changes from unanswered questions; (4) three questions for the finance owner. Round percentages to one decimal. Do not invent business causes, recommend layoffs, or describe an AI calculation as verified.
Use an AI tool approved for your work.
What a useful result looks like
The table contains all six accounts. Revenue is $70,000 below budget; cloud expense is $16,000 over budget, or 16.7%.
Budget profit is $477,000 and actual profit is $370,000. The six bridge contributions total a $107,000 decrease.
The commentary describes changes without claiming to know why they happened.
02Check the numbers without trusting the AIA spreadsheet with independent formulas and a signed review checklist that catches a deliberately wrong profit figure.40 min
Reliable review is the skill this example protects: a polished summary is useful only when its totals and signs are right.
Here’s how
1
Paste the original CSV into a blank spreadsheet at A1, splitting on commas. Confirm columns C and D contain numbers. Add E1 = variance_usd, F1 = variance_pct and G1 = profit_impact.
2
In E2 enter =D2-C2; in F2 enter =IF(C2=0,"N/A",E2/C2); in G2 enter =IF(B2="revenue",E2,-E2). Fill E2:G2 down to row 7 and format F as a percentage.
3
In C9 enter =SUM(C3:C7), D9 =SUM(D3:D7), C10 =C2-C9, D10 =D2-D9, E10 =D10-C10 and G10 =SUM(G2:G7). Verify E10 and G10 both equal -107000. This independently checks the AI table.
4
Run the prompt below to practice reviewing a deliberately flawed draft. Compare every correction to your own spreadsheet, not to a second AI answer.
5
Save a checklist with row count, expense totals, profit totals, bridge agreement and unsupported claims. Date your review and record any corrections; leave the draft unapproved if a check fails.
Your ready-to-use prompt
Review this fictional finance draft against the source figures. Do not introduce any new business facts.
Source figures: budget revenue 1250000; actual revenue 1180000. Budget expenses: payroll 420000, cloud 96000, marketing 140000, contractors 85000, travel 32000. Actual expenses: payroll 438000, cloud 112000, marketing 128000, contractors 103000, travel 29000.
Deliberately flawed draft: "Actual profit was $391,000, only $86,000 below plan. Cloud costs were favorable because usage fell. Marketing savings prove the campaign became more efficient."
Return a table with four columns: original claim, correct calculation or evidence, verdict, replacement sentence. Recalculate expense and profit totals. Distinguish an arithmetic error from an unsupported explanation. Finish with a corrected commentary of at most 80 words and three checks I must perform in my spreadsheet. Treat your answer as a draft requiring independent verification.
Use an AI tool approved for your work.
What a useful result looks like
Expense totals are $773,000 budget and $810,000 actual; E10 equals G10 at -107000.
The review rejects the $391,000 profit and $86,000 shortfall, and rejects both unsupported causal explanations.
Replacing any budget with zero produces N/A for that percentage, not an invented rate or a divide-by-zero result.
03Turn one course lesson into better commentaryA completed practice assignment: two prompt versions, a scored comparison and a corrected 80-word management note.60 min
This uses a reporting task to practice giving clear instructions and judging a draft, instead of collecting a course badge without applying it.
Here’s how
1
Open Anthropic's AI Fluency course. Study the lessons on Description and Discernment; this session covers selected lessons plus practice, not the whole course.
2
Start a fresh chat and use this baseline prompt: Summarize this for a manager: Fictional January case: revenue $1,180,000 vs $1,250,000 budget; expenses $810,000 vs $773,000; operating profit $370,000 vs $477,000. Profit is $107,000 below budget. The CSV contains no evidence about causes. Save the answer as version A.
3
In another fresh chat, run the complete prompt below and save the answer as version B. Using the same source, compare the effect of clearer instructions.
4
Give each version one point per check: correct revenue shortfall, correct expense increase, correct profit shortfall, no invented causes, and one useful next question. Score out of five, explaining every lost point.
5
Write the final note yourself using the better draft, then save both prompts, both outputs, the scorecard and your correction. That completed practice file is the assignment; do not claim you finished the full course.
Your ready-to-use prompt
Write a management note for a fictional finance review using only these checked facts:
Fictional January case: revenue $1,180,000 vs $1,250,000 budget; expenses $810,000 vs $773,000; operating profit $370,000 vs $477,000. Profit is $107,000 below budget. The CSV contains no evidence about causes.
Audience: a department head who needs to know what changed and what to ask next. Use plain words. In at most 80 words, give: one headline; two bullets with the revenue and expense changes in dollars; one sentence with the profit change; one specific question for the owner. Label explanations as unknown. Do not call lower marketing spend a proven efficiency gain. Do not invent a forecast, customer behavior, hiring changes, or savings. After the note, list the numerical checks I should repeat before sharing it.
Use an AI tool approved for your work.
What a useful result looks like
The practice file contains both outputs and a completed five-point scorecard with reasons, not just a course link.
The final note names the $70,000 revenue shortfall, $37,000 expense increase and $107,000 profit shortfall correctly.
The note has a next question but makes no unsupported claim about the cause.
04Assemble a report packet with MakeA manually run, three-module workflow that creates a private draft document from one checked practice row and records its document ID.90 min
Automating the handoff between a checked spreadsheet and a draft report is a small, visible project that keeps review with a person.
Here’s how
1
Use Make Academy's introductory material to learn modules and field mapping. In a practice Google Sheet, create these headers in A1:F1: period, status, revenue, profit, commentary, document_id. Add January-test, READY, 1180000, 370000, Profit is $107000 below budget; causes require review, and leave document_id blank.
2
With a practice Google account, build a Make scenario beginning with Google Sheets → Search Rows. Select this sheet, use the header row, filter status = READY and document_id empty, and limit the result to one row. Leave scheduling off.
3
Add Google Docs → Create a Document. Map period into the title with DRAFT before it. In the content, map period, revenue, profit and commentary into labelled paragraphs. Add: Fictional practice data. Human review required. This step assembles text; it does not generate or verify financial analysis.
4
Add Google Sheets → Update a Row using the row number returned by Search Rows. Map the existing fields back unchanged, set status to REVIEW and document_id to the new document's ID. Run once, open the private draft, and compare each field with the sheet.
5
Run once again: the REVIEW row must produce no second document. Review the draft manually and record reviewer/date in your practice notes. Keep the workflow without email or publishing steps. If a run fails after document creation, inspect the run and record that document's ID before retrying to avoid duplicates.
What a useful result looks like
The document is marked DRAFT, contains $1,180,000 revenue and $370,000 profit, and has no external sharing or sending step.
A successful run changes only the selected row to REVIEW and saves the created document ID; a repeat run creates no duplicate.
The private draft still needs a person's numerical and commentary checks before anyone uses it.
05Write a decision memo about cloud spendingA one-page memo with a checked $16,000 price-and-usage bridge, two options and one decision for the budget owner.55 min
Explaining a tradeoff and asking for a decision moves the example beyond producing a recurring report.
Here’s how
1
Use this fictional detail for the Cloud row: planned usage 80000 units at $1.20; actual usage 100000 units at $1.12. Keep the same unit definition in both periods.
2
Independently calculate usage impact = (100000-80000)*1.20 = $24000 and price impact = (1.12-1.20)*100000 = -$8000. Their sum is the $16000 expense increase; this bridge is an arithmetic decomposition, not proof of why usage grew.
3
Run the prompt below. Keep the options distinct: investigate usage while maintaining service, or ask an engineer to assess a 10% usage-reduction scenario. The second is only a scenario, not a committed saving.
4
Check the memo, add a named owner and review date only if someone actually agrees, and save the evidence and assumptions below the recommendation.
Your ready-to-use prompt
Draft a one-page decision memo for a fictional finance business partner. Facts: cloud budget = 80000 units * $1.20 = $96000; actual = 100000 units * $1.12 = $112000. Usage effect at budget price is +$24000. Price effect at actual usage is -$8000. Net expense increase is $16000. No information is available about service quality, customer demand, commitments, switching costs or the reason usage increased.
Compare option A: maintain service and investigate usage; option B: ask engineering to assess reducing current usage by 10%. At unchanged $1.12/unit, the illustrative gross reduction for B is $11200 per month before any implementation costs or service impact. Do not treat that as an approved or achievable saving.
Use these sections: Decision needed; What the data shows; Two options (benefit, risk, missing evidence); Recommendation for the next investigation; Owner and review date to agree. Include the arithmetic and distinguish facts, assumptions and questions. Keep it under 350 words. Do not invent an owner, deadline, financial forecast, or recommendation to change production systems.
Use an AI tool approved for your work.
What a useful result looks like
The $24,000 usage increase and $8,000 price reduction reconcile to the $16,000 expense increase.
The $11,200 scenario is labelled hypothetical and excludes implementation costs and possible service effects.
The requested decision is an investigation with missing evidence listed, not an automatic spending or operational change.
06Test the workflow on three different casesA short test log showing accuracy, active review time and the failure conditions that stop a draft from being used.75 min
One successful demo is not evidence of a dependable reporting process. Repeated checks reveal what can be reused and what still needs you.
Here’s how
1
Create three separate copies of the original CSV: baseline; zero-budget marketing (change its budget to 0); and missing-cloud actual (leave Cloud actual empty). Label them Case A, B and C. Keep the original baseline unchanged.
2
Write the expected results before running AI: A profit shortfall $107000; B budget profit $617000, actual profit $370000, shortfall $247000 and marketing percentage N/A; C incomplete data, so no validated actual expense or profit total.
3
Run the variance prompt on each case, adding this rule: Treat an empty numeric cell as missing, not zero; identify the account and stop totals that depend on it. Repeat your spreadsheet checks. Flag a missing input explicitly because a SUM formula alone may silently skip it.
4
For each case, record start/end time, review minutes, arithmetic errors, unsupported claims and pass/fail in a five-column log. Compare against a timed manual baseline only if you actually measured one; do not invent a time-saving percentage.
5
Write three stop rules: missing required values; a failed profit reconciliation; or unverified explanatory claims. Only a checked, complete result may enter the READY report-packet row; keep failed cases out of the automation.
What a useful result looks like
Case B changes the budget profit correctly and shows no marketing percentage; Case C flags missing data instead of silently treating it as zero.
All three cases have real test results, with review time separated from tool runtime.
A failed case cannot be marked READY without correction and a new human check.
07Package the work as a clear portfolio exampleA two-page case study and a tidy evidence folder that show your process, checks and decisions using fictional data.80 min
A small, inspectable example gives a hiring manager or team lead evidence of your judgment and workflow skills without exposing an employer's data.
Here’s how
1
Create a folder with the original CSV, checked spreadsheet, prompt comparison, automation screenshot, decision memo and three-case test log. Remove account names, IDs and private links from screenshots.
2
Run the outline prompt below. Fill measured results only from your test log; keep unmeasured benefits labelled as goals.
3
Build two readable pages in a document or Canva: page one shows the problem and a five-step flow; page two shows the reconciliation, one error caught, the decision memo and limits. Label both pages Fictional finance practice project.
4
Export a PDF and ask one trusted reviewer whether they can explain the problem, your contribution and the checks after a one-minute read. Keep it private until you choose to share it.
Your ready-to-use prompt
Create a two-page portfolio outline for a fictional finance practice project. The author used synthetic data, a spreadsheet, an AI chat and a manually run Make workflow. The project produced a variance table, an independently checked profit bridge, a comparison of two prompts, a private draft report packet, a cloud-spend decision memo and a three-case test log.
Known financial checks: budget profit $477000, actual profit $370000, shortfall $107000. A deliberately flawed draft claimed $391000 profit; the check caught it. The workflow creates drafts only and requires human review. No real business deployment, time-saving percentage, employer endorsement or hiring outcome has been established.
Return ready-to-edit headings and short copy. Page 1: problem, input, five-step workflow, my contribution. Page 2: evidence table, one error caught, decision supported, limitations, next test. Include explicit Fictional practice project labels. Where measured timing is needed, write Not measured rather than inventing a number. Keep responsibility for verification with the author.
Use an AI tool approved for your work.
What a useful result looks like
Every numerical result can be traced to a file, and fictional data is clearly labelled on both pages.
The case study shows a check and a decision, not just screenshots of tools or unsupported productivity claims.
The PDF contains no real employer information, private account details or working access links.
08Test two adjacent career paths with conversationsNotes from two career conversations and one next step supported by what people doing the work actually say.90 min
The worked project makes finance business partnering and finance systems analysis worth investigating; it does not establish fit or guarantee either role.
Here’s how
1
Identify one finance business partner and one finance systems analyst through your existing network or public professional profiles. Note one concrete question for each: decision ownership for the former; reporting controls and workflow maintenance for the latter.
2
Adapt this invitation and send it yourself if you choose: "I am exploring how finance work is changing. I built a small reporting example with fictional data and human checks. Could I ask you for 20 minutes about your role, the decisions you own and the skills a newcomer needs? I am looking for perspective, not asking for a job."
3
Ask both people: Which decisions are yours? Which recurring tasks are changing? What goes wrong when reporting is automated? Which part of my example is useful or missing? What small project would show readiness for this role? Ask permission before taking identifiable notes; a reply is not guaranteed.
4
Compare the two conversations in four columns: daily work, skill gap, evidence they want and my interest. Record what each person actually said and separate it from your interpretation.
5
Choose one next step based on the evidence: improve the portfolio, observe a workflow with permission, or study one named skill. If neither person replies, record that outcome and try another contact instead of inventing feedback or treating silence as a verdict.
What a useful result looks like
Each career idea is supported by a real conversation or explicitly remains an untested hypothesis.
Your next step names one skill or piece of evidence to build, with no claim that a role is AI-proof or a job offer is likely.
Put it into practiceComplete the commentary comparison in Weeks 3–4: two prompts, two outputs, a five-point scorecard and your corrected 80-word note. Pass only when the $70,000 revenue shortfall, $37,000 expense increase and $107,000 profit shortfall are correct and no cause is invented.
Build a small workflow with a review step
The example repeats a spreadsheet-to-document handoff. Learning modules and mapping gives that handoff a visible, testable structure without requiring code.
Put it into practiceAfter introductory Make lessons, build the three-module practice packet: search one READY sheet row, create a DRAFT document, then save its ID and change the row to REVIEW. Submit a screenshot, the private draft and a test note showing that a second successful run makes no duplicate.
Choose useful AI tasks and retain human decisions
The assumed role combines repeatable preparation with judgment. This noncoding course can help you decide which part of that work to delegate; access and certificates may cost money.
Put it into practiceAfter the lessons on applying generative AI at work, make a three-row table for variance calculation, commentary drafting and spending approval. For each, specify the AI task, independent check, human owner and stop rule. Use the missing-cloud case to show why incomplete figures must not become an approved report.
Your full skill priorities
Learn now
Prompting with an explicit output and evidence
Limited AI use makes a repeatable, verifiable starter prompt more useful than a complex agent.
Independent numerical verification
Frequent analysis and accountability make wrong figures a practical risk to manage.
Learn next
Workflow design with an exception path
Repetitive work is a good place to try an automation that stops for missing or inconsistent data.
Decision-focused writing
Your judgment answer supports practicing a short recommendation with alternatives and uncertainty.
Optional
SQL for a specific adjacent role
Consider it only if role descriptions or conversations show it is needed. Coding experience was not collected.
Facilitating a stakeholder conversation
Useful if business partnering interests you. Your answers do not establish a preference for people-focused work.
You have directions to explore.
Use these options to compare the work, the skills and the fit. You don’t need to choose a new career today.
Finance business partneringYour analysis and judgment answers make this worth investigating. Use the decision memo to test the work; your assessment did not establish presentation skill or an interest in changing roles.
Combines analysis with explaining choices to the people who act on them.
Skills you may bring
Financial analysis, if this is part of your current role
Working with numerical evidence
Reviewing consequential output
Skills to build
Stakeholder discovery and discussion
Presenting a decision with trade-offs
Suggested AI resilience: Moderate. Transition difficulty: Moderate. These are starting points to investigate, not guarantees.
Reporting and process improvement analystYour repeated digital work gives you a relevant sample process to examine. The draft report packet is a low-cost way to find out whether process design suits you.
Improves how information is gathered, checked and turned into useful reports.
Skills you may bring
Experience with repeated procedures
Analytical work
Awareness of errors and review
Skills to build
Mapping a workflow
Basic automation and exception handling
Suggested AI resilience: Moderate. Transition difficulty: Moderate. These are starting points to investigate, not guarantees.
Data quality analystYour frequent analysis and responsibility answers support an initial exploration. Look at real role descriptions before deciding whether additional technical study is needed.
Focuses on whether information is complete, consistent and suitable for a decision.
Skills you may bring
Frequent work with data
Checking numbers
Accountability for output
Skills to build
Documented validation rules
Root-cause investigation
SQL if required by the specific role
Suggested AI resilience: Moderate. Transition difficulty: Moderate. These are starting points to investigate, not guarantees.
What this means for your work.
Keep the strengths that make you useful. Look for the tasks worth doing differently.
Checking what is actually true
Frequent analysis gives you a useful place to practice catching wrong numbers and unsupported explanations.
Explaining a trade-off clearly
Your judgment answer supports practicing recommendations with alternatives, assumptions and a clear decision.
Owning the final review
You report accountability. Make your checks visible instead of treating a polished draft as finished work.
Tasks that could change6 tasks
Preparing a recurring report
High AI exposure
Example: let AI propose a first narrative from a fictional table. Keep reconciliation, source checks and final approval with you.
Your repeated, computer-based work suggests this could be relevant.
Explaining budget differences
High AI exposure
Use AI to calculate candidate differences and ask questions. Independently verify the numbers; never accept an invented cause.
You report frequent analysis and some content creation.
Collecting report inputs
High AI exposure
Move approved sample rows into a draft report packet. Pause at missing data and keep a manual review before distribution.
Defined procedures and repeated work can suit a simple automation.
Writing a recommendation
Moderate AI exposure
Use a draft to compare options and surface missing evidence. Decide which trade-off is acceptable yourself.
You report regular judgment rather than entirely rule-based work.
Discussing the numbers with a colleague
Moderate AI exposure
Prepare questions with AI, then clarify the actual business context with a person. Do not assume a model knows the decision maker’s priorities.
People are part of your work, even though they are not its largest reported feature.
Signing off a consequential analysis
Low AI exposure
Use a reproducible check sheet as support. Retain required professional review and your organization’s approval process.
You report responsibility for mistakes or outcomes.
Three ways your role could evolve
Your role stays; the workflow improves
One possibility is that useful tools reduce preparation work and let you spend more time on analysis. Test that with measured preparation plus review time, rather than assuming an AI draft saves time.
Keep a comparison of total effort before and after.
Agree which checks still need a person.
Use any verified benefit to propose one additional analytical responsibility.
A useful responseRun the repeatability test in action six, then decide whether the workflow deserves wider use.
The team is asked to cover more work
Another possibility is that your team’s workload or staffing changes. Your answer about changed responsibilities is a hypothetical expectation; it is not evidence that your employer plans cuts.
Document what your work contributes beyond producing a report.
Keep a portable, non-confidential example of your judgment.
Compare external role requirements before investing in retraining.
A useful responseUse the portfolio case in action seven to describe your contribution with evidence.
You choose an adjacent direction
You may find that you prefer interpreting decisions, improving reporting systems or managing data quality. Explore the actual work before committing to a new title or qualification.
Reuse your existing finance context where it fits.
Identify one skill gap for each option.
Treat conversations and small projects as information, not a promise of a job.
A useful responseComplete the two role conversations in action eight and compare their weekly work.
Other ways to use AI
Analysis and drafting
Turning a small table into a checked report and a draft recommendation.
Start here: Open ChatGPT using the linked quick start. Use the fictional data and exact prompts in the first two actions.
Workflow automation
Gathering report inputs and creating a draft packet consistently.
Start here: Use Make with a sample sheet and manual review. The fourth action includes a stop condition and no automatic email sending.
Learning and practice
Understanding how to brief an AI system and evaluate what it returns.
Start here: Use the linked AI fluency resource with the practice assignment. Bring back one revised prompt and one example of an error you caught.
Looking beyond 90 days
Month 1
Build a checked example you can repeat
A reconciled fictional variance report
A tested draft report packet with manual review
Months 2–3
Use evidence to choose your next direction
A decision memo and repeatability comparison
A non-confidential portfolio case and two role conversations
Months 4–6
Deepen the skill your evidence supports
A second project addressing a different task
One agreed learning priority based on the role or responsibility you want to explore
Months 7–9
Ask for relevant responsibility or explore openings
A specific discussion about useful new scope in your current role
If you want to move, a comparison of actual job requirements with your evidence
Months 10–12
Review what worked and choose again
A review of the tools, checks and skills still useful in your work
An updated career choice based on experience, not an AI forecast
Your assessment in context
Routine preparation is the clearest place to start.
Most of your reported work is on a computer, with frequent analysis and repeated procedures. Those answers make data preparation, first drafts and routine checks useful tasks to investigate. The budget examples in this plan are assumptions you can replace with a comparable low-risk task.
You also reported judgment and responsibility. That makes verification, interpretation and explaining trade-offs worthwhile skills to practice. It does not establish that your job is protected. These scores describe your answer pattern, not a probability of losing your job.
Your AI exposure score is 74/100. It reflects your answers; it is not a probability of job loss.
Fictional example. Preparation ideas, not predictions. Use approved tools and keep confidential information out of prompts.
Start with your work.
Your assessment and work snapshot are free. About 3 minutes.