ModeLoop

Designed illustrations and two-voice synthetic narration. Standalone five-minute technique with a synthetic single-product revenue bridge. Planned 100 units at £50 versus actual 120 at £45: volume +£1,000, price -£600, net +£400 using volume-first order. Reverse order allocates the interaction differently but reaches the same total. No FX, mix, returns, rebates or timing effects assumed. A reconci...

Designed illustrations and two-voice synthetic narration. Synthetic annual end-period DCF in pounds thousands, not a live AI demonstration, tested workbook or investment recommendation. Review FCFF scope, terminal cash-flow timing, discounting, the narrow equity bridge and discount-rate sensitivity. Hypothetical WACC 10%, growth 2%, cash flows 80/90/100; steady-state growth and reinvestment assume...

Designed illustrations and two-voice synthetic narration. Two teaching roles debate calculation precision and false certainty. Synthetic planning price £79.995 per unit, 1,250-unit central case and illustrative 1,150/1,350-unit scenarios. These are not probability intervals. Preserve input precision, separate calculation from display, and test a hypothetical £100,000 threshold against the agreed m...

Designed illustrations and two-voice synthetic narration. Synthetic one-month teaching case in £000, not a live AI demonstration or a tested workbook. Follow working-capital, fixed-asset, debt, equity and cash roll-forwards. Reconcile direct and indirect operating cash, test a receipt delay and detect a deliberate revenue-versus-receipts error. Excludes tax, interest, dividends, FX, new equity and...

Designed illustrations and two-voice synthetic narration. Synthetic teaching example, not a software recording, tested workbook or measured AI product test. AI answer check format: an explicitly illustrative AI-style answer, not a measured response from a named model. Distinguishes arithmetic mean, total change and CAGR by rebuilding the endpoint. Positive, consistently measured revenue levels at...

Designed illustrations and two-voice synthetic narration. Synthetic teaching example, not a software recording, tested workbook or measured AI product test. Plan model scope, outputs, evidence, selected assumptions, dependencies, timing, signs and checks before formulas. Illustrative two-month operating-cash bridge. Excludes tax, capital expenditure, inventory, financing and dividends; expenses pa...

Synthetic teaching example with designed illustrations and two-voice synthetic narration. Not a software recording, tested workbook or measured AI failure. A self-contained lesson on source locators, units, periods, approval, conflicts and change logs. Synthetic 8%/12% volume-growth case and simple receivables estimate. Not a complete cash forecast. Sources: https://hodacfo.com/financial-model-ass...

Synthetic teaching example with designed illustrations and standard two-voice synthetic narration. Not a software recording, tested workbook, measured AI failure or financial advice. Operating margin equals operating profit divided by revenue. Synthetic divisions retain 10% and 40% margins while revenue mix moves the group from 13% to 37%. Central costs and eliminations are excluded; commercial ca...

Synthetic teaching example with designed illustrations and standard two-voice synthetic narration. Not a software recording, tested workbook, measured AI failure or financial advice. A controlled AI and Power Query workflow: preserve raw rows, document locale and duplicate rules, expose exceptions, and reconcile refresh totals. Example: seven source rows, one removed duplicate, five accepted and o...

Synthetic teaching example with designed illustrations and standard two-voice synthetic narration. Not a software screen recording, measured demo, lender quote or financial advice. Two hypothetical fully amortising loans: principal 100,000 currency units, nominal annual rate 6% divided by 12, payments at month-end, 60 versus 120 months. Fees, tax, insurance, early payments and rate changes exclude...

Synthetic teaching example with designed illustrations and standard two-voice synthetic narration. Not a software screen recording, measured demo, lender quote or financial advice. A bounded AI extraction workflow for reported figures: preserve source page, reporting period, scale and signs, then check the candidate input. Sources: https://learn.microsoft.com/en-us/power-query/connectors/pdf and h...

Synthetic four-month GBP revenue check. Define the rule, build an independent Excel benchmark and test failures. Standard synthetic narration, not a recorded Excel run or full workbook audit. Use synthetic data and follow your organisation’s rules.

A synthetic GBP teaching case. Reconcile profit to cash using receivables, inventory and supplier payables. Standard synthetic narration, not a real forecast or measured AI test. Not financial advice.

A ModeLoop myth check using a hypothetical one-product business. A 10% price cut and 10% more units do not automatically preserve profit. The worked example separates revenue recovery, zero-profit break-even and preserving operating profit. A sales target is not a sales forecast. Standard synthetic narration; illustrative example, not a recorded AI-tool test.

Companion podcast episode to the PolyMind Explained video. PolyMind is a ModeLoop product: a browser dashboard that monitors Polymarket (with Kalshi available as a second source) and turns market activity into alerts with AI analysis attached. This episode covers the twelve signal types in the app's own grouping, the severity tiers, and the free tier. Try PolyMind: https://polyminds.netlify.app -...

Which parts of the analyst's job have moved with current AI tools, and which parts have not moved at all? This episode separates the two, with one worked illustration you can reuse tomorrow. Audio companion to the video explainer. Not financial advice.

When should your model use IRR, and when XIRR? In this episode we look at the difference between the two in Excel: IRR assumes evenly spaced cash flows, while XIRR pairs every cash flow with its actual date. We cover the two mistakes one wrong AI answer can hold (annualising a quarterly rate by multiplying by four, and ignoring real payment dates), the checks to run before you trust the number, an...

AI can write VBA in seconds, but a wrong macro can overwrite the wrong workbook, and running a macro clears Excel's undo stack. This episode shows how to use AI to write VBA macros safely: save the workbook as .xlsm, open the Visual Basic Editor with Alt+F11, start every module with Option Explicit, enable content one file at a time, and always run new code on a copy, never the live model. Include...

Excel dynamic arrays can look risky in a financial model, especially when a spilled range feeds covenant checks or project cashflows. This episode walks through the same worked examples as the ModeLoop video lecture and blog, using a small project table with SPILL errors, covenant style checks, and AI prompts you can adapt in your own work. We cover how to use UNIQUE and SUMIFS for duplicates, FIL...

Excel LAMBDA lets you turn repeated financial model calculations into your own named functions, and AI makes writing them much faster. This lecture shows how to build custom Excel functions with LAMBDA and AI for a project finance model, how to test them, and when not to use them at all. In this episode: What LAMBDA does: parameters plus a calculation, defined once in Name Manager and used everywh...

AI can write and explain complex Excel formulas for financial models, but only if you give it the logic first. This lecture shows how to describe a calculation in words before asking for syntax, how to make AI explain an inherited formula and its failure modes, and how to rewrite it with LET without hiding errors. In this episode: Logic before syntax: a sculpted senior debt repayment written from...

Using AI on confidential deal models is safe only if you control what leaves your machine. This lecture shows how financial modellers can use Claude, ChatGPT and Copilot on live M&A and project finance deals without leaking client data: which account to use, how to describe the structure instead of the deal, and how to anonymise a model before you share it. In this episode: What counts as conf...

AI prompts for Excel formulas only work when the AI knows your model: the layout, the sign convention, the timing and the Excel version. This lecture gives you a five-part prompt that turns Claude, ChatGPT or Copilot into a fast junior analyst on project finance, M&A and FP&A models. In this episode: Why vague prompts fail: invented cell references, sign conventions, timing, Excel version,...

Claude vs ChatGPT vs Copilot in Excel vs Gemini for financial modelling: which AI should you actually use? The tool matters less than where it sits relative to your workbook. This lecture compares the four through a modeller's eyes, with project finance, M&A and FP&A examples. In this episode: In the workbook vs beside it: why Copilot and Excel add-ins behave differently from chat tools Cl...

AI in financial modelling: what Claude, ChatGPT, Copilot in Excel and Gemini can already do for your Excel models, and where they still fail. The first episode of Part 3 sets the ground rules for using AI on project finance, M&A and FP&A models without giving up control of the numbers. In this episode: Where AI already saves real time: writing and explaining formulas, structuring a model,...