The short answer
Use Excel Copilot if you're already inside Excel and want AI as an assistant — to suggest a formula, clean up a table, build a quick chart, or analyse a column of data. The value is in-flow speed for someone who already knows how to model.
Use Smalt AI when you don't want to start with a blank workbook. Brief in plain language — "build a DCF for ABC, 5-year horizon, three scenarios" — and a 14-sheet model lands in your inbox with formulas live, scenarios wired, sensitivity tables built. Smalt AI is the AI that builds the workbook; Excel Copilot is the AI that helps you while you're in a workbook.
Capability comparison
| Capability | Excel Copilot | Smalt AI |
|---|---|---|
| Where it works | Inside Excel only — needs the workbook open | Web chat — outputs a working .xlsx file |
| Build a 14-sheet DCF from scratch | No — helps you build inside Excel but doesn't construct the multi-sheet structure | Yes — Assumptions through Sensitivity, all sheets linked, in one delivery |
| Suggest a formula in a cell you have selected | Excellent — that's the core feature | Not the use case (you're not in Excel; Smalt AI hands you the formula in the file it builds) |
| Clean a messy spreadsheet | Strong — sort, dedupe, identify outliers, format consistently | Available — upload the file, ask it to clean and return |
| Build a sensitivity table | Helps you build it once you know the structure — you set up the data table | Three 5×5 tables built automatically alongside the base case (75 full-recalc formulas) |
| Pull data from filings (10-K, 10-Q) | Limited — you'd paste in or import the data first | Direct — Claude 1M context ingests entire 10-Ks; pulls historical financials into the model |
| Real =SUM, =NPV, =XIRR formulas | Yes when it suggests them in your sheet | Yes throughout the workbook it builds — every cell traces back to a driver |
| Pricing | Bundled with Microsoft 365 Copilot at $30/user/mo, requires E3/E5 license | Free tier 500 credits, Basic ~$18/mo, Pro ~$45/mo, Enterprise self-host |
Where Excel Copilot is the right tool
For an experienced modeller working inside an existing workbook, Excel Copilot is genuinely useful. You know the structure you want; Copilot accelerates getting there. "What's the formula to compute working capital days?", "Suggest a chart for this range", "Highlight cells where revenue declined YoY." All in-flow, all without leaving the spreadsheet.
For data hygiene tasks — cleaning, sorting, deduping, formatting — Copilot's table-aware features are excellent. Drop a messy CSV in, get a clean table out without writing the cleanup formulas yourself.
For quick analysis on data already in your sheet — averages, distributions, outliers, simple charts — Copilot is fast and right next to where the data lives.
Where Smalt AI is the right tool
For building a model from zero, Smalt AI is the right shape. The hard part of a DCF is not writing one cell — it's structuring 14 linked sheets correctly, wiring scenarios through INDEX, computing interest on beginning balance to break circularity, building three 5×5 sensitivity tables with center-cell sanity checks. Smalt AI does that construction work; Copilot helps you do it inside Excel.
For institutional formatting and conventions — blue-font hardcoded inputs, green cross-sheet links, mid-year discounting, source comments on every input, balance-sheet check tabs — Smalt AI follows IB-grade standards by default. Copilot follows whatever the open workbook follows.
For data sourcing — pulling 5 years of historicals from 10-Ks, fetching analyst estimates, looking up risk-free rates and ERPs — Smalt AI does it as part of the build. With Copilot, you'd source the data separately and paste it in.
For round-trip workflows on existing workbooks — upload your model, ask Smalt AI to add a sensitivity tab on EBITDA × multiple, or to extend the projection by two years — Smalt AI handles it as a workbook operation. Copilot can edit cells inside the file you have open; it can't append a new sheet wired to existing drivers from a verbal brief.
How they fit together
The complementary workflow: Smalt AI builds the model and hands back a working .xlsx. You open it in Excel; Copilot helps you tweak. You change a driver assumption based on something you saw in the latest earnings call; Copilot suggests the cleanest formula refactor. You want to add a chart for a new view; Copilot generates it.
Most analysts will end up using both. Smalt AI for the construction job. Copilot for the in-Excel refinement after.
Honest caveats
- Smalt AI's models are strong defaults, not custom-fit. If your firm has a house-style three-statement template that differs from Smalt AI's defaults, you'll need to either ask for adjustments in the brief or refactor in Excel after.
- Excel Copilot is not a build tool. Asking it to "build me a DCF for ABC" inside Excel won't yield a working multi-sheet model — it will offer suggestions, but the structure is on you.
- Pricing math differs. Copilot is bundled with M365 (~$30/user/mo on top of E3/E5). Smalt AI is credit-metered standalone (Free / $18 / $45 / Enterprise).
The summary
Excel Copilot is the assistant inside Excel. Smalt AI is the AI virtual employee that builds the workbook. Different jobs. Most finance teams benefit from both — Smalt AI to construct, Copilot to refine.
500 free credits, no credit card. Try Smalt AI on a real workflow and see for yourself.