500 free credits, no card. Try Smalt AI →

Showcase

See the work.
Not the promise.

Anyone can generate a slide. The question is whether the output survives an investment committee. Here's what your AI coworker actually hands back from a one-line brief — working files, sourced claims, and the model that did each step.

The platform in action

Every capability, in motion.

One plain-language brief goes in. A working model, a board-ready deck, and sourced research come out — watch it happen.

Smalt AI one brief, finished work
Plain language in

alder-dcf.xlsx

Excel · 14 linked sheets

“Build the Alder DCF from the data room — board call at three.”

What came back

  • Cells hold =NPV, =XIRR, =INDEX/MATCH — change a driver, the model recalculates
  • Assumptions → revenue → 3 statements → WACC → valuation, all linked
  • Three 5×5 sensitivity tables on WACC × growth

Shows its steps: Read 47 files → built 14 sheets → ran sensitivities → flagged 1 assumption

project-helix-lbo.xlsx

Excel · returns waterfall

“Model the Helix buyout — 6× entry, 5-year hold, 60% leverage.”

What came back

  • Sources & uses, debt schedule, cash sweep — live formulas
  • IRR and MOIC by exit year, sponsor vs management split
  • Entry / exit multiple sensitivity grid

Shows its steps: Parsed the brief → structured the waterfall → solved returns

series-b-deck.pptx

PowerPoint · 18 slides

“Turn this investment memo into a Series B raise deck.”

What came back

  • Native editable .pptx — real charts, not screenshots
  • Consulting-grade typography with an analyst right-rail per slide
  • Watermark-free, on your template

Shows its steps: Read the memo → built the narrative → laid out 18 slides

sea-digital-banks.pdf

Research memo · sourced

“Research the Southeast Asia digital-bank landscape for an IC.”

What came back

  • Paragraph-level citations to filings, news, and expert calls
  • Every claim traces to a verifiable source link
  • Entire 10-Ks read in one pass — no chunking

Shows its steps: Swept sources → read filings end-to-end → cited every claim

data-room-index.xlsx

Due diligence · 200 files

“What's in this data room, and what's missing?”

What came back

  • 200 documents classified, tagged, and indexed
  • Gaps flagged against a standard DD checklist
  • Fast triage handed to deep analysis for the items that matter

Shows its steps: Classified 200 files → escalated the material ones for synthesis

payments-comps.xlsx

Excel · trading + txn comps

“Pull trading and transaction comps for the SEA payments space.”

What came back

  • EV/EBITDA, EV/Revenue, P/E — pulled and formula-linked
  • Precedent transactions with announced multiples
  • Median / mean / quartile rows that recompute on edit

Shows its steps: Sourced the set → built the table → linked the stats

These are representative of the deliverables Smalt AI produces. Want to see one on your data? It takes a one-line brief and a few minutes.

Under the hood

The strengths of many models, for each step.

One chat. Behind it, Smalt leverages the strengths of many models — open-source and frontier alike — routing each step to whichever is best optimised for it: deep reasoning where the work has to be right, long context where the document is huge, fast and efficient where speed is what matters. You never pick a model. You never see the seams.

1

Reasoning & structure

Models, decks, valuation logic — the work that has to be right.

2

Long-document analysis

Full 10-Ks, multi-year transcripts, hundreds of pages — read in one pass.

3

Fast classification

Triaging a data room, tagging files, routing — speed where speed is fine.

Not locked to one vendor's mistakes. We review and update the platform continuously, so your work always runs on the models best suited to it — when a better one arrives, we adopt it and your workflow doesn't change.

See it on your work.

Build a DCF, generate a deck, run a sourced research scan. 500 free credits, no card.

Try Smalt AI See more use cases