Approve the story.
Then build the deck.
Every other tool is one-shot: describe, generate, then fix. The expensive mistake isn’t a crooked logo. It’s forty slides off a misread story.
Built for deal teams
Every deck is rendered and inspected before a job reports done, and every change is recorded with the instruction behind it.
The approval step
The plan screen is the product.
Ask for a pitchbook, get an outline. Reorder it, cut it, argue with it. Then it composes.
Approve, then compose
The planner returns an outline and stops. Nothing expensive runs until you agree.
Checked by looking at it
We render the deck and look at it before handing it over. Overflow, dead space, bad contrast.
On-brand by construction
House style goes in at build time. Nothing to snap into compliance afterwards.
House style
On-brand by construction.
Everyone else snaps a finished deck into compliance. We put the palette and the template into the build, so there’s nothing to correct.
What people say
The people who sign off the deck like it most.
I used to find out the story was wrong on page thirty, at midnight. Now I disagree with an outline for four minutes and the deck gets built once.
Every deck comes out in our template. Nobody on my team has opened the brand guidelines PDF in months.
The approval step sounds like an extra click until you count how many decks you've thrown away.
Honest comparison
What we do, and what we don’t.
Everyone sells on-brand and traceable. Nobody sells the step before either.
| Capability | Bankerdeck | Hebbia, Auxi, Gamma, Prezent |
|---|---|---|
| Approve the storyline before slides are built | Yes | One-shot: describe, generate, then fix |
| House style applied at build time | Yes | Snapped on afterwards as a compliance pass |
| Deck rendered and inspected before completion | Yes | Ships whatever it generated |
| Deck-level audit trail | Yes | "A file was produced" |
| Edit by instruction, most edits with no model call | Yes | Regenerate the slide and hope |
| Comps median and sources & uses computed, not recalled | Yes | Whatever the model remembered |
| Builds from your model, filing or memo | Yes | Prompt-only, or a research corpus you rent |
The leaders embed in PowerPoint and inherit its muscle memory. An add-in operates on a deck that already exists, which is exactly the point at which a storyline mistake has already been paid for.
Security
Built for people who have to stand behind the pack.
Client material leaves the building with your name on it. Every deck change is on the record: which job ran, on whose instruction, and from which sources.
Tenant isolation on every route
Every API route pulls org_id off the authenticated user; data is scoped to your organization at the query, not the application layer.
Deck-level audit trail
Every deck change records which job ran, on whose instruction, from which attached sources, and whether it landed.
Your deck is not a training set
Your decks and source documents are used to answer your prompts and nothing else. No training on customer data.
Scoped API keys
Machine access runs through revocable, org-scoped keys, metered and logged like any other actor.
Who it’s for
From one deck to a house standard.
Solo banker
Turn a filing and a model into a book without waiting on a designer.
Start freeDeal team
One thread per deck, so the storyline is agreed once and built once.
Start freeFirm
House style enforced at build time, plus the trail of who asked for what.
Talk to usResources
Win more of the work you already pitch for.
8 min read
Why the storyline gets approved before the pitchbook is built
Every deck tool automates the formatting half. The expensive half is the argument, and on a pitch it is decided in ten minutes by an MD who never opens PowerPoint.
6 min read
From the operating model to the IC memo
The numbers already exist in the model. What the memo adds is a case. Re-keying the outputs into slides is where the errors get in.
5 min read
Comps tables that compute, not remember
A median an LLM recalled is silently wrong on a client page. Why trading-comp arithmetic belongs in deterministic code and never in a model's head.
7 min read
The credit memo, drafted the same afternoon
Covenant tables, sources and uses that balance, and a downside case, assembled from the filing and the model rather than last deal's template.
6 min read
LP updates when every quarter looks the same
Twenty portfolio companies, one house format, four weeks after close. Where the time actually goes, and which parts are genuinely repeatable.
5 min read
Why output in your house template is the whole bar
A deck that is 90% on-brand costs more to fix than one built from the master. What .potx fidelity means beyond matching the colours.
Questions we get asked
Is this a PowerPoint add-in?+
No. Auxi and the other add-ins operate on a deck that already exists, which is after the expensive mistake has been made. We work on the storyline first, then compose. You get a real .pptx out whenever you want one.
What exactly do I approve?+
An outline: the slide order, what each slide argues, and its title. The planner stops there and waits. Nothing expensive runs until you have agreed what the deck says.
Can it use our template?+
Yes. Extract the style from a deck you already send to clients, or pick one of twenty institutional palettes. It goes in at build time, so slides are drawn on-brand rather than corrected into compliance afterwards.
Do you check the deck before handing it over?+
The produced file is rendered and looked at before the job reports done. Overflow, dead space, unreadable contrast. Generation is not the last step.
Do you have a research corpus or market data?+
No. Hebbia and Rogo have PitchBook, Capital IQ and FactSet. We read the filings, models and notes you attach.
Ask for a pitchbook.
Read the outline first.
That pause is the whole pitch. Everything after it is formatting.
Start a deck