Finance · Financial Analyst
A live financial forecast, built from your own books.
Every month it derives a driver-based forward model straight from your accounting and billing data — then gives you a plain read on cash, runway and what moves them. It works for any business — SaaS, services, agency, retail, e-commerce, manufacturing and more. A second AI from a different vendor re-reads the narrative before you see it, and you approve it before it sends. It is read-only on every connected tool: it never posts an entry or closes a book.
From the books closing to a narrative in your approval queue
Books close
A monthly refresh fires from QuickBooks or Xero.
Model derived
Drivers from the ledger; cohort curves too, for subscription businesses.
Scenarios run
The scenario library runs; a drift narrative is drafted.
You approve
A second AI re-reads it; then it waits for your sign-off.
Books not closed within 35 days? A books-quality gate refuses to run rather than narrate a forecast on stale numbers — it tells you the gap instead. Need a read between refreshes? Ask /runway in Slack any time.
The contract
What it runs on its own, what it checks with you, what it won't touch.
Runs on its own
- Auto-derives a driver-based forward model from your QuickBooks or Xero ledger every month — revenue, cost, margin, cash and headcount, whatever your business sells.
- For subscription businesses, fits cohort-retention curves from your real Stripe history — no manual cohort spreadsheet.
- Overlays live macro benchmarks — the Fed funds rate, sector growth and burn norms — so the model is grounded in outside context, not just your last 12 months.
- Runs a scenario library on every refresh — base, conservative, aggressive, plus any custom scenario you have saved.
- Surfaces the top-3 assumptions driving your forecast, each with the concrete number it moves.
- Narrates the drift between each assumption and the actuals in plain English, with the source figure behind every claim.
- On the upper tiers, drafts your quarterly board pack from the same model — a 12-14 page PDF, narrative and charts, with every dollar and percent cited to a source transaction.
Checks with you first
- The monthly drift narrative queues for your approval before it sends — you approve it, edit it, or reject it. Nothing leaves on its own.
- A natural-language assumption change shows you a preview of the parsed deltas before it commits to a scenario.
- When a macro-overlay benchmark swings more than 20% from your own number, that assumption routes to you instead of being applied silently.
- If the books-quality gate blocks a refresh, it surfaces the specific gap and waits — it does not run a forecast on numbers it does not trust.
Won't touch
- Close your books, post a GL entry, or modify any ledger — it is read-only on every connected tool.
- Recategorize transactions or do bookkeeping — it reads your books, it does not fix them.
- Tax filings, sales-tax, or 1099s — it is a forecasting tool, not a tax preparer.
- Investor-facing fundraising material — the drift narrative is for internal board context only.
- Email your customers, or take any action outside the dashboard and Slack.
Multiple legal entities? Multi-entity consolidation is a Business-tier feature and up — Solo and Team forecast a single entity.
Full capability matrix from the role registry
PROFICIENT
- Driver-based forward model auto-derived from QB + Stripe + HubSpot. Revenue model auto-shapes to your data: cohort × ARPU × retention for SaaS, repeat-purchase rate × AOV for e-commerce, recurring-contract retention for services. OpEx model from headcount × loaded cost + tools + marketing efficiency. 3-statement linked through AR-DSO, AP-DPO, deferred revenue. Driver-derivation runs nightly; operator can override per-driver in plain English.
- Append-only versioned assumption ledger with audit trail. Every assumption ('hiring pace 2/quarter', 'Q3 enterprise close rate 22%', 'churn 3.5%') is a row in an append-only ledger — never updated, never deleted. Each edit creates a new version with author, timestamp, prior value, and the rationale you typed. The model rebuilds from the current assumption-version set on every refresh. CLAUDE.md principle #2 made operational for the Financial Planning & Analysis surface.
- Cohort retention curve auto-fit from Stripe actuals. Reads Stripe subscription events, computes per-cohort retention curves (monthly cohort × month-since-signup → retained %), fits a decay model, and uses the fitted curve as the forward retention driver. Non-SaaS equivalents: repeat-purchase rate decay from QB invoice patterns for e-commerce; recurring-contract retention for services. Operator can override the fitted curve.
- Macro auto-overlay — Fed rate + wages + sector benchmarks. Daily refresh from public sources: Fed funds rate (cost-of-capital input for runway-extension decisions), Carta + Levels.fyi wage benchmarks by role/level/region (headcount cost assumption sanity), SaaS Capital + OPEXEngine sector growth / burn-multiple / NRR norms for SaaS, NFIB SMB index for non-SaaS. The model surfaces every assumption against its macro-overlay range — 'your Q3 enterprise close-rate assumption 28% sits at the 90th percentile of SaaS Capital benchmark — verify or revise.'
- Scenario library — base / conservative / aggressive + custom. Three pre-built scenarios run on every model refresh (base = your assumptions; conservative = −20% growth / +20% costs; aggressive = +20% growth / +20% costs). Custom scenarios accept natural-language input ('what if I raise $2M at 15% dilution and hire 4 engineers') — the scenario builder parses the prompt into assumption deltas you preview before the rerun. Every scenario saves with a name + creation timestamp + cited assumption changes.
- Sensitivity analysis — top-3 assumptions driving runway. On every refresh, the model computes which 3 assumptions would shorten runway most if they moved by ±10% — surfaced as a ranked list with delta-runway in weeks per +10% / −10% move per assumption. The founder sees not just 'runway 47 weeks' but 'runway is 64% driven by your Q3 enterprise close-rate assumption — if that drops 5pts you lose 8 weeks.'
- Monthly AI-narrated assumption-vs-actuals drift. On the day books close (≤35 days post month-end), the Tier-3 cross-family supervised worker narrates the drift between every assumption and what actually happened in plain English. 'CAC payback slipped 14→17 because Q1 cohort retention dropped 2pts; the macro benchmark for your stage is 14 — your prior assumption was at the benchmark, the actual is below it.' Every claim cites a source GL transaction or Stripe event. Lands in /app/fpa + Slack #finance digest channel.
- Slack /runway slash command + Slack-native model Q&A. Real-time read from the latest model state. /runway returns the current runway + the macro context + the top-3 sensitivity-driving assumptions. Free-form Slack Q&A ('what's our gross margin trend over the last 6 months?', 'how much would runway extend if I cut $50K headcount?') returns answers grounded in the live model + actuals. Quoted answers always cite the relevant model row + the source transaction chain.
- 13-week rolling cash forecast with P10/P50/P90 confidence bands. Pulls cash actuals from QB + Stripe + Mercury + Brex + Ramp (read-only); projects the next 13 weeks of cash-in / cash-out from the driver model + AR-DSO + AP-DPO + scheduled payroll. Confidence bands derived from scenario library (P10 = conservative, P90 = aggressive). Surfaces in /app/fpa + emits a weekly Slack digest every Monday 07:00 local-time with the next-13-week chart.
- Append-only audit log + Tier-3 supervisor on monthly narrative. Every assumption change, scenario rerun, model refresh, and narrative emission is logged to the append-only audit_log table with tenant_id, user_id, timestamp, metadata. The monthly drift narrative routes to Tier 3 cross-family supervisor (Anthropic worker → OpenAI reviewer → Gemini tiebreaker on disagreement) — the narrative is the externally-visible artifact, so its guardrails sit at the strictest tier in the marketplace.
- Books-quality gate — refuse-to-run if books not closed in 35d. At onboarding and on every monthly refresh, the worker checks whether the customer's QB/Xero books are closed within the last 35 days, reconciled, and categorized above a minimum quality bar. If not, refuse to run the model — surface the specific gaps + recommend a bookkeeper (Pilot / Bench-refugee / Puzzle partner list). Brand-defense discipline: a narrated drift on garbage books is a credibility bomb, not a deliverable.
- Read-only by API-scope across every connected vendor. QuickBooks / Xero / Stripe / HubSpot / Mercury / Brex / Ramp tokens carry NO write scopes. Cannot post a journal entry, cannot recategorize a transaction, cannot modify a subscription, cannot move a deal stage, cannot move money. Read-only at the API-scope layer, not the prompt layer — schema-enforced refusal.
ASSISTED
- Macro-overlay variance > 20% triggers operator review. When the model finds any assumption sitting outside ±20% of the macro-overlay benchmark for your stage / sector / region, the monthly narrative banners it in /app/fpa and the worker pauses the auto-send pipeline for that month. Operator either revises the assumption or approves the overlay-variance with a rationale that audit-log records.
- Three-way supervisor disagreement — operator review. When worker / supervisor / tiebreaker all disagree on a narrative claim (e.g., direction of the drift, magnitude of the assumption miss), the narrative row routes to operator review in /app/fpa/inbox instead of auto-resolving. Today: pick the worker's verdict, log all three in audit_log, surface the conflict in the per-narrative detail. ≥50 instances needed before a deeper investigation.
- Natural-language assumption updates — operator preview before commit. When the operator types 'bump enterprise close rate to 25%' or 'churn assumption goes to 3.8%', the worker parses the language into a structured assumption-delta + previews the runway / ARR impact in /app/fpa before the operator confirms commit. Audit-log records the prompt + the parsed structured delta + the operator's confirmation.
- Monthly narrative review queue — operator approves before send. The Tier-3 monthly narrative lands in a review queue in /app/fpa/inbox by default. Operator can edit, approve, or reject before it ships to the Slack #finance channel or is included in the Board Pack Drafter's monthly internal update. Auto-send opens after the first 90 days once the operator has approved ≥6 narratives without material edits.
REFUSED
- Investor-facing fundraising material — internal board only. FTC / SAFE / 506(b) risk discipline. We never draft a fundraising pitch deck, never draft a SAFE, never draft any material that goes to a current/prospective investor in a fundraising context. We draft INTERNAL board updates (post-funding, post-info-rights). Board Pack Drafter (boris-v0) consumes our model output strictly for internal board communication. Schema-enforced — output type is 'internal_board_update', never 'fundraise_doc'.
- Opaque score outputs without underlying transaction citations. Every quantitative claim in any narrative, model output, or Slack answer cites a source transaction chain (stripe_invoice_X, qb_transaction_Y, hubspot_deal_Z) — never a single 'health score: 0.34' or 'risk index: 7'. Post-parse validator drops outputs that fail the citation check. The citation IS the unlock for finance trust.
- Modifying the chart of accounts or recategorizing transactions. The chart of accounts (CoA) is the bookkeeper's domain. We READ the CoA + the categorized transactions, we never WRITE to it. If we find a transaction that looks miscategorized (e.g., personal expense in COGS), we surface it in /app/fpa as a recommendation for the bookkeeper — never recategorize ourselves. API-scope + schema enforcement.
- Multi-entity consolidation (V1 single-entity; Phase 2 enterprise). V1 ships single-entity modeling — one QB/Xero file per tenant. Multi-entity consolidation (parent + subsidiaries + intercompany eliminations + currency translation) opens as Phase 2 enterprise after the first 25 paying SMB customers, per ADR-0031 §Enterprise Phase 2. Refusing in V1 protects model quality — multi-entity is a hard modeling problem we won't half-ship.
- Customer-facing outreach — never emails / Slacks your customers. AI Financial Analyst is an INTERNAL worker for your founder / operator / finance team. Never sends a message to YOUR customers (overdue invoice chasing is Renewal Hunter v2 territory; customer-success outreach is RH + AHW). Output channels: /app/fpa (web UI), Slack #finance (internal channel), Board Pack Drafter (internal board update). No external send paths.
- Closing books, posting GL entries, modifying any ledger. AI Financial Analyst is NOT in the bookkeeping business. Books are closed by Pilot / Bench-refugee / Puzzle / your human bookkeeper. The output schema has no journal-entry field; the API tokens for QB/Xero carry no write scopes. Hard refusal at the API-scope + schema layer, not the prompt layer.
- Tax filings, sales-tax computation, 1099 generation. Regulated / UPL-adjacent territory. We never file a tax return, never compute sales-tax owed, never generate a 1099. Surface a recommendation list of tax software (Avalara / TaxJar / Anrok / your CPA) when the operator asks; refuse the action itself. Founding-doc principle: methods, not credentials — we describe our methods, never claim a license.
What you get
Not a memo — a working model you open, plus the read on what changed.
Three concrete deliverables, every month. Every figure traces back to your own ledger — or your billing data, for subscription businesses — the narrative is re-read by a second, independent AI before you see it, and nothing is final until you approve.
Always on
A live model at /app/fpa
Cash, runway and monthly burn or build rate on one screen — derived from a driver-based 12-month forward that is re-built from your books. Open it any day, not just at month-end.
Every month
A cited read on what changed
Why the forecast moved, in plain words — every number traceable to a ledger or billing row, queued for you to approve, edit or reject.
Export & ask
Scenarios, Excel and Slack
A base / conservative / aggressive scenario library plus your own saved ones, an Excel export of the drivers and scenarios, and runway on demand in Slack with /runway.
May refresh · books closed May 31
Drafted — awaiting your approvalRunway — cash on hand ÷ current burn
11.2 months
1.4 months shorter than April — was 12.6
Projected cash · next 12 months
What moved it this month
Retention, not spend. The January cohort renewed at 94% against a modelled 101%, so net revenue retention came in below plan. Headcount cost tracked the model within 2%, and your cost of capital still sits under the current Fed funds benchmark — no spending flag this month.
Traces to: January-cohort renewal rate from Stripe subscription history · modelled NRR from the prior driver-model run · Fed funds benchmark from the live macro snapshot.
Change one assumption — runway re-runs
- Hire the 3 planned engineering roles runway → 9.4 mo
- The next raise slips a quarter runway → 8.9 mo
- January-cohort retention recovers to plan runway → 13.1 mo
Every figure traces to a ledger or Stripe row, and the whole brief is re-read by a second, independent AI from a different vendor before it reaches you — nothing is final until you approve.
May refresh · books closed May 31
Drafted — awaiting your approvalCash on hand — and the monthly build rate
$612K
building +$34K/mo — half of Q1's +$71K pace
Projected cash · next 12 months
What moved it this month
Margin, not demand. Revenue grew 6% on the quarter, but gross margin slipped about 2 points — materials ran over plan on two large installs. Cash still builds; the pace just halved from Q1. Payroll tracked the model within 3%.
Traces to: revenue and materials cost from the May ledger close · prior-quarter margin from the last model run · payroll from the general ledger.
Change one assumption — the forecast re-runs
- Win the two pending maintenance contracts year-end cash → $1.0M
- Materials cost holds at plan year-end cash → $940K
- Add a third install crew in Q3 year-end cash → $620K
Every figure traces to a ledger row, and the whole brief is re-read by a second, independent AI from a different vendor before it reaches you — nothing is final until you approve.
Same brief, downloadable: an Excel export of the drivers and scenarios, and a Slack /runway command for the number on demand. Every month's assumption set is kept in an append-only ledger — what the model assumed, what the actuals were, and what you approved is always traceable.
On the upper tiers
Board-ready, from the same numbers
On the Business and Practice tiers the Financial Analyst also drafts your quarterly board pack — built on the model it already maintains, so the deck and the forecast can never disagree. You approve every section before it is final; nothing is ever sent to investors automatically.
Quarterly deck
Draft the pack
A 12-14 page board-pack PDF — narrative and charts, coherent quarter to quarter — with every number cited to a source transaction in your model.
Before the room
Q&A rehearsal
A pre-board rehearsal of about a dozen tough questions with cited answers, roughly 48 hours before the meeting, plus a monthly internal update between packs.
Across packs
Promise Ledger
Tracks every commitment made to the board across packs and re-grades it next quarter — you mark each line met, missed, or re-scoped.
The board pack is reviewed independently by a second AI vendor before it reaches your queue, and there is no external send path — the pack stays internal until you approve and share it.
Pricing
Three tiers by entity count and seats.
Every tier gets the same engine — the auto-derived driver model, the macro overlay, the cohort fit for subscription businesses, and the second-vendor AI review on every drift narrative.
What it replaces
Doing this in-house means a full salary, benefits, and months of ramp — for one function. Ataski ships the output from day one at a flat, predictable monthly price.
Enterprise-grade quality for a fraction of a hire. Scale up or pause anytime — you only pay for the months you use.
Solo
$599 / month
1 entity · 1 user
≈ $599.00 per entity per month
- Driver-based forward model, auto-derived monthly
- Cohort-retention fit — subscription businesses
- Cross-vendor reviewed drift narrative · Slack /runway
A founder running a single entity who wants a real forecast off real books.
Choose SoloTeam
$1,499 / month
3 entities · 5 users
≈ $499.67 per entity per month
- Everything in Solo, across 3 entities
- Scenario library + top-3 sensitivity
- Natural-language custom scenarios · custom macro upload
A finance lead modelling a few entities who needs scenarios on demand.
Choose TeamBusiness
$2,999 / month
5 entities · 20 users
≈ $599.80 per entity per month
- Everything in Team, across 5 entities
- Multi-entity consolidation
- SSO · audit-log export · custom retention · DPA
- Quarterly board pack + pre-board Q&A rehearsal + Promise Ledger
Mid-market finance with procurement — SSO and a DPA are sign-off requirements.
Choose BusinessPractice — fractional CFO
$399 / mo base + $199 per entity-month
$399/mo base + $199/entity per month · multi-tenant sub-orgs · designed for fractional CFO practices running 4-10 client books on a single Ataski seat. Up to 10 entities; usage metered through Stripe.
Includes the quarterly board pack for every client book — deck, Q&A rehearsal, and Promise Ledger, built on each entity's own model.
Per-entity billing wires through your account manager. Self-serve checkout flips on once Stripe usage reporting is wired.
Enterprise — beyond 5 entities
For unlimited entities, intercompany eliminations, and FX translation, or when procurement needs SAML, bring-your-own LLM keys, regional data residency, and DPA redlines. Sold inbound.
Setup
Three steps — then it runs itself every month.
-
01
Connect your books and billing. QuickBooks or Xero for the ledger, Stripe for subscription cohorts (optional) — read-only access on every one.
-
02
Confirm the model. It detects your business shape from the data and derives the first model — you review the drivers it picked.
-
03
Wait for the close. Each month the refresh fires once your books close; the narrative lands in your approval queue.
Under the hood
How the forecast is built, and what reviews it.
- Builds the model
- A deterministic engine derives the driver-based forward model — and fits cohort curves for subscription businesses; the math is plain code, never a language model
- Writes the narrative
- An AI turns the run into plain-English prose — it explains the numbers, it does not compute them
- Reviews the narrative
- A second, independent AI from a different vendor re-reads every drift narrative; a tiebreaker decides when the two disagree on direction
- External data
- The live Fed funds rate from the Federal Reserve's public feed, plus sector benchmarks you can upload — overlaid on every refresh
- Access
- Read-only on QuickBooks, Xero, and Stripe — it cannot post an entry, close a book, or change a ledger
- Privacy
- Private to your workspace, every run and assumption in an append-only audit ledger, 30-day data deletion on offboarding
A forecast that rebuilds itself — from your books, every month.
Connect your books and billing once. Each month the model is re-derived, the scenarios run, and a plain read on cash, runway and what moves them lands in your approval queue — re-read by a second AI before you ever see it.