Work in Progress
This document is a draft and not ready for use. Do not reference for decision-making yet.
Palm: 12-Month Product Vision¶
What Palm needs to ship to be a commercial success
The Thesis¶
Palm's current positioning: "AI native cash forecasting that sits on top of TMS/ERP"
Where Palm needs to be in 12 months: "The treasury decision platform that tells you not just what your cash will be, but what you should do about it"
The shift is from visibility → action.
Genuinely Hard, Unsolved Problems¶
These are problems no one in the market has cracked well. Solving them creates defensible differentiation.
1. Scenarios & Assumptions Studio¶
The problem: Treasurers don't just want a forecast - they want to stress test it. "What if collections drop 10%?" "What if this payment doesn't happen?" Currently they export to Excel to do this.
Why it's hard: Need to layer user assumptions on top of ML predictions without breaking the model. Need to preserve patterns while applying percentage adjustments.
Customer evidence: - Personio: Wants ±% flex on categories, seasonality adjustments - Instacart: Budget reconciliation scenarios - Treasury Dragons poll: "What-if scenarios" called out as key industry gap - Levi's: Push-button forecast with assumption layers
What "done" looks like: - Apply percentage adjustments to specific categories/entities - Model "what if this payment doesn't happen" instantly - Compare base vs. pessimistic vs. optimistic scenarios side-by-side - Layer assumptions while preserving ML patterns underneath - Save and share scenarios with stakeholders
2. Investments & Liquidity Management¶
The problem: Treasury teams manage both operational cash and investments, but these live in separate systems. They can't see total liquidity or make informed decisions about what to invest.
Why it's hard: Investment data comes from different sources (money market portals, custody banks, internal spreadsheets). Need to model maturities, rates, counterparty limits.
Customer evidence: - ON: Building operational + investments chart for Christoph presentation; $350M idle cash unlocked - Personio: Investment tracking + connectivity; manual G-sheets prone to error - Tom (Personio): "The upkeep of this sheet is quite manual... prone to error"
What "done" looks like: - Unified view: operational cash + investments + maturities - Investment ladder / maturity calendar - "Available for investment" calculation based on forecast + minimum cash - Rates, counterparty exposure, credit ratings tracking - Automatic detection when investments mature back into operational cash
3. Debt Management¶
The problem: Companies with significant debt have complex tracking needs - multiple facilities, tranches, covenants, securities, floating rates tied to benchmarks. Currently managed in spreadsheets with high error risk.
Why it's hard: Debt structures are complex (syndicated loans, multiple tranches, penalty rates, securities). Need to calculate interest accruals with changing benchmark rates. Must reconcile with bank statements.
Customer evidence: - Avramar: "We have many different securities... it is very important to put them all in a very nice system and press a button" - Avramar: Syndicated loans with 6+ different tranches, each with different amounts and rates - Avramar: "The calculation is easy, but then you have complexities - delays, penalties, changes in EURIBOR"
What "done" looks like: - Facility/loan register with all terms (amount, rate, maturity, covenants) - Automatic interest accrual calculations (including floating rate adjustments) - Drawdown and repayment tracking - Covenant compliance monitoring - Securities/collateral tracking across loans - Integration with cash forecast (scheduled repayments as known outflows) - Reconciliation with actual bank movements
4. Forecast Explainability & Confidence¶
The problem: "Can I trust it?" is the universal question. Treasurers can't explain to their CFO why the forecast says what it says. ML is a black box.
Why it's hard: Need to make complex model selection transparent without overwhelming non-technical users. Need to show confidence intervals in a way that's actionable, not academic.
Customer evidence: - ON: "Can I trust it?" - the fundamental question from Lucia - Ferguson (Treasury Dragons): "AI is a tool - do not trust it implicitly" - Gurjit: "It doesn't matter how accurate your forecast is if you're not confident in how it was generated" - ON's Federico: Building WMAPE tracking in Excel because Palm doesn't have it
What "done" looks like: - Model performance dashboard with simple → detailed slider - Category-level accuracy tracking (which categories forecast well vs. poorly?) - Forecast vs forecast comparison (did recategorization help?) - "Here's why we predicted this" explanations in plain English - Confidence intervals that translate to actionable ranges - Historical accuracy trends over time
5. Intercompany Forecasting¶
The problem: IC transactions are excluded from forecasts because they net to zero at global level. But at entity level, variance looks wrong. Long-term, treasurers want to forecast IC for operational planning.
Why it's hard: IC must net to zero across counterparties - you can't forecast one side without the other. Need to handle the pairing problem.
Customer evidence: - Amanda (ON): "Long term strategy should be to somehow forecast intercompany because otherwise we need to add that manually" - Emma: "The tricky part is the counterparties and forecasting in pairs" - ON: Entity-level variance looks wrong without IC forecast
What "done" looks like: - IC forecasts that automatically net to zero across counterparties - Clear entity-level visibility when IC is included - Differentiate sweeps (automatic) from intentional IC payments - Manual IC forecast input that enforces zero-sum constraint - Variance analysis that can include or exclude IC
6. Proactive Reminders & Pattern Detection¶
The problem: ML can detect patterns in historical data, but doesn't proactively remind users about expected recurring items that haven't appeared yet.
Why it's hard: Need to distinguish "this payment is late" from "this payment isn't happening this year". Need to avoid alert fatigue.
Customer evidence: - Tom (Personio): "Machine learning based on last year to say, Tom, in March last year there was this payment. Don't forget, it's not in the system yet" - Rodrigo (ON): "If we see something required every couple of months, we can input it in advance" - Volvo Cars: Wanted "discrepancy reporting" - report deviations vs overwriting
What "done" looks like: - "Expected but not seen" alerts for recurring patterns - Configurable thresholds (don't alert for small variances) - Distinguish between "late" and "not happening" - Let users teach the system about known recurring items - Pattern detection for biweekly, monthly, quarterly, annual items
7. Direct ↔ Indirect Bridging¶
The problem: Treasury forecasts (direct method - bank transactions) don't reconcile with FP&A budgets (indirect method - P&L). CFOs want to understand why cash differs from budget.
Why it's hard: Need to map transaction categories to budget line items. Working capital movements (AR/AP timing) create differences. Requires collaboration between Treasury and FP&A.
Customer evidence: - Personio: Wants to compare Palm forecast vs. FP&A budget visually - Euroports: "Ideally we would have the working capital movement" - Levi's: "Permanent vs timing differences" concept
What "done" looks like: - Category mapping: Palm categories → FP&A budget lines - Visual bridge showing forecast vs. budget with variance drivers - Working capital impact visualization (AR/AP timing effects) - Drill-down from variance to underlying transactions - Exportable for board/CFO reporting
8. AP/AR Data Integration¶
The problem: Bank statements show what happened. AP/AR data shows what's coming. Without open items, forecasting is guessing.
Why it's hard: Every company's ERP is different. Data quality varies wildly. Need to predict when invoices will actually be paid (not just due date).
Customer evidence: - ON: "When we have BigQuery integration... especially the AP data" - Treasury Dragons: "The biggest pain is gathering all the data into one platform" - TIS presentation: AI for analyzing customer payment behaviors
What "done" looks like: - Automated AP/AR data ingestion from major ERPs - Open items visibility in forecast - Payment behavior prediction (when will this customer actually pay?) - Aging analysis and collection risk flagging - Reconciliation: match bank transactions to open items
9. Access & Entitlements¶
The problem: Enterprise customers need multi-user deployments with different access levels. Commercial scaling requires feature gating for different tiers. This is the foundation for scaling ACVs from $50K to $200-400K with large enterprise logos.
Why it's hard: Need granular permissions (entity-level, action-level) without complexity explosion. Feature entitlements must integrate cleanly with existing architecture.
Why it matters commercially: - Land: Enterprise procurement requires RBAC, audit trails, and SSO. Without these, we can't pass security reviews. - Expand: More users per customer = higher ACV. ON wants 3 regional users today, could scale to 10+. Feature tiers enable upsell motions. - Retain: Customers with more users are stickier. Regional teams become stakeholders, not just central treasury.
Customer evidence: - ON: "We would love that [access control]... With Kyriba, we feel like we had so many training sessions, every two weeks. Palm is way more intuitive." Want to share Palm with 3 regional hubs (APAC, Americas, EMEA) but need entity-level permissions first. - ON: Lucia after visiting APAC: "We need to enable them way more than we do." Regional teams need visibility without overwhelming them.
What "done" looks like: - Role-based access control (regional treasurer sees only their region) - Entity-level visibility restrictions - Read-only vs edit permissions - Feature tiers (Basic/Premium/Enterprise) - Feature gating that enables upsell motions - Audit trail of user actions - SSO integration (enterprise requirement)
Competitive Positioning¶
| Problem | Kyriba | TIS | Panax | Palm (Today) | Palm (12mo) |
|---|---|---|---|---|---|
| Scenarios | Basic | Yes | Limited | No | Full |
| Investments | Module | Limited | No | Basic | Full |
| Debt Management | Module | No | No | No | Full |
| Forecast Explainability | No | No | Some | Limited | Full |
| IC Forecasting | No | No | No | Excluded | Paired |
| Proactive Alerts | Basic | Basic | Some | No | Smart |
| Direct/Indirect Bridge | No | No | No | No | Yes |
| AP/AR Integration | Via TMS | Yes | Limited | No | Native |
| Access & Entitlements | Full | Full | Basic | Limited | Full |
Commercial Success Metrics¶
| Metric | Current | 12-Month Target | Long-Term (24mo) |
|---|---|---|---|
| Paying customers | ~5 | 15-20 | 30-50 |
| ACV | Variable | $50K+ average | $200-400K (enterprise) |
| NRR | Unknown | >120% | >130% |
| Time to value | Weeks | <4 weeks | <2 weeks |
| Customer ROI case studies | 1 (ON) | 3-5 with $ values | 10+ |
| Users per customer | 2-3 | 5-10 | 10-20 (regional teams) |
Path to $200-400K ACVs: 1. Land: Win enterprise logos with core forecasting + visibility 2. Expand users: Enable regional teams with access control (3 → 10+ users) 3. Expand features: Upsell scenarios, investments, debt modules via feature tiers 4. Expand scope: Add more entities, more banks, more integrations
The Narrative¶
For sales:
"Palm isn't just a forecasting tool - it's your treasury co-pilot. It tells you what's coming, explains why, lets you stress-test scenarios, and connects your cash, investments, and debt in one view. Other tools give you data. Palm gives you decisions."
For product:
"We're solving the problems that make treasurers stay in Excel - flexibility, explainability, and the ability to answer 'what if?' We're not building another TMS. We're building the intelligence layer that makes every treasury tool better."
For enterprise expansion:
"Start with central treasury, expand to regional teams. Palm is intuitive enough that regional finance teams can use it without the training burden of Kyriba. More users = more value = higher ACV. Access control unlocks this."
Last updated: 2026-02-04