Enterprise Extensions: Operations and Strategy
In the previous lessons, you built finance agent capabilities that work across organisations — model comprehension, scenario testing, variance analysis, and the financial-services plugin suite. Each of those capabilities applies general financial logic. Now you confront the limitation: general financial logic does not know your treasury policy, your tax provision methodology, your CFO's preferred variance bridge format, or your fund's NAV calculation rules. This lesson covers seven enterprise extension domains where the gap between generic capability and organisation-specific requirement is widest.
The pattern is consistent across all seven. Identify what the generic plugin lacks. Name the institutional knowledge the extension must encode. Write the SKILL.md instructions that close the gap. The methodology is the same one Chapter 16 taught — what changes is the professional context in which you apply it.
Extension 3: Treasury and Cash Management
A generic finance plugin produces financial statements and variance reports. It does not know which banks hold which accounts, what sweep mechanics apply, what minimum operating balances are required by currency, or what your treasury policy says about hedging ratios and instrument selection.
A treasury extension encodes your organisation's actual cash structure and applies your treasury policy when monitoring FX exposures and hedging positions.
Concept: Treasury Management
Treasury manages three interconnected workflows. Cash positioning tracks daily balances across all bank accounts, currencies, and entities to ensure the organisation can meet its obligations. FX exposure management identifies currency risks from cross-border receivables, payables, and intercompany loans, then hedges them using instruments specified in the treasury policy — forwards, options, or cross-currency swaps. Bank relationship management maintains the structure of accounts, sweep mechanics, and credit facilities across the organisation's banking partners.
Without these specifics encoded in the SKILL.md, a treasury agent applies generic logic that may violate the firm's actual risk mandate or board-approved hedging policy.
Key SKILL.md instructions to write:
- Hedging policy rules: Which exposure categories are hedged, minimum exposure size triggering a hedge, acceptable instruments by exposure type, and maximum tenor per instrument class
- Banking relationship structure: Which banks hold which accounts, sweep mechanics (overnight sweeps, notional pooling, zero-balance accounts), and minimum operating balances by currency
- FX rate source: Whether positions are valued at mid-market, bid/offer, or a specific fixing — this may differ from the rate used in management accounts
- Intercompany loan structure: Which entities lend to which, currencies, interest rates, transfer pricing methodology, and maturity schedule
Extension 4: Tax Provision and Compliance
Quarterly tax provision under ASC 740 (US GAAP) or IAS 12 (IFRS) is one of the most technically demanding recurring tasks in corporate finance. The generic plugin cannot calculate your effective tax rate because it does not know your jurisdictional mix, your permanent and temporary differences, or your deferred tax asset valuation methodology.
Concept: What Is a Tax Provision?
A tax provision estimates the income tax a company owes for a period. The effective tax rate (ETR) is total tax expense divided by pre-tax income — it differs from the statutory rate because of permanent differences (items never taxable or deductible), temporary differences (items taxable or deductible in different periods), and credits. Deferred tax assets arise when a company pays more tax now than the accounting expense recognises (future benefit). Deferred tax liabilities arise when accounting expense exceeds current tax payable (future obligation). ASC 740 governs how US companies calculate, recognise, and disclose income taxes — including uncertain tax positions and valuation allowances for deferred tax assets.
Key SKILL.md instructions to write:
- Jurisdictional tax profile: Every jurisdiction where the entity files, the statutory rate, any tax holidays or incentives, and the withholding tax rates on intercompany dividends, royalties, and management fees
- Permanent and temporary difference classification: Which items in your chart of accounts create permanent differences (stock compensation, meals, penalties) and which create temporary differences (depreciation methods, revenue recognition timing, lease accounting)
- Deferred tax asset valuation: The methodology for assessing whether deferred tax assets are more likely than not to be realised — the positive and negative evidence framework, the weighting approach, and the conditions that trigger a valuation allowance
- Uncertain tax position assessment: The two-step recognition and measurement process under ASC 740-10, your threshold for recognising a position, and the routing rule for positions that require external tax counsel review
Extension 5: FP&A and Budget Ownership
The variance analysis command calculates variances. It does not know your planning cycle, your reforecast trigger thresholds, your seasonal adjustment factors, or your CFO's preferred bridge format.
Concept: Driver-Based Forecasting
Driver-based forecasting builds financial projections from operational inputs rather than accounting line items. Instead of forecasting "revenue will be $5M," it forecasts "4,800 units at $1,042 average selling price." This decomposition — volume, price, and mix — makes variances actionable: a revenue shortfall caused by fewer orders requires a different response from one caused by deeper discounting. The variance bridge presents these components visually, showing how each driver contributed to the gap between budget and actual results.
Key SKILL.md instructions to write:
- Variance bridge format: Whether revenue variances split into volume, price, and mix; whether cost variances distinguish volume-driven from rate-driven; the presentation order and labelling convention your CFO expects
- Reforecast policy: The variance magnitude (absolute dollars and percentage of budget) that triggers a formal reforecast versus a commentary note, and which P&L lines trigger at which thresholds
- Seasonal adjustment factors: Which lines have known seasonal patterns, the adjustment factor for each, and the reasoning so the agent can explain seasonality in narrative form rather than treating it as an unexplained variance
- Forward-looking statement standards: The standard disclaimer language required on documents containing forecasts
Exercise 16: FP&A SKILL.md
What you need: 30 minutes. Your organisation's most recent budget-vs-actual report, or the practice model from Exercise 4.
- Identify three revenue or cost lines where the variance was material last quarter.
- For each, decompose the variance into its operational drivers (volume, price, mix, rate, efficiency).
- Write a SKILL.md instruction encoding your preferred bridge format: which components to name, how to sequence them, and how to translate numbers into decision-relevant language.
- Write one reforecast trigger rule: at what threshold does this variance require a formal reforecast versus a commentary note?
The key learning: The discipline is not variance calculation — Claude handles arithmetic. It is the CFO-ready framing: knowing which components to name and how to translate accounting language into the language of decisions. The SKILL.md instruction you write encodes this framing permanently.
Extension 7: M&A Integration PMO
Deal execution gets the attention. Post-merger integration determines whether the deal creates value. A generic plugin does not know your integration playbook, your synergy tracking methodology, or your Day-1 readiness checklist.
Concept: What Is a PMO?
A Project Management Office in the M&A context is the central coordination function that manages integration after a deal closes. It tracks three things: Day-1 readiness (can the combined entity operate on the first day — payroll, IT access, customer communication, regulatory notifications), synergy delivery (are the cost savings and revenue synergies promised in the deal model being realised on schedule), and cultural integration (are retention targets being met, are employee engagement scores stable, are key talent departures within acceptable bounds).
Key SKILL.md instructions to write:
- Day-1 readiness checklist: The minimum operational requirements by function (HR, IT, Finance, Legal, Operations) that must be confirmed before close, the escalation path for items not confirmed 48 hours before close, and the fallback procedures for each
- Synergy tracking methodology: How synergies are categorised (cost vs. revenue, run-rate vs. one-time), the measurement baseline, the cadence of synergy reporting, and the variance threshold that triggers a formal remediation plan
- Cultural integration scoring: The metrics used to assess integration health — voluntary attrition of key talent, employee engagement survey delta, cross-team collaboration indicators — and the scoring thresholds that trigger intervention
- Integration governance: The reporting cadence (weekly steering committee, monthly board update), the escalation matrix, and the decision rights framework for integration trade-offs
Extension 8: ESG Reporting and Metrics
ESG reporting is moving from voluntary disclosure to mandatory compliance. The frameworks are specific, the calculations are technical, and the data requirements cross every function in the organisation.
Concept: ESG Reporting Frameworks
CSRD (Corporate Sustainability Reporting Directive): The EU's mandatory sustainability reporting standard, effective for large companies from 2024. Requires reporting under the European Sustainability Reporting Standards (ESRS) with third-party assurance. Covers environmental, social, and governance topics with double materiality — both financial impact on the company and the company's impact on people and planet.
ISSB (International Sustainability Standards Board): Issues IFRS S1 (general sustainability disclosures) and IFRS S2 (climate-related disclosures). Focused on enterprise value — what sustainability matters mean for investors. Adopted or under adoption in multiple jurisdictions outside the EU.
Emissions classification: Scope 1 (direct emissions from owned sources), Scope 2 (indirect emissions from purchased energy), Scope 3 (all other indirect emissions across the value chain — typically the largest and hardest to measure). The calculation methodology, emission factors, and organisational boundary choices all require encoding in the SKILL.md.
Key SKILL.md instructions to write:
- Framework applicability: Which frameworks apply to your organisation (CSRD, ISSB, voluntary GRI, SEC climate rules if US-listed), the reporting boundary, and the materiality assessment methodology
- Emissions calculation methodology: The organisational boundary approach (equity share vs. operational control), emission factor sources, and Scope 3 category coverage with data collection methods per category
- ESG KPI definitions: The specific metrics your organisation tracks, their calculation methodology, the data sources, and the internal controls that ensure data quality
- Assurance readiness: The level of assurance required (limited vs. reasonable), the evidence trail the SKILL.md must maintain for each disclosed metric, and the internal review process before external reporting
Extension 9: Fund Administration and NAV
Fund administration is a domain where precision is contractual. The net asset value calculation determines investor returns, fee calculations, and regulatory reporting. A generic plugin does not know your fund's valuation policy, fee waterfall, or investor reporting requirements.
Concept: What Is NAV?
Net Asset Value (NAV) is the value of a fund's assets minus its liabilities, divided by the number of outstanding shares or units. For an open-end fund, NAV determines the price at which investors buy and redeem. For a closed-end fund, NAV is the reference against which market price is compared (premium or discount to NAV).
Fund administration encompasses NAV calculation, investor reporting, fee calculations (management fees, performance fees with high-water marks or hurdle rates), capital account maintenance, and regulatory filings. The fee waterfall defines the sequence in which returns are allocated — typically: return of capital first, then preferred return to investors, then carried interest to the general partner, with catch-up provisions that vary by fund.
Key SKILL.md instructions to write:
- Valuation policy: The hierarchy of valuation inputs (Level 1 quoted prices, Level 2 observable inputs, Level 3 unobservable inputs), the frequency of valuation, the treatment of illiquid positions, and the valuation committee governance process
- Fee waterfall: The exact calculation sequence — management fee basis and rate, hurdle rate, catch-up percentage, carried interest split, clawback provisions, and the treatment of recycled capital
- Investor reporting requirements: The content, format, and cadence of investor statements — quarterly NAV statements, annual audited financials, capital account statements, K-1 tax reporting (US), and ad hoc investor requests
- Regulatory filings: Form PF (US), AIFMD Annex IV (EU), or equivalent filings in your jurisdiction — the data points required, the calculation methodology, and the filing deadlines
Exercise 17: Fund Admin SKILL.md
What you need: 30 minutes. Your fund's offering memorandum or partnership agreement, or a sample LPA.
- Extract the fee waterfall from the document: management fee rate and basis, hurdle rate, carried interest percentage, catch-up provisions.
- Write SKILL.md instructions encoding each step of the waterfall as a calculation rule the agent can execute.
- Write one escalation condition for the most common fee calculation dispute in your fund type (e.g., treatment of unrealised gains in carried interest calculation).
- Design one validation scenario: provide a set of inputs (fund returns, capital contributions, distributions) and calculate the expected fee allocation manually. This becomes your test case.
The key learning: Fee waterfall errors are not rounding differences — they are contractual disputes. A SKILL.md instruction that says "calculate carried interest" is insufficient. The instruction must specify every step of the waterfall in the exact sequence defined by the partnership agreement, including the edge cases that cause disputes.
Extension 11: Insurance Underwriting and Claims
Insurance is a domain where the agent must encode actuarial judgment, not just process flows. Premium pricing, claims triage, and reserving each require institutional knowledge that no generic plugin provides.
Concept: Insurance Underwriting and Reserving
Underwriting is the process of evaluating risk to determine whether to insure it and at what price. The underwriter assesses the probability and severity of potential losses using rating variables — characteristics of the insured (age, location, claims history) that predict future loss experience. An experience modification factor adjusts the base premium based on the insured's actual loss history relative to expected losses for their class.
Reserving estimates the total cost of claims that have been reported but not yet settled (case reserves) and claims that have occurred but not yet been reported (IBNR — Incurred But Not Reported). The chain ladder method projects ultimate losses by applying historical development patterns to current claim data. The Bornhuetter-Ferguson method blends actual experience with an expected loss ratio, making it more stable for immature accident years with limited development.
Key SKILL.md instructions to write:
- Premium pricing factors: The rating variables for your book of business, the base rate tables, the experience modification methodology, and the underwriting guidelines that specify when a risk requires referral to senior underwriting
- Claims triage rules: The severity classification thresholds, the assignment rules by claim type and complexity, the escalation conditions for claims requiring specialist adjustment or legal referral, and the documentation requirements at each triage stage
- Reserving methodology: The actuarial method used (chain ladder, Bornhuetter-Ferguson, or frequency-severity), the data inputs required, the assumptions that must be documented and reviewed quarterly, and the conditions that trigger a reserve re-estimation
- Regulatory reporting: The jurisdiction-specific filing requirements, the statutory accounting principles that differ from GAAP, and the annual statement schedules that require population
Prioritising Your Extension Roadmap
Not every extension is relevant to every organisation. The prioritisation framework uses three criteria.
| Criterion | Question | Example |
|---|---|---|
| Operational pain | Where does the workflow gap cause the most daily friction? | Multi-entity consolidation taking three days each month |
| Knowledge risk | Where is expertise concentrated in individuals who could leave? | The one treasury analyst who knows the hedging policy details |
| Regulatory exposure | Where do compliance requirements create urgency? | ESG reporting deadlines with mandatory assurance |
Practical sequencing: Run three extensions in parallel in the first quarter — one for the highest-volume workflow, one for the highest-risk compliance area, one for the knowledge most at risk. Sequence the remainder across quarters two and three.
Final Exercise: Design Your Extension Roadmap
What you need: 30 minutes. A blank document.
-
Write one sentence describing your role and your organisation's finance function.
-
From the seven extensions in this lesson (plus the four from prior lessons), identify the three most relevant to your context. For each: (a) the specific gap in the generic plugin that matters most, and (b) the one SKILL.md instruction that would close the most important part of that gap.
-
Rank the three by operational pain, knowledge risk, and regulatory exposure.
-
Write a one-paragraph specification for the first extension: what the SKILL.md should do, what data sources it needs, and what three test scenarios would validate it.
This specification is the starting point of a real deployment.
Try With AI
Use these prompts in Anthropic Cowork or your preferred AI assistant to design your enterprise extensions.
Prompt 1: Extension Gap Analysis
I work as [YOUR ROLE] in [YOUR INDUSTRY/ORGANISATION TYPE].
My organisation's finance function covers [LIST KEY AREAS:
e.g., treasury, tax, FP&A, fund admin, etc.].
From these seven enterprise extension domains — treasury and cash
management, tax provision, FP&A/budget ownership, M&A integration
PMO, ESG reporting, fund administration, insurance underwriting —
help me identify the three most relevant to my context.
For each of my top three:
1. What specific institutional knowledge would the extension need
to encode about MY organisation (not generic finance knowledge)?
2. What is the single highest-value SKILL.md instruction I should
write first?
3. What validation scenario would test whether that instruction
works correctly?
Be specific to my context — I need organisation-specific gaps,
not textbook definitions.
What you're learning: How to translate abstract extension categories into concrete, organisation-specific SKILL.md requirements. The gap between "treasury management extension" and "encode our EUR/GBP hedging policy with 75% minimum hedge ratio on committed exposures using forwards only" is the gap between a concept and a deployable instruction. This prompt forces you to cross that gap for your specific context.
Prompt 2: SKILL.md Instruction Drafting
I am building a [EXTENSION TYPE] extension for my organisation's
finance agent. Here is the context:
- Organisation type: [e.g., mid-cap manufacturing, PE fund, bank]
- Key workflow: [e.g., quarterly tax provision, daily cash positioning]
- Current pain point: [e.g., takes 3 days, concentrated in one person]
Help me draft four SKILL.md instructions for this extension. Each
instruction should:
1. Encode a specific organisational rule (not generic finance logic)
2. Include an escalation condition for edge cases
3. Be testable — I should be able to design a scenario that passes
or fails against the instruction
After drafting, critique each instruction: is it specific enough
that two different people reading it would produce the same output?
If not, what additional specificity does it need?
What you're learning: The discipline of writing SKILL.md instructions that are specific enough to be testable. Generic instructions ("calculate the tax provision") produce generic outputs. Organisation-specific instructions ("apply the 25% statutory rate to UK entity profits, add back the GBP 340K disallowable entertainment expense as a permanent difference, and assess the GBP 2.1M deferred tax asset against three years of projected taxable income using the Board-approved budget") produce outputs that match how your organisation actually works. The critique step teaches you to evaluate your own instructions for sufficient specificity.
Prompt 3: Extension Dependency Mapping
I am planning to build multiple enterprise extensions for my
organisation. My priority list is:
1. [EXTENSION A — e.g., Treasury and Cash Management]
2. [EXTENSION B — e.g., Tax Provision and Compliance]
3. [EXTENSION C — e.g., FP&A and Budget Ownership]
Before I start building, help me understand the dependencies:
1. Do any of these extensions share data inputs? If Extension A
needs the same data that Extension B produces, which must be
built first?
2. Are there SKILL.md instructions that should be shared across
extensions rather than duplicated? Identify any cross-cutting
rules (e.g., entity structure, chart of accounts mapping,
approval hierarchies).
3. Design a build sequence: which extension should I build first,
second, and third based on dependencies and quick-win value?
4. What is the minimum viable version of each extension — the
smallest set of SKILL.md instructions that produces useful
output — so I can deploy incrementally rather than waiting
until all three are complete?
What you're learning: Enterprise extensions do not exist in isolation. Treasury needs the entity structure that tax provision also uses. FP&A needs the chart of accounts mapping that treasury references. By mapping dependencies before building, you avoid duplicating instructions and ensure that shared data definitions are consistent across extensions. The minimum viable version discipline prevents the common mistake of attempting a complete extension before validating that the approach works.
Flashcards Study Aid
Continue to Lesson 11: Your Extension Roadmap →