INTU Analysis: INTU — SaaS Survival V-Score: 3.34 (κ=0.34)

V-Score Card

TICKER           INTU
V-SCORE          3.34
VERDICT          EMBEDDED
kappa            0.34
E[kappa]         0.16  (post E/C adversarial review)

GATES
  G1 = 1[E > 1] = 1[3 > 1] = 1              PASS
  G2 = 1[A > 1 v (C+E+U) >= 12]
     = 1[4 > 1 v (4+3+4) >= 12]
     = 1[TRUE v FALSE] = 1                   PASS

DIMENSIONS
  C  (0.25)   4    GenOS (KE+ML+financial LLMs), 15 inter-module deps,
                   7 agents, industry verticals. Single domain caps at 4.
  E  (0.22)   3    Transaction rails (payments, capital, payroll),
                   IRS e-file auth, 50-state compliance. ZERO NRR disclosure.
                   Xero migration = data portable. Revenue ≈55% pricing.
  U  (0.18)   4    15 workflows across 6 departments. Cross-module data flows:
                   QBO -> Capital underwriting, CK -> TurboTax (+1pt growth).
  A  (0.12)   4    MCP endpoints built. OpenAI App Directory (all 4 apps).
                   Anthropic multiyear deal. 3M+ agent users, 85%+ repeat.
  M  (0.15)   4    ≈100M customers. Accountant ProAdvisor counterparty network.
                   Lightbox: FIs place proprietary credit models exclusively.
  F  (-0.06)  2    Consumer-grade UX, 80%+ AI data entry, self-serve.
                   QBO 5-tier complexity, accounting domain knowledge required.

ARITHMETIC
  V = 0.25(4) + 0.22(3) + 0.18(4) + 0.12(4) + 0.15(4) - 0.06(2)
    = 1.00 + 0.66 + 0.72 + 0.48 + 0.60 - 0.12
    = 3.34

FAST SCREEN (Bustamante)
  Proprietary data?       YES   7M+ SMB financial records, 130M+ credit profiles
  Regulatory lock-in?     PARTIAL   IRS e-file auth, but no routing mandate
  Transaction embedding?  PARTIAL   Payments/capital/payroll = rails; rest = information
  b(INTU) = 2/3

Regime Context (T = 15 Weeks)

BENCHMARKS (Dec 13, 2025 - Mar 28, 2026)
  SPY     -6.3%
  IGV    -27.0%
  INTU   -36.3%    (-9.3% vs IGV, -30.0% vs SPY)
  RSI     29.2     (deeply oversold)
  FWD P/E 27.09    (was ≈35 at peak)

FACTOR REGRESSION: r_INTU = alpha + beta*SPY + gamma*IGV_orth + epsilon
  alpha (ann)    -130.2%
  beta (SPY)       0.849
  gamma (IGV_orth) 1.358
  sigma_idio (ann) 31.8%
  R-squared        0.584
  %Idio Var       41.6%   (factor-dominated window)

IR_INTU = alpha_ann / sigma_idio = -4.098   (t-stat = -2.11)

INTRA-SECTOR CORRELATION
  rho_intra (ex-market residuals) = 0.487
  INTU avg rho with peers (resid) = 0.589

RETURN DECOMPOSITION
  Market component  (beta * r_SPY):       -5.4%
  Sector component  (gamma * r_IGV_orth): -20.4%
  Idiosyncratic     (alpha + epsilon):    -10.5%

IR = -4.098 looks catastrophic. It is not. Two diagnostics explain why IR does not gate the verdict:

Diagnostic 1: rho_intra = 0.487. Elevated but not unity. When rho approaches 1, epsilon collapses to zero for all names and IR measures the regime, not the stock. At 0.487, the idiosyncratic channel is partially open. IR is partially informative but regime-contaminated.

Diagnostic 2: Cross-sectional IR dispersion. The market IS discriminating within software -- this is not an indiscriminate selloff:

Ticker    IR_i     Tot Ret%    Cluster
-------   ------   --------    -------
WDAY      -5.537   -42.2%      Pure SaaS workflow
NOW       -5.241   -35.0%      Pure SaaS workflow
ADBE      -4.483   -33.1%      Pure SaaS workflow
CRM       -4.226   -29.5%      Pure SaaS workflow
INTU      -4.098   -36.3%      Pure SaaS workflow
FICO      -3.528   -44.1%      Regulatory moat
PANW      -2.340   -20.9%      Security
CRWD      -1.931   -24.2%      Security
SNPS      -0.385   -16.3%      Hardware-adjacent (EDA)
CDNS      -0.306   -14.7%      Hardware-adjacent (EDA)

IR cross-sectional sigma = 1.786 (high)

The market is repricing AI displacement risk along a gradient: EDA tools (IR near 0) are untouched. Pure SaaS workflow (IR -4 to -5) is getting crushed. Security sits between. INTU's IR = -4.1 places it squarely in the "workflow software the market believes agents will replace" cluster.

The V-Score says the market is wrong about INTU -- transaction rails, regulatory barriers, and compound cognition place it in EMBEDDED, not AT_RISK. But the market is not entirely wrong about the cluster -- workflow SaaS IS more exposed than EDA or security. The question is degree, not direction.

Dimension Analysis

C = 4 -- Compound Cognition (Challenged)

40+ years of tax/accounting knowledge encoded in GenOS: knowledge engineering + ML + proprietary financial LLMs. Not generic frontier models. 15 inter-module dependencies create superlinear re-derivation cost (QBO Accounting to Payroll to Tax to Payments to Capital). 7 shipped agents with measured compound value: accounting agent categorized 237M transactions in January (≈50% of all categorized). Industry verticals launching (construction edition with measured 90% reduction in month-end reconciliation).

Adversarial challenge: How much cognition is proprietary vs codified public knowledge? Federal tax rules, state tax codes, GAAP standards -- all public. A frontier model with access to IRS code + 100M categorized transactions could re-derive single-domain cognition within 12-18 months. The defense is the compound integration: tax x accounting x payroll x lending x compliance across 50 states x multiple industry verticals. No frontier model team has attempted end-to-end replication.

Honest assessment: C = 3.5 rounded to 4. The compound cross-module effect saves it. If frontier models crack compound integration within 2 years (assigned 40% probability), C drops to 3.

Sources: 10-K FY2025 lines 337-380, 369, 390; Q2 FY2026 transcript lines 36-42, 81, 114; Q1 FY2026 transcript lines 36-38.

E = 3 -- Irreducible Infrastructure (Challenged)

Transaction rails are real: QBO Payments volume +29% in Q2, QBO Capital originated $2.6B in H1 FY2026, payroll processes disbursements across all jurisdictions. CFO names it explicitly: "Our moat comes from being the core of the flow of funds." IRS e-file authorization blocks AI tax filing (92% probability no AI product gets IRS auth by April 2027). IRS Direct File program shut down for 2026 season with ≈1% adoption.

Adversarial challenge: The claim that switching cost equals infinity for ≈35-40% of GBS overstates the case. Decomposing actual switching cost for a typical SMB: accounting data migration 1-2 weeks (Xero tool exists), payment processing 1-2 weeks (Stripe/Square), payroll 4-8 weeks (Gusto/ADP handle compliance routinely), QBO Capital N/A (loans continue, alternative lenders exist). Total: $5,000-15,000, or 1-3 years of subscription value. Real but finite.

Three behavioral signals confirm switching cost is high, not infinite:

ZERO NRR/GRR disclosure. Every SaaS company with NRR above 120% discloses it -- it is the single most investor-friendly metric. Intuit reports nothing. Bayesian update: P(NRR > 120% given non-disclosure) is approximately 10-15%.
Xero built a free QBO migration tool. Engineering investment in migration tooling reflects real customer demand for switching. Revealed preference evidence.
QBO customer growth decelerating (+6% FY2024, +5% FY2025, below expectations Q1 FY2026) while ARPC rises +14%. This is extraction approaching the switching cost ceiling, not infrastructure lock-in.

Revenue-weighted switching cost reality: ≈20-24% of total revenue sits behind truly high switching costs (transaction rails). ≈30-35% moderate (QBO accounting). ≈45-50% low-to-zero (TurboTax annual decision, Credit Karma free service, Mailchimp commodity).

Honest assessment: E = 3 holds at ceiling. The accountant ProAdvisor network (counterparty gravity) is the underappreciated moat -- harder to quantify than transaction rails but arguably more durable. E = 2 case (30% probability) requires NRR below 105% and Xero gaining more than 5% QBO share.

Sources: 10-Q Q2 FY2026 lines 1461-1464, 2672-2677; Q2 FY2026 transcript lines 167-174; 10-K FY2025 lines 319-324, 3472-3478.

U = 4 -- Ecosystem Breadth

15 enumerated workflows across 6 departments: bookkeeping, invoicing, bill management, merchant payments, lending, payroll, time tracking, project management, multi-entity accounting, personal tax, professional tax, credit monitoring, email marketing, checking/savings, inventory management. Cross-module data flows create superlinear switching cost: QBO transaction data feeds Capital underwriting, Credit Karma drove 1 point of TurboTax growth, Intuit Intelligence queries across all data.

Not 5: no manufacturing, supply chain, procurement, or facilities. That is SAP territory.

Sources: 10-K FY2025 lines 319-324, 396-530; 10-Q lines 2015-2022; Q1 transcript line 146.

A = 4 -- Distribution and Discoverability

MCP endpoints explicitly built ("APIs and MCP that spans a customer's financial life"). All 4 Intuit apps in OpenAI App Directory. Multiyear Anthropic partnership: Intuit platform becomes foundation for Claude Agent Builder. No revenue share on either partnership -- Intuit owns the experience and economics. 3M+ customers using AI agents, 85%+ repeat engagement, accounting agents handling ≈50% of all categorized transactions.

Not 5: agents do not route through Intuit by default for all financial tasks. Enterprise goes to SAP/Oracle. But upgrade candidate if OpenAI/Anthropic routing scales.

Sources: Q2 FY2026 transcript lines 81-85; Q1 transcript line 92; 10-K line 520-521.

M = 4 -- Ecosystem Gravity

≈100M customers worldwide. Accountant ProAdvisor network ("hundreds of thousands") creates genuine counterparty network effect -- nearly 1/3 of new IES contracts were accountant-influenced in Q2, 10 points higher than Q1. Lightbox: financial institutions place proprietary credit models exclusively on Credit Karma platform. IES mid-market contracts growing ≈50% QoQ. Multi-year tax return data creates temporal gravity.

Not 5: Xero migration exists, NRR undisclosed, not ADP-level industry standard.

Sources: Q2 FY2026 transcript lines 29, 36, 55-56; Q1 transcript lines 45-48, 149-150; 10-K line 504.

F = 2 -- Ecosystem Friction (Penalty)

Consumer-grade UX across core products. TurboTax guided interview with 80%+ AI data entry. QBO self-serve signup. Credit Karma free and fully self-serve. 600 local service centers. Natural language Intuit Intelligence reducing interaction cost.

Not 1: QBO has 5 tiers, accounting requires domain knowledge, mid-market onboarding requires configuration.

Sources: Q2 transcript lines 41, 62, 70-72; 10-K line 523-527.

Sensitivity and Adversarial Scenarios

SCENARIO                  E    C    V      TIER        kappa   P(scenario)
--------------------------------------------------------------------
Base (as scored)          3    4    3.34   EMBEDDED    0.34    30%
Cognition re-derived      3    3    3.09   EMBEDDED    0.09    40%
Switching overstated      2    4    3.12   EMBEDDED    0.12    15%
Double bear (E+C)         2    3    2.87   AT_RISK     0.00    15%
NRR >120% (bull)          4    4    3.56   EMBEDDED    0.56    --
Agent default (bull)      3    4    3.46   EMBEDDED    0.46    --

E[V]     = 0.30(3.34) + 0.40(3.09) + 0.15(3.12) + 0.15(2.87) = 3.13
E[kappa] = 0.30(0.34) + 0.40(0.09) + 0.15(0.12) + 0.15(0.00) = 0.16

The C dimension is the closer call. Single-domain cognition (tax alone, accounting alone) is C = 3 -- the knowledge is primarily codified public law applied through good software. The compound cross-module integration is the only thing holding C = 4. If frontier models crack that integration within 2 years (40% probability), C falls.

The E dimension holds at 3 but the infinite switching cost claim for transaction rails is overstated. Real switching cost is $5,000-15,000 for a typical SMB -- high enough to sustain 10-15% annual price increases, finite enough that a materially better product could trigger migration. The NRR black hole is the single largest source of scoring uncertainty. Resolve NRR, resolve E, resolve the score.

Thermodynamic Summary

Intelligence cannot flow around Intuit's transaction rails -- payments, payroll disbursements, and lending require regulated banking infrastructure that no local agent can replicate. That is the irreducible core, representing roughly 20-24% of total revenue with genuinely high switching costs.

Everything else is information processing on a kill cycle:

TASK                    c_local    t* (years)    % of Revenue
---------------------------------------------------------------
Payment processing      infinite   never          ≈8-10%
Payroll disbursement    infinite   never          ≈6-8%
Lending (QBO Capital)   infinite   never          ≈4-6%
Complex tax prep        high       5-8            ≈15%
Simple tax prep         moderate   3-5            ≈11%
Complex bookkeeping     moderate   5-8            ≈15%
Simple bookkeeping      low        2-4            ≈15%
Email marketing         low        1-2            ≈6%
Credit monitoring       low        2-3            ≈6%
Lead-gen/referral       low        1-2            ≈6%

Resistance trajectory:
  R(INTU, 2026) ~ 0.85
  R(INTU, 2030) ~ 0.65
  R(INTU, 2035) ~ 0.50

R(INTU, infinity) > 0 because transaction rails have infinite local cost. Intuit survives because it moves money, not just tracks it. The information layer compresses; the infrastructure layer persists.

The accountant ProAdvisor network is the underappreciated structural asset. It creates counterparty gravity that reinforces every other dimension: accountants recommend QBO, clients adopt QBO, accountants deepen QBO expertise, more accountants join. This network effect operates independently of AI capabilities and may actually strengthen as AI makes the platform more valuable to accountants (12-14 hours saved per month).

delta: The Structural Discount Gap

V_scored           = 3.34  (EMBEDDED)
V_market_implied   ~ 2.8   (inferred from FWD P/E 27x vs peer 30-40x for growth profile)
delta              ~ 0.54  (base)

E[V]               = 3.13
delta_adjusted     ~ 0.33  (with E/C adversarial haircut)

The market is applying an AI displacement discount that compresses INTU from EMBEDDED pricing (≈35x for +24% QBO growth) to near AT_RISK pricing (≈27x). The V-Score says the structural properties do not justify AT_RISK classification -- transaction rails, regulatory barriers, and counterparty networks are not impaired by the software selloff.

However, delta is not pure mispricing. Part of the discount reflects legitimate uncertainty about E and C that the V-Score sensitivity analysis quantifies: 15% probability of AT_RISK (V = 2.87). The market may be overweighting this tail but not inventing it.

Insider behavior corroborates the structural view: founder and executive leadership terminated ALL pre-scheduled 10b5-1 stock sale plans in March 2026. Combined with $3.5B remaining buyback authorization and H1 FY2026 repurchases of $1.8B (+40% YoY). Insiders choosing not to sell at these levels is Tier 1 behavioral evidence.

Conviction Weight

kappa_base     = (V - 3.0)+ = (3.34 - 3.0)+ = 0.34
kappa_expected = E[kappa]   = 0.16

w_INTU = W_S * kappa_INTU / sum_j(kappa_j)

Use kappa = 0.16 for basket weighting. The adversarial review on E and C halves the conviction weight. This is not bearish on INTU -- it is honest about where the scoring confidence interval is wide. NRR disclosure resolves the ambiguity; until then, half-weight is the correct position.

Watchpoints:

QBO paying customer growth in FY2026 10-K (July 2026): below 4% while ARPC rises confirms extraction, not stickiness
Any NRR/GRR disclosure: immediately resolves E dimension uncertainty
Frontier model financial product launch (OpenAI/Google/Anthropic partnering with payroll/payments provider): would accelerate C re-derivation timeline
Mailchimp strategic review outcome: divestiture removes commodity drag, improves portfolio purity

Basket Verdict

KEEP. V = 3.34, EMBEDDED. kappa = 0.34 (base), E[kappa] = 0.16 (adjusted).

INTU belongs in a SaaS survival basket as a mid-weight position. The structural thesis is sound: transaction rails persist, counterparty network reinforces, AI makes the platform more valuable. The exposed flanks (TurboTax computable task risk, Mailchimp commodity drag, NRR black hole) are real but bounded -- even the double-bear scenario (V = 2.87) barely crosses into AT_RISK.

IR = -4.098 measures the regime, not the stock. Software is repricing AI displacement risk along a gradient. INTU sits in the "workflow SaaS" cluster getting the harshest treatment. V-Score says the market is overcorrecting for INTU specifically because it conflates information processing (exposed) with financial infrastructure (durable). The delta exists, but size to E[kappa], not kappa_base.

At 34 analysts and $116B market cap, there is zero informational edge. Any position here is structural conviction (V-Score says EMBEDDED while market prices AT_RISK), not idiosyncratic alpha. That is the trade: you are betting the V-Score framework is more correct than the market's current AI displacement discount. The framework has primary source backing. The market has momentum.