VS. NOTHINGBY HERITAGE HOLDINGS

HOUSE UNDERWRITING · WORKING PAPER · SEPTEMBER 16, 2026

SHADOW MODE · NOT A LIVE CALL

The interface is disappearing.
The record is not.

CURRENT HOUSE ACTIONADVANCE CRM

At the September 15 close, Salesforce clears the proposed three-year return hurdle and bear-downside rail in the first explicit model. Build the complete Call packet; do not promote this working paper into a position.

Ben Thompson argues that agents will make enterprise software interfaces less defensible. HOUSE agrees with the mechanism but not with the shortcut. The investable question is which system of record still controls governed data, permissions, and action when the agent owns the screen.

THESIS ORIGINBEN THOMPSON · STRATECHERY · SEPTEMBER 15, 2026READ THE ORIGINAL
PRODUCT THESIS · CREDIBLEHOUSE POSITION · NONEFIRST SCREEN · CRMBENCHMARK · IGV
ATTRIBUTION BOUNDARY

Thompson supplies the product observation: “Salesforce should be headless.” He does not recommend CRM, specify an entry, set a horizon, or define a loss condition. Every security choice and number below is HOUSE inference.

01

INVESTMENT CONCLUSION FIRST

CRM clears. The basket does not.

OBSERVATIONPASS

Agents can bypass interface friction. UI familiarity alone is becoming a weaker moat.

MECHANISMCONDITIONAL PASS

Systems of record may retain power through governed data, permissions, identity, and write access.

VALUATIONCRM · FIRST PASS

The modeled base case clears a proposed 13% annualized hurdle without requiring a heroic exit multiple.

ALLOCATION$0 · SHADOW

No live HOUSE position, sealed basis, or AI panel exists. Advance one name, not a correlated software basket.

THE CLAIM HOUSE MAY EVENTUALLY LOCK

From a future eligible close, CRM should outperform IGV held unchanged over 36 months if Salesforce converts its system-of-record position into governed agent actions faster than agent interfaces erode seat pricing.

THE THESIS IS NOTAI is good for software.
THE THESIS ISValue migrates from the screen to the governed record and action layer.
02

HOW THE THESIS MAKES MONEY

Lose the face. Keep the authority.

01 · INTERFACEAgents own the screen

Users ask a model to complete work instead of navigating each vendor's interface.

02 · RECORDEnterprise truth persists

Customer, workflow, entitlement, and transaction data still require an authoritative home.

03 · CONTROLPermission becomes product

The valuable layer authenticates the agent, limits its actions, records its work, and reverses mistakes.

04 · ECONOMICSUsage replaces seats

Winners monetize governed reads, writes, workflows, and data movement faster than UI revenue decays.

ALL FOUR LINKS MUST HOLD

A correct product observation can still produce the wrong stock.

  1. Agents must act, not merely summarize. Copilots that stop at answers do not create a new transaction layer.
  2. Customers must keep the record vendor. Open formats, warehouses, and agent memory cannot make the application database interchangeable.
  3. Control must remain monetizable. API and action revenue must replace lost seat value instead of becoming a free compatibility feature.
  4. Growth must reach each share. Acquisition spending, stock compensation, infrastructure cost, and buybacks decide the equity result.
03

MAP THE SECURITIES

Four names. Three different trades.

CRM, NOW, and SNOW share enterprise-AI exposure; owning all three does not diversify this thesis. ORCL's present return mechanism is even further away. HOUSE ranks the claims instead of hiding them in a basket.

CRM · PRIMARY CANDIDATESYSTEM OF RECORD

Sales, service, customer data, permissions, and workflow can remain authoritative even when another agent owns the interface.

WHAT MUST PROVE · AGENTFORCE + DATA 360 GROW PER-SHARE CASH FASTER THAN SEAT VALUE DECAYS
NOW · SECOND SCREENSYSTEM OF ACTION

Workflow authority and Action Fabric may make ServiceNow a cleaner cross-agent execution layer.

WHAT MUST PROVE · THE SUPERIOR GROWTH SURVIVES A FULL VALUATION AND DILUTION MODEL
SNOW · WAITNEUTRAL DATA LAYER

Governed access across application silos is strategically useful, but adjusted cash flow cannot erase dilution.

WHAT MUST PROVE · PER-SHARE ECONOMICS CATCH THE PRODUCT-REVENUE STORY
ORCL · SEPARATE BOOKDATABASE + COMPUTE

Database gravity is relevant, but the current earnings debate is AI infrastructure, backlog, and capital intensity.

GRADE IT AS AI INFRASTRUCTURE, NOT AS HEADLESS-SAAS DIVERSIFICATION
04

PRICE IS PART OF THE THESIS

CRM earns the next round.

The model uses reported FY26 free cash flow, the midpoint of FY27 cash-flow growth guidance, guided diluted shares, explicit per-share growth, and an exit free-cash-flow multiple. It is intentionally simple enough to attack.

Loading the HOUSE model…

PRICE BASIS. CRM $255.65, NOW $141.90, SNOW $322.98, ORCL $140.35, and IGV $105.55: September 15, 2026 daily closes from Yahoo Finance via the VPS, queried September 16. NOT LIVE.

CRM INPUTS. FY26 free cash flow · FY27 guidance and diluted shares · Q2 Form 10-Q.

BENCHMARK. IGV tracks North American software and selected interactive-media companies. The 10% base assumption and +3-point hurdle are HOUSE proposals, not iShares forecasts.

05

THE OPPOSITION CASE

Headless can be right and CRM can still lose.

ATTACK 01 · COMMODITIZATION

The model owns the customer.

Third-party agents reduce Salesforce to a replaceable database and pressure API access toward zero-margin compatibility.

ATTACK 02 · CANNIBALIZATION

Usage does not replace seats.

Premium seats and UI bundles contract before Agentforce consumption becomes large or durable enough to compensate.

ATTACK 03 · PORTABILITY

The record moves.

Warehouses, open tables, MCP servers, and model memory make enterprise context less dependent on the application vendor.

ATTACK 04 · PER SHARE

The company wins. Owners do not.

Acquisitions, stock compensation, infrastructure cost, and buybacks consume the operating benefit while IGV compounds elsewhere.

06

REALITY GRADES

Usage is real. Capture is unresolved.

These facts support rapid agent adoption, governed-data demand, and enterprise willingness to spend. They do not yet prove that record owners capture more economics than model vendors, warehouses, integrators, or customers.

07

RTI · REAL-TIME INTELLIGENCE

Watch the value migrate.

NOW · DREAMFORCEInvestor and Analyst Session · Sep 16

Listen for headless-platform pricing, Agentforce consumption, Data 360 attach, and explicit evidence that usage replaces rather than cannibalizes seats.

NEXT REPORT · DATE PENDINGSalesforce Q3 FY27

Track organic cRPO and subscription growth excluding Informatica, Agentforce ARR, cash conversion, stock compensation, and diluted shares.

COMPETITOR TAPENOW · SNOW · ORCL

Which layer records agent actions, controls permissions, and charges for them? Product announcements matter only when revenue and per-share cash follow.

PRICING TAPESEATS VS ACTIONS

New free tiers, usage units, API charges, bundled access, and customer migration reveal where bargaining power is moving.

RTI means real-time intelligence: recurring public evidence checked against the same written mechanism. It is not a claim that market data or the page updates tick by tick.

08

WHAT WOULD CHANGE OUR MIND

Make the thesis mortal.

PROPOSED KILL CONDITIONS · NOT SEALED

Dates, series, and exact thresholds must freeze with the future Call packet. These are underwriting proposals, not a graded record.

  1. ORGANIC GROWTH FAILS. By the FY28 year-end report, Salesforce's organic subscription and support growth has not reaccelerated despite material Agentforce and Data 360 usage.
  2. CASH CAPTURE FAILS. Three-year free cash flow per share grows below 7% annualized from the future lock basis.
  3. HEADLESS PRICING FAILS. Governed agent reads, writes, or actions remain bundled or free and do not replace measured seat pressure by FY28.
  4. THE RECORD MOVES. Disclosed customer migrations show warehouses, open data layers, or model platforms replacing Salesforce as the authoritative customer record.
  5. PER-SHARE DISCIPLINE FAILS. Acquisitions and stock compensation offset operating cash growth, leaving diluted per-share economics below the sealed path.
  6. THE TRADE LOSES. At 36 months, CRM fails to outperform IGV held unchanged. The money verdict stands even if headless architecture wins.
09

WHAT REMAINS BEFORE CAPITAL

Turn the screen into a record.

01 · CRM MODELRebuild organic revenue and per-share cash.

Separate Informatica, model seat pressure, action consumption, infrastructure cost, stock compensation, buybacks, and acquisition spending.

02 · NOW COMPARISONUnderwrite the strongest alternative.

Normalize free cash flow, dilution, and valuation instead of awarding the trade to the fastest current growth rate.

03 · METHODFreeze IGV, 36 months, and the evidence series.

Same close, no rebalance, corporate actions, benchmark treatment, and exact kill-condition settlement.

04 · AI PANELAttack one complete packet.

Do not ask frontier models to vote on an article. Ask them to challenge the security, entry, cases, and falsifiers.

05 · LOCKUse the next eligible close.

Never backfill September 15 as a basis. If price or facts change first, rerun the model.

06 · PUBLISHOnly then does a Fight begin.

The source observation, HOUSE inference, machine attack, and future evidence remain separate on the permanent record.