From System of Record to System of Context & Work

Part 1: The Argument

1.1 The Inversion

For forty years, enterprise software was built on one architectural assumption: the human is the interface. Schemas, forms, role-based dashboards, per-seat pricing, configured rules. All of it derives from the assumption that knowledge work is humans typing structured intent into rows so other humans can query those rows later. The system of record (SoR) was, fundamentally, a database with a UI thick enough to keep the data clean.

Agent-as-interface breaks this assumption. When a model sits between the work and the database, almost every architectural primitive of the previous era becomes wrongly shaped. Schemas grow too rigid for work that no longer arrives as form fills. Forms themselves drop out of the user surface, because the model handles intent directly. Reports collapse into on-demand reasoning rather than persisted pages, customization moves from rules-as-code to markdown skills, and pricing settles on per-outcome instead of per-seat. The deeper shift sits underneath all of these: the job of the SoR itself changes. It stops being the place where truth is stored and starts being the substrate that feeds a learning loop.

This is the actual architectural inversion. The surface-level reframings (“AI-native CRM,” “AI-native ERP,” “agent-first support”) describe the marketing. The deeper change is that the entire data shape pivots from “validated state at point-of-entry” to “trajectory plus outcome flowing through a loop.” Companies that get this right become systems of context and work. The rest ship agents on top of the same legacy substrate.

1.2 The Flywheel Architecture

The learning loop is the actual structural unlock. Five primitives must interlock for it to spin. Take any one out and you have a chatbot rather than a learning system.

(1) Outcome-Grounded Signal. Every agent run terminates in a verifiable outcome that arrives back from the world without human annotation. A trial balance either reconciles or it doesn’t; a support ticket either gets resolved without escalation or it doesn’t; a claim either pays or stalls in audit. The signal is binary or graded but mechanical, with the cognitive trajectory captured along the way rather than judged by a human after the fact. This is what the traditional SoR fundamentally lacked. SoRs persisted states (the row got created) rather than outcomes (the work that row represented succeeded against an external criterion). Domains where (1) is dense (code, accounting, customer service, leasing) adopt agents fastest. Domains where (1) is sparse or slow (strategy, creative work, multi-year legal matters) adopt slowest.

(2) Trajectory Persistence. Every step of every run preserved as data: what was retrieved, which tool was called, which model gave which intermediate output, what was rejected, what alternative was considered. The trajectory IS the training corpus. Schema-first SoRs persisted final committed state only; they had no representation of how that state was produced because the producer was a human and the cognitive trajectory lived in the human’s head. (2) only matters because of (1): without an outcome to attribute the trajectory to, the trajectory is exhaust.

(3) Attribution + Eval Architecture. Structured machinery for answering “what caused this outcome.” Regression evals on every change. Replay against new prompts and models. Isolated subagent traces. Sierra’s stress-tests with trick questions before live launch; Decagon’s Trace View as an in-product debugger; Numeric’s variance-explanation gates before any close moves to publish. Without (3) you have data but no gradient. You cannot tell which parts of the trajectory caused success or failure, so you cannot update the substrate intelligently.

(4) Editable Policy Substrate. The writable surface that the loop updates. Markdown skills, prompt templates, eval suites, routing tables, memory graphs, fine-tunes. Multi-versioned, expert-authorable, model-agnostic. SoR customization was rules-as-code (Salesforce custom objects, NetSuite SuiteScript, Workday Studio). Every customer became a divergent fork and learning never transferred back. Substrate-based customization is convergent: the loop writes back into a shared structure that all customers benefit from. The substrate is also where domain experts now author logic directly, collapsing the talent-scarcity moat.

(5) Cross-Customer Aggregation Surface. The substrate is multi-tenant on the learning layer even when data is isolated on the facts layer. Campfire’s LAM trains across customers; customer 100 inherits the categorization patterns and reconciliation heuristics that customers 1 through 99 produced. Sierra’s routing optimization learns across deployments. Clay’s recipe library propagates. Without (5), each customer is a fresh start, and (4) collapses back to fork-style customization (i.e. SoR economics).

The mechanic. An agent run executes; it produces an action in the world. The world returns an outcome (1). The full trajectory is captured (2). Attribution (3) maps the outcome backward through the trajectory to identify what caused success or failure. The substrate (4) updates with the lesson. Aggregated across customers (5), the lesson flows to all future runs. The next outcome arrives more likely to succeed.

The speed of one cycle of this loop is the durability metric. Sierra’s constellation routes each new conversation through whichever model is currently winning on cost-quality, with the routing decision itself updated continuously from outcome data. Campfire’s LAM ingests every customer transaction as it lands and feeds the categorization patterns back into the shared substrate. Cycle time is the moat. Companies whose architecture lets them turn the crank faster compound faster.

1.3 Why Incumbents Cannot Retrofit

Three structural reasons.

First, architectural absence of (1) through (3). Incumbents have at most partial outcome signal capture; near-zero trajectory persistence; near-zero attribution/eval architecture. They can ship agents on top of legacy schemas (Salesforce Agentforce, NetSuite AI Close, Workday Illuminate, Service Cloud Einstein) but cannot rebuild the substrate without a multi-year data architecture rewrite that no incumbent has yet committed to publicly.

Second, commercial misalignment with (4) and (5). Per-seat pricing penalizes substrate compounding. Every successful agent deflection is a cancelled seat license. Salesforce cannot ship a substrate-aggregating agent that genuinely compresses headcount because doing so breaks its own revenue model. ServiceNow’s $2.85B acquisition of Moveworks in December 2025 was the way out: pay an external party to build the harness, then bolt it on without disturbing the seat economics in the short term.

Third, organizational dependence on the customization economy. Salesforce has $30B+ per year of SI partner revenue (Accenture, Deloitte, Capgemini, Slalom) that is structurally threatened by markdown-as-substrate. The same applies to NetSuite (BDO, Armanino, RSM) and Workday (Mercer, Deloitte, KPMG). The political coalition inside the incumbent that resists the substrate move is bigger than the engineering team that would have to build it.

Combined effect: incumbents move faster than expected in shipping agent surfaces (Agentforce, Joule, Illuminate, Now Assist) and slower than required in rebuilding the substrate. The gap widens between 2026 and 2028.

1.4 The Org-Level Consequence

The line worker disappears. The “Salesforce admin,” the “NetSuite implementer,” the “Zendesk agent,” the “billing analyst,” the “claims processor”: roles that existed because someone had to sit inside the SoR moving rows around. They collapse into two new shapes.

• The agent operator configures, monitors, and improves the agent: writes and edits skills, designs evals, tunes routing, manages model upgrades. Sierra’s Workspaces productize this role explicitly with parallel-development branches, reviewable changes, controlled releases. The Campfire customer running accounting at $250M ARR with a single controller is operating the agent rather than doing the books.
• The exception handler solves cases the agent escalates. Each handled exception becomes training data; the gate widens; the role shrinks in volume but remains for the highest-judgment edge cases.

Tribal knowledge (“what to do when X happens” that used to live in a senior person’s head, a Slack thread, a half-written runbook) is forced into one of three structured forms: a markdown skill, an eval, or a curated trajectory. Knowledge accrues to the company, not the individual. This holds only at companies whose architecture supports the artifact form.

The buying center shifts from CIO IT spend to COO/CFO labor or operations budgets. The labor budget is 35× the size of the software budget ($11T vs $315B per NEA / FRED data); the buyer has a faster decision cycle and a higher tolerance for outcome-based contracts. This is why AI-native P&Ls can sustain at 50 to 70 percent gross margins (Redpoint 2026). They’re priced into a much larger pool against a much higher willingness-to-pay.

Part 2: The Field

Five SoR segments matter for this thesis: ERP / Financial Systems, CRM / Sales, Customer Service, ITSM, and Regulated / Operational verticals. The same flywheel logic plays out in each one, with different velocities and different equilibrium structures. The main text walks through finance as the exemplar, because finance has the densest signal surface and the clearest replacement story so far. The other four segments live in Appendix A with the same depth treatment.

Incumbents. SAP, Oracle, Microsoft Dynamics 365 F&O at the enterprise level; NetSuite, Sage Intacct, Workday Financials at mid-market; QuickBooks and Xero at SMB. Adjacent stacks: BlackLine and FloQast (close), Bill.com / Tipalti / AvidXchange (AP/AR), Anaplan / Pigment (FP&A), Coupa / Ariba (procurement). Per Redpoint’s 2026 CIO survey, 50% of CIOs are open to replacing their ERP with an AI-centric vendor, tied with procurement, second only to CRM (83%) and CX (56%).

Why finance is the densest signal surface. Three properties make ERP the most architecturally suitable SoR for the flywheel: outputs are numerically verifiable against GAAP (the trial balance reconciles or it doesn’t); the workflows (categorize → match → reconcile → close) repeat millions of times per customer per year, generating dense trajectory data; outcome attribution is mechanical, every journal entry is right or wrong against external bank/vendor truth. (1), (2), (3) are dense by default. (4) and (5) are the engineering problem.

Campfire, system of action, not record

John Glasgow founded Campfire in 2023 to “upend 1990s-era ERP like NetSuite” with an LLM-powered alternative. $100M raised in 12 weeks across Series A and B (Foundation + Ribbit + Accel), $375M post-money. 10× revenue increase YTD. 100+ companies migrated from NetSuite/QuickBooks. CareRev calls it “Ramp for accounting.”

The architecture maps cleanly to the five primitives. (1) Outcome: every reconciliation either ties to bank/vendor truth or doesn’t; every journal entry passes the trial-balance check or doesn’t, dense per-transaction signal. (2) Trajectory: Campfire persists every reconciliation attempt, every category assignment, every allocation decision with the LLM’s reasoning, not just the final journal entry. This is what makes the LAM (Large Accounting Model) trainable. (3) Attribution: LAM is benchmarked at 95%+ accuracy on reconciliation and variance analysis. (4) Substrate: LAM weights themselves, plus Ember, the conversational AI assistant where finance teams query data and automate workflows in natural language. (5) Aggregation: LAM trains across all 100+ customers; a new customer’s books inherit categorization patterns and reconciliation heuristics learned from prior customers’ books.

What Campfire does that NetSuite couldn’t. NetSuite persists committed state. GL accounts, journals, customer/vendor masters. It has no schema for “the alternative allocation the agent considered and rejected because of a vendor pattern.” It has no LAM because it has no trajectory store to train one on. NetSuite’s customization is SuiteScript, fork-per-customer, unaggregable. Its pricing is per-seat plus implementation, which actively discourages reducing the number of accountants required.

Customer evidence (JPMorgan, January 2026). Flex (~80 employees, fintech): close cycle 10 days → 3 days, 60K transactions auto-mapped per month, 67% lower headcount needs. Fooji (experiential marketing): close 15+ days → 3 days, eliminated NetSuite consulting spend, finance “shifted from book-keeping to strategic business partnership.” CFO testimonial: “Switching to Campfire from NetSuite has been an absolute game-changer. As a CFO running a multi-currency, multi-legal entity operation, I’ve dealt with clunky, frustrating accounting software before. Campfire? A totally different story.”

What this segment teaches. Finance is the canonical proof-of-thesis segment. The five primitives are dense by default, and the engineering problem reduces to shipping (4) and (5) faster than incumbents can retrofit (1) through (3). Campfire is one of several GL Replacers building this layer in parallel (Rillet, DualEntry, Digits, Light, Numeric, Basis sit alongside it in Appendix A.5 with the segment depth). By 2027, one or two GL Replacers cross $100M ARR. By 2028, the question is whether NetSuite acquires one or rebuilds.

2.2 The Same Logic, Across Four More Segments

The flywheel argument generalizes from finance to the four other SoR categories. The velocities are different, the equilibrium structures are different, but the architectural inversion holds.

CRM / Sales has the highest CIO replacement openness (83 percent) and the weakest incumbent moats. Clay reached $100M ARR with a credit-priced signal-graph substrate that Salesforce cannot replicate without rebuilding its activity model. Customer Service / CX has the densest flywheel signal of any segment, with Sierra at $150M ARR after seven quarters and Decagon at $35M ARR powering Chime, Notion, and Rippling. ITSM splits between a horizontal employee front-door (Moveworks, now ServiceNow-owned) and a vertical access wedge (Serval at ~$50M ARR, 500 percent growth since Series A). Regulated and Operational verticals refuse SoR replacement (Epic, Workday, iManage are untouchable) and force AI-native value into adjacent layers: Abridge and Ambience as ambient scribes paying Epic’s Workshop tax, Mercor at $1B run-rate as labor bypass that sidesteps Workday entirely, EliseAI touching one in twelve US apartments as the operational layer that may yet eat Yardi’s GL.

Each segment has its exemplar deep dive, its company comparison set, and its one-paragraph teaching in Appendix A.

Part 3: The Implications

3.1 What Becomes Clear

The unit of value capture has moved from the record to the run. Software earned its rent for forty years by being the place where state was stored. The new rent comes from being the place where the work happens. That is the whole shift, said simply.

The toll holds in three positions and fails in a fourth. It holds above the user, where productivity orchestrators sit (Microsoft, Anthropic); below the user, where the model substrate sits (OpenAI, Bedrock, the Anthropic API); and inside the compliance hub, where regulated SoRs sit (Epic, Veeva, Workday). The horizontal SoR layer between them is exactly where the toll breaks down. Agentforce shipped four pricing models in eighteen months, SAP customers route around the platform to non-SAP solutions, and the consumption shock at production scale is real.

Whoever owns the workflow runtime owns the customer. Sierra won at a named financial-services bake-off with ninety-percent resolution rates while Salesforce sat on every byte of that customer’s data. The incumbent telemetry produces a governance moat: high-margin, durable, and capped at the size of regulation itself. The intelligence moat compounds elsewhere.

The replacement compounds in months. Sixty AI-native SoR-replacement companies crossed a hundred million in ARR by early 2026, with another fifty projected by year-end. Sierra reached a hundred million in seven quarters; Mercor reached a billion in run-rate seventeen months later. The constraint that matters in 2027 is no longer demand or quality. It is power: ground grids saturate by 2028, and the frontier labs are already negotiating space-based compute.

The ten-billion-dollar AI businesses live at the orchestration layer and at the workflow runtime in customer-facing categories. Microsoft and Anthropic above the user; Sierra, Intercom Fin, Decagon, Harvey, and Glean at the runtime. Vertical regulated software caps around two to four billion AI ARR per vendor. This is a real ceiling, and it tells you where the largest equity-value pools sit for the next five years.

3.2 The Four Meta-Archetypes, Confirmed

The 2026 evidence base validates the archetype framing. Each archetype now has a quantitative scoreboard from broker triangulation, with specific examples.

GL Replacer (Greenfield SoR Rebuild). Examples: Campfire, Rillet, DualEntry, Digits (ERP); Day.ai, Attio, Reevo, Rox (CRM substrate); EliseAI on its long-term arc. Primitives that dominate: (4) editable substrate and (5) cross-customer aggregation. The pitch is “rebuild the SoR from scratch with the substrate as the primary architectural commitment.” Moat profile: strong on three dimensions for finance and ERP (proprietary GL data, GAAP/SOX regulatory, transaction embedding); weak on all three for greenfield CRM rebuilds. The 2026 validation came fast. Campfire replaced NetSuite and QuickBooks at 100+ companies within nine months of formation. 21% of partners reported HubSpot displacement, 18% reported SAP displacement in favor of third-party AI agents or vibe-coded solutions (TBR, 24 Apr 2026). The lock-in moat that protected NetSuite for two decades has collapsed at mid-market scale. Head of Finance Ops at Sierra: “switching between modern ERPs is feasible with low risk.”

Workflow Wedge. Examples: Numeric (close), Basis (audit firms), Spellbook (legal drafting), Harvey (legal), Glean (knowledge work), Document Crunch (construction contracts), Codametrix (medical coding), Norm AI (regulatory compliance). Primitives that dominate: (1) outcome signal and (3) attribution / eval. Wedges live on the densest single workflow inside a vertical. Their architectural commitment is narrow but deep eval. Harvey nearly doubled from $100M to $195M ARR between August 2025 and January 2026, with DLA Piper picking Harvey over LexisNexis Protégé for “superior agentic automation.” Harvey delivers a 10 to 15 percent efficiency increase for senior lawyers on non-billable work. The Wedge survival path bends Wedge → Workflow OS → SoR; Numeric’s “compound startup” framing is the canonical play.

Concierge / Brand Layer. Examples: Sierra, Decagon, Parloa, Intercom Fin (CX); EliseAI (real estate operational layer); Hippocratic (nursing); Abridge (clinician-side scribe). Primitives that dominate: (2) trajectory persistence and (4) editable substrate, with (5) constrained by brand isolation. Moat profile: strongest on data, since trajectory volume and brand-customized substrate are both proprietary by construction. The best three-question profile of any archetype. Compounding rate: quadratic in (conversation density × customers). Sierra reached $100M ARR in 7 quarters (Feb 2024 to Nov 2025), doubled to ~$150M by Jan 2026 at a $10B valuation. Sierra’s installed base reaches 95% of American Black Friday shoppers, 50% of US families in healthcare, >90% of US media, >70% of US fintech. The strategic-move-up-the-stack pattern (operational layer becomes SoR) has gone from theory to live. WeightWatchers, Sonos, SiriusXM, and OluKai now have AI agents as their primary text-interaction channel as of October 2025.

Labor Bypass. Examples: Mercor (placement), Crescendo (CX BPO hybrid), Eve Legal (plaintiff PI vertical labor), Hippocratic (nursing tasks). Primitives that dominate: (1) outcome and (5) aggregation. Moat profile: outcome-ownership. Compounding rate: sublinear-to-linear with placement count; agency-style unit economics. Mercor hit $1B run-rate by early 2026, seventeen months after reaching $500M. Mercor’s $1B is driven by AI-lab demand for specialized training-data talent rather than SoR replacement. The category-archetype mapping holds. Mercor is best read as a primitive supplier to the AI economy. Caution: 11x demonstrated that AI-Employee role-replacement without measurable outcome accountability is a trust-collapse archetype masquerading as Labor Bypass.

3.3 The Compounding Dynamic, Quantified

Each archetype runs the flywheel at a different cycle time and accumulates a different aggregation surface. The product determines compounding rate. At least 60 AI-native companies reached $100M ARR by early 2026; 50 more are projected to hit $250M by year-end (Greenwich, 3 Apr). The replacement layer has graduated from thesis to observed history.

Incumbents cannot match this rate. Three reasons restated: architectural absence of (1) through (3); commercial misalignment with (4) and (5) under per-seat pricing; the customization-economy political coalition. The combined effect: incumbents move fast in shipping agent surfaces and slow in rebuilding the substrate. The gap widens between 2026 and 2028.

The unit-economic picture, corrected. AI-native gross margins land at 55 to 75 percent at maturity, against traditional SaaS at 80 to 90 percent. The intelligence tax of inference accounts for the spread. The historical cost drop is real (a ~155× drop in median inference price from early 2024 to early 2026, from $37.20 to $0.24 per million tokens). The 50× annual compound rate is decelerating. NVIDIA’s roadmap projects 2 to 10× improvement at Vera Rubin (H2 2026), 2 to 3× at Rubin Ultra / Kyber (H2 2027), 14× at Feynman (2028). Hardware bottlenecks (HBM4 yields, CoWoS packaging, power infrastructure) create a price floor for premium models.

Two countervailing forces make this complicated. The Inference Multiplier: moving from chatbot (~10K tokens) to agent (~1M+ tokens) workflow means 100× compute consumption per task. Per-token costs drop while total spend climbs (Jevons Paradox). ServiceNow’s “10% Rule”: AI reasoning sits at less than 10 percent of cost-to-serve for workflow orchestrators, so incumbents preserve 80 percent margins even with rising inference costs.

Net: AI-native margins settle at 60 to 70 percent rather than climbing back to classic SaaS levels. Palo Alto Networks already showed contraction from 77.6 percent (FY24) to 76.4 percent (FY25). The inference cost tailwind is muted by the inference multiplier; workflow orchestrators with thin AI usage retain their margin advantage. The biggest medium-term constraint is power infrastructure. Ground-based grids saturate by 2028; Anthropic and xAI are exploring space-based AI computing partnerships.

M&A pattern continues. Incumbents buy the flywheel pieces they cannot rebuild. Already visible: ServiceNow → Moveworks ($2.85B), Trimble → Document Crunch (Q2 2026), Salesforce → Informatica ($8B), Clio → vLex ($1B), Capital One → Brex ($5.15B), Workday → Sana and HiredScore, Zendesk → Ultimate.ai and Forethought, Automation Anywhere → Aisera. Accenture acquired Faculty, Decho, RANGR Data, Cabel, and NeuraFlash to absorb AI-native deployment capability.

3.4 The Seven Forces

The compounding curves above resolve into seven dynamics that play out at different speeds across the SoR stack. Each force has named broker evidence from 2026.

Force 1: Replacement is layer-specific. The single-phenomenon framing of “agents replace SoRs” misses the structure. Four phenomena are moving at different velocities. Mid-market finance ERP is in active replacement (Campfire 100+ rip-outs, 21% / 18% HubSpot / SAP partner displacement). Customer support runs as substrate-on-top with a new interface (Sierra at named brands, Chime running Sierra plus Decagon at 60M annual inquiries, Intercom Fin at $100M ARR across 8,000 businesses). Enterprise CRM sits as contested substrate stuck in pilot (Agentforce 4K live of 9.5K paid, 42 percent; only 12 percent of Salesforce’s 150K customer base has engaged Agentforce deals at all). Regulated SoRs have hardening incumbent moats (Workday 97 percent gross retention; Epic Connection Hub forces AI-natives to integrate; SAP “Indirect Access 2.0” actively blocks third-party agents from APIs). The unifying claim: the unit of value capture is shifting from “record-of-state” to “record-of-work.” What changes per layer is who captures the work-economy.

Force 2: Value capture stabilizes at 5 to 15 percent vendor take. The most decisive 2026 data: vendors capture 5 to 15 percent of brand savings for specialized vertical agents (Campfire, Decagon broker analysis). Brands keep the rest. Campfire prices in the “low five digits” ($10K to $50K per year) for what replaces six-figure NetSuite implementations. One finance leader: “$5,000 to Campfire captures at least $100,000 in total value of ownership” (a 20× ratio brand:vendor). Decagon at one named client delivered 80 percent reduction in support headcount; if the team was $5M, Decagon captures perhaps $0.5M to $1M of that. Salesforce Agentforce is attempting to move from $1,200 per seat annually to $10,000 per automated workflow (an 8× revenue uplift), conditional on proving human-labor replacement that the pilot-to-production gap suggests they cannot yet deliver. AI is funded by reallocating existing P&L: 26 percent from headcount reduction, 24 percent from legacy infrastructure, 22 percent from IT services and consultants, 11 percent from SaaS license cuts. The bull case that vendors extract 30 to 50 percent of saved labor cost overshoots the empirical data by a factor of two to three.

Force 3: The tax-collector model holds in three positions. Microsoft and Anthropic sit above the user, OpenAI and Bedrock below, the regulated SoRs inside the compliance hub. The horizontal SoR layer between them is squeezed: Agentforce shipped four distinct pricing versions in eighteen months, customers report consumption shock with 30 percent of their annual token allocation consumable in weeks, and SAP’s own bear signal is that “the vast majority of AI use cases that have reached production are built on non-SAP solutions.” The interesting counter-play is that SoR vendors stopped trying to be the orchestrator and started embedding inside one. Workday launched its Sana Self-Service Agent inside Microsoft 365 Copilot in May 2026, and Moody’s deployed MCP Apps inside Claude Desktop the same quarter.

Force 4: Who owns the customer? Whoever owns the workflow runtime. The killer 2026 datapoint: in a head-to-head bake-off at a named financial-services customer (Deutsche Bank, 9 Dec 2025), Sierra achieved >90 percent resolution rate against Agentforce’s lower containment, with implementation cut from 16 weeks to 8 weeks. The customer chose Sierra while Salesforce sat on every byte of customer data. This collapses the SoR’s “conductive spinal cord” framing. The data moat is properly bounded: incumbent telemetry is leveraged for governance and internal workflows more than for raw agentic reasoning. SoR data scale is a governance asset. The orchestration race is shaped, with Microsoft Copilot Studio at 230K+ organizations and 3M+ custom agents (90 percent of Fortune 500); Anthropic Claude Desktop at $44B ARR; Salesforce Agentforce at $800M ARR with 14 percent live-to-paid ratio; Workday ASOR at $400M AI ARR (embedded inside Copilot); ServiceNow AI Control Tower with 1,700+ live customers and 4× overperformance against 2025 targets in regulated industries.

Force 5: The SI customization economy bifurcates. HCLTech’s segmentation tells the story cleanly. AI-Disrupted: 40 percent of the IT services market, shrinking 3 to 5 percent CAGR. AI-Amplified: 55 percent, growing 10 percent+. AI-Native: 5 percent, growing 30 percent. Mid-tier $15M to $100M boutiques are the AI-Disrupted segment. Cognizant “Project Leap”: 12,000 to 15,000 global job cuts ($230 to 320M severance). Oracle 30K cuts. Intel 25K. Tier-1 IT firms (TCS, Infosys, Wipro, HCLTech, TechMahindra) collectively reduced 31,500 lateral positions in FY26. Top-tier (Accenture, Deloitte) sits in the AI-Amplified segment: Accenture AI bookings projected to DOUBLE in FY26; 85,000+ AI / data professionals; 30,000 staff trained on Claude; 740,000-seat Microsoft Copilot deployment (the largest in the world). The AI-native deployment-as-a-product cohort is already capitalized: Campfire Series A $35M (Accel, Jun 2025); Avoca Series B $125M (General Catalyst, Apr 2026); Exaforce Series B $125M (Khosla, May 2026); Vapi Series B $50M (Peak XV, May 2026). Decagon’s “Agent Operating Procedure” framework is the cleanest example of the model: enterprises build and manage agents without custom SI code. Named customers: Notion, Duolingo, Hertz, Eventbrite, Avis Budget, Block, Deutsche Telekom.

Force 6: Regulated SoRs hold, capped at $2 to 4B AI ARR per vendor. Compliance moats are hardening, not eroding. SAP “Indirect Access 2.0” (Apr 2026) actively restricts autonomous third-party agents. AI startups in regulated sectors face 5 to 15 percent of total AI spend on compliance; organizations with mature governance frameworks deploy AI 5× faster (an accelerant for incumbents). The ceiling is real. Workday $400M+ AI ARR runs to a bull case of $4.3B by FY30 (88 percent five-year CAGR); Veeva remains “immaterial” through FY27 with a $6B total revenue 2030 target; Epic stays bundled; SAP €2.1 to 2.5B AI cloud revenue by 2028 (~10 percent of cloud total). The hard implication: no $10B+ AI vendor emerges from regulated verticals alone. Industry-specific AI tools total just $3.5B across all sectors as of early 2026; healthcare AI (the largest sub-segment) is $1.5B. Critical caveat: incumbent AI ARR may be substitution rather than net-new. As AI-driven headcount reductions cannibalize seat revenue at customers, AI upsells become a defensive hedge rather than additive growth.

Force 7: The data moat is governance, not quality. The big rhetorical claim from incumbents in 2026: “Our data makes our AI better.” The evidence: their data makes their AI more compliant rather than smarter. Salesforce Data 360 ingests 112 trillion records (+114 percent YoY); ServiceNow processes 100 billion annual workflows / 7 trillion transactions; Workday Illuminate logs 1.7 billion AI actions. What this data does: powers governance, powers deterministic workflow execution, creates portability friction (a “12 to 24 month project to reconstitute equivalent data pipelines” if you switch). What it has yet to do: produce measurably better customer-facing agent quality (Sierra beat Agentforce in the named bake-off above) or compound a quality moat faster than AI-natives can improve (the Decagon-Sierra delta at Chime closed from 5 to 6 percentage points to under 1 point in twelve months). The SoR data moat is real for governance, a high-margin and defensible position. It remains capped, because compliance regulation does not multiply. The intelligence moat compounds at AI-natives.

3.5 What Would Break the Thesis

Four falsifiers with time-bound triggers grounded in 2026 evidence.

F1: Incumbent tax-collector model stabilizes. The thesis is rejected if by Q3 2027 all of the following hit: Microsoft Copilot ARR crosses $10B (currently $4 to 5B, +250 percent YoY); Salesforce Agentforce live customers cross 10,000 (currently ~4,000); Workday AI ARR crosses 10 percent of total ARR (currently 4 to 5 percent); M365 E7 bundle adoption reaches 5M seats. If all four hit, substrate-with-toll wins at the enterprise tier and the AI-native replacement claim weakens.

F2: AI-native replacement reverses. Rejected if by Q3 2027: Campfire, Rillet, and DualEntry combined customer count crosses 2,500; at least two of the AI-native ERPs reach $100M+ ARR with disclosed numbers; named Fortune 500 brands publicly cite NetSuite or Workday replacement; per-resolution pricing reaches >50 percent of total CS AI spend.

F3: Quality gap closes (the most predictive falsifier). Today, Glean wins 92 percent preference against Copilot’s 8 percent when users are given a choice (Greenwich, 3 Apr). Sierra wins 90 percent+ resolution against Agentforce in bake-offs. The thesis collapses if by Q4 2026 Microsoft Copilot achieves >50 percent preference when given a choice between it, Glean, ChatGPT, and Gemini. That outcome would mean distribution plus data context wins on quality, and best-of-breed dies.

F4: Outcome attribution becomes politically untenable. Rejected if outcome-based pricing adoption drops below 5 percent by 2027 (currently 7 percent per PiperSandler 1Q26). If finance teams refuse to defend variable spend and procurement loathes audit complexity, per-outcome pricing collapses and AI-natives fall back to per-seat. The thesis holds if AELA adoption continues; Salesforce closed 120+ AELAs in F4Q26 at $500K to $5M+ minimum tiers, with named customers including Amazon, Ford, AT&T, Pfizer, Dell, FedEx, PepsiCo, and Williams-Sonoma.

Open Questions

Things I genuinely don’t know, in order of analytical importance.

1. Where does the substrate / model boundary settle? Today, (4) lives partly in markdown skills (editable, substrate-side) and partly in fine-tunes (model-side). Anthropic’s Cowork plugins suggest skills stay editable. Sierra’s per-customer fine-tuned constellation members suggest model weights also absorb policy. The boundary determines whether AI-natives’ substrate work is durable infrastructure or dissolves into the model. The deepest open question.

2. Do slow-binding verticals match fast-binding ones’ compounding velocity? Legal (matter outcomes years-long), healthcare (treatment outcomes months-long), strategy (decisions decades-long) cannot run the loop at coding/CX/accounting velocity. Do dense proxy signals (BigLaw Bench, clinical guideline adherence) substitute well enough? Or do these verticals stay structurally less competitive and AI value gets captured by Concierge/Labor-Bypass plays sitting adjacent rather than GL Replacers?

3. Does cross-customer aggregation survive privacy regulation? GDPR, HIPAA, EU AI Act, and emerging regional data-residency laws all push back on (5). The companies that win their segment will be the ones who design (5) to be regulator-compatible, typically by aggregating patterns rather than records. The legal status of “learned patterns derived from customer data” is unsettled. EU healthcare AI is the live frontier.

4. Does the operator role become high-value-add or get compressed by meta-harness automation? Sierra’s Workspaces show the operator role being productized. But meta-harness research suggests harness self-improvement may absorb the operator’s job over time. If the operator gets compressed, the org-level thesis weakens; no new role replaces the line worker, and the line worker disappears.

5. What is the actual M&A price for a Workflow Wedge? Document Crunch → Trimble (modest); Moveworks → ServiceNow ($2.85B); Campfire → ? (untested but likely $1B–$3B if NetSuite buys). The price multiple a Workflow Wedge gets at acquisition tells us how durable the standalone path is.

6. Does Salesforce’s installed base buy enough time? ~150K customers × multi-year contracts × political resistance to swap = real defensive capacity. Replacement-openness percentages don’t translate immediately into replacements. The 2026–27 ELA renewal cycle is the real test of the 83% openness number.

Sources

Primary. JPMorgan Research, First Principles - AI Agents 2.0: The rise of AI-native new entrants, 8 January 2026 (91 pp): sections on Rillet, Campfire, DualEntry, Digits, Sierra, Decagon, Parloa, Clay, with G2 customer reviews and case studies; the C.H. Robinson AI-agent ROI case; software revenue model evolution. Redpoint Ventures, 2026 Market Update: CIO replacement openness data, AI-native P&L economics, two-playbook framework.

Frameworks referenced. The Harness Thesis (Verifiability × Regulation; Delegatability × Last-Mile; harness half-life). Vertical AI Platforms three-question moat test (proprietary data + regulatory + transaction embedding). The Memory IS Learning thesis. The Fintool 10-moats analysis (“agent IS the bundle”; “software is becoming headless”). Context-engineering landscape ($10B+ historical context systems). Sierra constellation architecture analysis.

Web evidence (selection). Sequoia partnership posts: Day.ai, Rillet, Rox. TechCrunch: Campfire $35M Series A, Sierra $10B raise, Sierra $100M ARR, Decagon Series D / tender, Doss $55M. BusinessWire: Basis $100M / $1.15B val. Bloomberg: Decagon $4.5B Series D. CNBC: OpenEvidence $12B val. STAT News: Epic AI Charting threatens scribe. Josh Bersin: Workday “Platform of Agents” reinvention. ABA Journal: Clio acquires vLex $1B. Crunchbase: Clay $100M Series C. BlackLine Verity launch. NetSuite 2026.1. SAP at Hannover Messe 2026.

Living document. The five-primitive flywheel framework is the central new contribution; the segment evidence is the proof; the compounding-rate argument is the load-bearing claim. Falsification conditions are listed in §3.5. Updates on receipt of new evidence.

Appendix A: The Segments in Depth

The main essay walks through finance as the exemplar segment. This appendix carries the same treatment across the other four SoR segments (CRM, CX, ITSM, Regulated / Operational) plus the broader set of ERP-replacement companies that sit alongside Campfire. Each segment follows the same shape: incumbents, why the segment behaves the way it does, one company in depth, the comparison set, and what the segment teaches.

Campfire sits in a wider field of GL Replacers. The seven that matter most for the 2026 thesis:

By 2027 we expect one or two of these companies above $100M ARR. By 2028 the question is whether NetSuite acquires one or rebuilds.

Incumbents. Salesforce Sales/Service Cloud, HubSpot, Microsoft Dynamics 365 CE, Zoho, Pipedrive. Sales engagement: Outreach, Salesloft (merged with Clari Dec 2025, ~$450M combined ARR), Apollo, Gong. 83% CIO replacement openness, the highest of any category.

Why CRM is the most exposed SoR. All of Salesforce’s lock-in was UI and process, with no architectural or data root. Three Salesforce moats were UI-based: admin labor encoded as custom objects, validation rules, page layouts; rep muscle memory across pipeline views and dashboards; the SI implementation economy. When the agent IS the interface, all three vaporize. The data is portable (it’s the customer’s, not Salesforce’s), the regulatory lock is absent (CRM is unregulated), and the transactions don’t pass through Salesforce. The three-question moat test (proprietary data, regulatory, transaction embedding) yields zero structural moats for Salesforce.

Clay, system-of-action for GTM

Kareem Amin (ex-WSJ VP Product) and Nicolae Rusan, founded 2017. $204M total; $100M Series C at $3.1B (CapitalG, June 2025) following a $1.5B Sequoia tender and a $1.3B Series B extension. $100M ARR (CEO confirmed to NYT, tripling YoY). Customers: OpenAI, Anthropic, Canva, Intercom, Rippling. 1.5B lifetime AI agent tasks. Credit-based pricing.

1. Outcome: enriched record matched correctly, signal that triggered a converted outbound, response-rate uplift. (2) Trajectory: every workflow’s full execution trace preserved, which data provider was tried first, which was tried second, what was rejected, which signal triggered which action. (3) Attribution: customers A/B test workflows; recipes can be benchmarked. (4) Substrate: spreadsheet-based programmable workflows plus Claygent (research agent). The substrate is the workflow recipe, a versionable, shareable, copy-pasteable unit of GTM logic. (5) This is Clay’s real moat. 60+ Clay Clubs worldwide, 400+ GTM engineer roles posted in a single spring 2025 hiring cycle, customer-led GTM agencies scaled to $1M+ ARR within a year. The recipe library aggregates across all customers.

What Clay does that Salesforce couldn’t. Salesforce processes Activity rows. Clay processes signals (job changes, news, intent, technographics) and executes on them via a programmable workflow that is itself a learned artifact. The “GTM engineer” role didn’t exist before Clay because the substrate didn’t exist before Clay. Salesforce’s response (Data Cloud) adds signal but cannot match the recipe-library compounding, because Data Cloud is not a cross-customer workflow surface.

Customer evidence (JPMorgan). OpenAI: 2× enrichment coverage, 100% research automated, 8,500+ enrichment runs by team members. Vanta: 80%+ enrichment coverage, 1,000+ contacts/month added. CapitalG partner Jane Alexander: “Clay is the first and only company to take an engineering approach to go-to-market.”

The rest of the segment

What this segment teaches. Three archetypes coexist. CRM Replacer (Day.ai, Attio, Clarify, Reevo, Monaco) is a UX race with the weakest moat. Role Replacer (11x, Rox) is outcome-priced labor replacement with severe trust risk, as 11x demonstrated. Signal Graph (Clay, Common Room, Apollo, Unify) sits alongside the SoR and owns the signal layer. The Signal Graph archetype is the most defensible because (5), the recipe library plus contact graph plus signal feed, is genuinely cross-customer. Greenfield CRM rebuilds without (5) compete on UX and lose to whoever ships fastest.

Incumbents. Zendesk, Salesforce Service Cloud, ServiceNow CSM, Intercom, Freshworks, Kustomer (Meta-owned), Front. CCaaS: Five9, NICE, Genesys, Talkdesk. CIO replacement openness: 56%, second-highest after CRM.

Why CX has the densest flywheel signal of any segment. Resolution is per-conversation, per-minute verifiable. The unit of work is bounded (one ticket). The outcome surface is huge. Sierra alone handles hundreds of millions of interactions per year, and conversation latency is seconds rather than days. (1) and (2) are dense by default. (3), (4), (5) are where the moats live.

Sierra, the constellation harness

Bret Taylor + Clay Bavor, founded 2023. $635M total, $10B valuation (Greenoaks, September 2025). $100M ARR in 7 quarters since launch (TechCrunch, November 2025), growing from ~$20M a year prior. 50%+ of customers have $1B+ revenue; 20%+ have $10B+. Reach: 95% of US Black Friday shoppers, 50% of US healthcare families, >90% of US media ecosystem, >70% of US fintech. Outcome-based pricing, “if the AI agent has to transfer to a real person, it’s free” (Bret Taylor).

1. Outcome: resolution-without-escalation per conversation; the pricing model gives Sierra commercial incentive to maximize signal density. (2) Trajectory: every conversation, with all branching and tool-call decisions, preserved; 2M+ conversations per month feed the loop. (3) Attribution: stress-tests with trick questions before launch, built-in guardrails and audit systems, Workspaces model for parallel testing, supervisor models acting as “Jiminy Cricket” on factuality and policy. (4) Substrate: the constellation of 15+ models is the substrate, with adaptive routing per task (low-latency for tool calls, high-precision classifiers for fraud, long-context reasoners for knowledge); AIMD admission control adapted from TCP congestion control; planner-executor-validator pattern. Workspaces productize the substrate edit cycle with branches, reviews, and controlled releases. (5) Aggregation: routing optimization compounds across deployments within brand-isolation constraints.

What Sierra does that Zendesk couldn’t. Zendesk’s data model is ticket → conversation → customer → agent assignment. Sierra’s primitive is the agent run, a graph of (intent → plan → tool calls → KB references → policy checks → resolution outcome → memory update). Zendesk’s macros plus KB articles are static substrate; Sierra’s constellation is dynamic substrate that learns. Zendesk’s per-seat economics ($55 to $169 per agent per month plus a $50-per-agent AI add-on plus $1.50 per automated resolution) sit in commercial conflict with deflection. Sierra is paid more only when a human is bypassed.

Customer evidence (JPMorgan). ADT: 2 million+ monthly interactions handled autonomously, 70% containment, “warm, conversational, empathetic” tone. WeightWatchers: 70% containment in week one, CSAT 4.6/5. “I knew the AI agent would answer questions quickly, but I didn’t expect the responses to be so genuine and empathetic. I was reading chat transcripts with members exchanging heart emojis with the AI agent.” Maureen Martin, VP Customer Care, WeightWatchers.

The rest of the segment

What this segment teaches. CX is the most-validated agent-native SoR replacement, because (1) and (2) are densest, (3) is most measurable, and customer pain is acute. Sierra dominates F1000 enterprise; Decagon dominates high-growth tech enterprise; Parloa dominates voice-first enterprise. By 2027, Service Cloud’s per-seat revenue is in visible decline as deflection eats the seat count. Zendesk’s hybrid pricing is a transitional artifact that collapses to per-resolution.

Incumbents. ServiceNow, Atlassian Jira Service Management, BMC Helix, Ivanti.

Serval, the access-management wedge

$127M total, $1B valuation (Sequoia, December 2025). ~$50M ARR; revenue grew 500% since Series A in August 2025. Wedge: provisioning access requests (SaaS apps, permissions, on/offboarding), 30–50% of help desk volume.

1. Outcome: the access either got granted to the right person on the right system or it didn’t, binary and verifiable with minute-grain feedback. (2) Trajectory: chain of identity lookups, group memberships checked, approval routing decisions. (3) Eval on access-grant accuracy across hundreds of SaaS integrations. (4) Substrate: identity policies plus per-app workflow templates plus supervisor agents. (5) Cross-customer access patterns: Salesforce plus Slack plus Notion plus Datadog plus GitHub plus AWS is the same canonical onboarding shape across thousands of orgs.

What Serval does that ServiceNow couldn’t. ServiceNow’s app marketplace was always thin, with every customer building integration scripts manually. Serval ingests hundreds of SaaS integrations natively. Access-grant is one of the few IT processes where (1) is binary AND (5) is highly transferable. ServiceNow’s $2.85B Moveworks acquisition in December 2025 was the cleanest possible admission: rebuilding the front-end harness from scratch was slower than buying it.

The rest: Moveworks (now ServiceNow), F500 employee front-door covering HR + IT + Finance intent surface. Atomicwork ($40.3M, $25M Series A from Khosla/Okta Ventures), agentic ITSM plus employee experience. Aisera (acquired by Automation Anywhere), pre-LLM AI that couldn’t compete with Serval’s substrate quality.

What this segment teaches. Two archetypes hold: horizontal employee front-door (Moveworks-style, high TAM but commoditized post-acquisition) and vertical access wedge (Serval-style, deeper compounding because the workflow is more verifiable). ServiceNow’s installed base will absorb Moveworks but cannot ship a cross-app substrate as deep as Serval’s by 2027.

Healthcare, legal, HRIS, real-estate / construction / field-service. The pattern across all four: regulation depth is inversely correlated with SoR replaceability. Healthcare EHR (Epic), HRIS (Workday), BigLaw matter management (iManage) are not replaceable, and AI-native value capture is forced into adjacent layers: scribe, RCM, recruiting, vertical practice rebuilds, labor delivery. Lightly-regulated SoRs (mid-market real estate operations, SMB field service) are full-replacement targets.

Healthcare (EHR-adjacent)

Epic is not replaceable. HIPAA, HL7/FHIR pipes, claims adjudication, hospital deployment cycles measured in years. CIO openness to replacing the EHR sits around zero. AI-native value capture is forced into adjacent layers.

What this teaches. Epic absorbs ambient scribing via Workshop with revenue-share extraction, so avoid pure scribe plays. Best plays: RCM and coding (verifiable transaction loop generates real training signal), direct-to-clinician reasoning that bypasses both EHR and HIPAA-aggregation constraint, and patient-facing labor displacement.

Legal (Practice & Matter Management)

Incumbents. Clio (200K+ lawyers; $5B val post-vLex acquisition), MyCase, NetDocuments, iManage (BigLaw DMS), LexisNexis/Westlaw/Bloomberg Law.

What this teaches. Avoid horizontal practice-management replacements (Clio absorbed vLex; iManage is too entrenched). Best path: vertical practice-area rebuilds (Eve in plaintiff PI; immigration, IP, and M&A as next targets). Harvey is a workflow layer with a defended-but-not-fortress moat; its long-term durability requires moving up to outcome ownership.

HRIS / Recruiting

Workday, Rippling, ADP, BambooHR are structurally protected. FLSA, EEOC, ACA, multi-state-payroll, and SOX plumbing is identical to Epic’s. No AI-native HRIS unicorn exists. Workday’s “platform of agents” repositioning (April 2026) and Sana acquisition signal the absorption pattern. CIO openness to replacing HRIS sits at 25 to 35 percent, the lowest of any segment in this study.

The unlock is in the labor flow, not the SoR.

What this teaches. Bypass the SoR via labor delivery. Best plays: Mercor-style labor-as-a-service, vertical recruiting agents (clinical, sales, blue-collar) where outcome is measurable, and compliance-automation wedges (immigration, multi-state payroll). Avoid horizontal HRIS replacement. Workday will eat them.

EliseAI, the operational SoR by interaction-layer expansion

$2.2B+ valuation (Series E). It touches 1 in 12 US apartments, covers 70 percent of NMHC Top 50, and runs 90 percent interaction automation. (1) Outcome: lease conversion (binary, high-frequency). (2) Trajectory: leasing-conversation graph at apartment density. (4) Substrate: Fair-Housing-compliant policy substrate. (5) Strongest cross-customer aggregation in any operational vertical, with conversation patterns aggregating across one-in-twelve US apartments.

EliseAI is the cleanest “AI-layer becomes operational SoR” arc in the study. The path: own the resident interaction, then own delinquency recovery, then own maintenance triage, then eventually displace the GL. Yardi increasingly just becomes the GL while EliseAI owns the operational layer. The same playbook is now extending into healthcare.

Construction and field service. Procore is becoming an agentic platform itself; ServiceTitan is mostly safe at enterprise but vulnerable at SMB (around 60% replacement openness). AI-native value: Document Crunch (acquired by Trimble Q2 2026, modest exit, confirming the “AI-layer thesis,” no path to standalone SoR), Buildots (computer-vision construction progress as a data moat), Trunk Tools (Procore-side, with Jasper-risk if Procore native AI catches up), Fieldproxy / Fixlify / Quantra (AI-native ServiceTitan alternatives at SMB).