Inviting the Agent In | Inferential<br>Intuit told 3,000 employees on Wednesday that their jobs were gone — 17% of an 18,200-person global workforce — in an internal memo sent hours before the Q3 FY26 earnings release. The same release raised full-year revenue guidance to $21.341–21.374 billion, posted Q3 revenue of $8.6 billion (+10%), Online Ecosystem revenue of $2.5 billion (+19%, +22% ex-Mailchimp), and non-GAAP EPS of $12.80 (+10%). The stock entered the print down 41% year-to-date. The arithmetic is now familiar — beat the quarter, raise the guide, cut the workforce, watch the multiple compress — but the mechanism Intuit is funding with the cut is different from the ones SAP, Atlassian, and Cisco have used in the past six weeks.<br>CEO Sasan Goodarzi’s framing is that the company is becoming “faster, leaner, and more focused.” The actual reallocation is toward two multi-year partnerships, both visible in product before the layoff hit the calendar. The Anthropic deal, announced in February, brings TurboTax, Credit Karma, QuickBooks, Mailchimp and Intuit Enterprise Suite directly inside Claude via MCP integrations — a TurboTax customer can get a real-time refund estimate and connect to a live tax expert without ever opening Intuit’s own app. The OpenAI deal, announced alongside the Q3 print, layers a $100M+ commercial agreement on top of the same architecture, embedding GPT models inside Intuit’s own products and surfacing Intuit’s tax, accounting, and financial workflows inside ChatGPT.<br>The Anthropic press release is unusually direct about what the company is buying: “Intuit’s world-class financial intelligence will be surfaced directly inside Anthropic products.” Translated into seat economics, that sentence describes a TurboTax customer whose entire experience is routed through Claude — the UI, the workflow chrome, the prompt-to-resolution loop, the front door. What remains inside Intuit is the tax engine, the expert escalation, the privacy boundary, the audited compliance layer, and the data Intuit has accumulated about 90 million returns. The seat doesn’t disappear from the income statement immediately — TurboTax revenue still grew 7% to $4.4 billion this quarter — but the unit being defended at renewal stops being a logged-in user and starts being an API call inside someone else’s product.<br>This is the canonical “AI eats the SaaS UI” trade made voluntarily and in public. Atlassian’s Self-Funded Pivot cut headcount to fund a seat-stapled agent SKU that keeps customers inside Jira and Confluence. ServiceNow’s Action Fabric at Knowledge 2026 built a metered toll-bridge so Claude and other external agents pay to traverse the platform. SAP’s Tables and Gates bought tabular-AI capability and closed the API to unauthorised agents. Intuit is doing the opposite of all three on the inbound side — opening the front door so wide the assistant is the front door — while doing the same thing as Atlassian on the cost side, severing seats to pay for it.<br>The financial signature differs in one important way. Atlassian, ServiceNow, and SAP defend enterprise platforms with multi-year contracts and procurement-tested switching costs. Intuit’s exposed franchises are consumer (TurboTax, Credit Karma) and SMB (QuickBooks, Mailchimp) — segments without procurement-mediated retention, where a Claude-mediated tax conversation that auto-imports a W-2 and routes to a live expert is a credible direct substitute for the standalone TurboTax app, not a complement to it. The market is pricing that gap. Intuit’s competitors at the application layer who reported earlier this cycle — Palantir’s Above the Model Layer print and Datadog’s first $1B quarter inside The Substitution Line — produced 85%+ gross margins and 30%+ revenue growth defending application-layer context with no comparable consumer-disintermediation risk. INTU at -41% YTD is the same fortress thesis being tested at the layer of the stack where the fortress is thinnest.<br>The hypothesis this complicates is whether Above the Model Layer’s cleavage point — value expands where customer-specific context cannot be re-derived externally — holds when the front door is the AI assistant rather than the application. Intuit’s bet is that 90 million tax returns of accumulated data, the network of human experts, the compliance certifications, and the IRS submission pipeline cannot be re-derived inside Claude. That bet is structurally identical to Palantir’s ontology bet — but consumer tax is not a defence-contractor data trust, and a small-business owner asking Claude for an invoice template has weaker switching cost than a Fortune 500 customer with a five-year ontology investment. The fortress quadrant may exist at the consumer/SMB layer, but its walls are thinner.<br>Hypothesis 54% → 54%. Verify the UI-compression mechanism (a flagship vertical-SaaS leader...