For buyers watching AI reshape software.
Outcome pricing goes GA and the data cloud gets a $188B endorsement: the week the agentic business model stopped being an experiment
Week 29 of 2026 · July 18, 2026
Big read
Salesforce made Agentforce Help Agent generally available at $2 per successful autonomous resolution — no charge when the agent escalates to a human or the customer is dissatisfied, sold in 1,000-resolution pack minimums. The unit carries over from the per-conversation model, but the trigger condition changed: the vendor now gets paid only when the outcome lands. Salesforce brought vendor-reported proof — 4.3 million inquiries on its own support site, 70% resolved autonomously — and a pending definitive agreement to acquire Fin, the SMB customer-service agent used by 30,000+ companies. Two days earlier, Sierra launched Horizon, extending its agents from single conversations to long-horizon goals spanning weeks or months: loan origination, prior authorization, specialist-referral scheduling. Read together, the largest CRM incumbent and the best-funded CX startup converged on the same commercial architecture in the same week. Outcome pricing is no longer a pilot SKU to watch — it is the default track, and W27-W28's Zendesk, HubSpot, and Coupa moves now look like the early adopters rather than the outliers.
Capital delivered the same verdict one layer down. Databricks put Inkling — Thinking Machines' open model — on Unity AI Gateway day-0, and the next day announced a Coatue-led strategic round at a $188 billion valuation, explicitly earmarked for Unity AI Gateway, Genie, and Lakebase. The caveat matters: this is a signed term sheet expected to close later in the summer, not closed capital — but the earmark is the signal. The largest private-market valuation event of the year is a bet on the gateway, not the model. The CEO's own framing — 'tokenmaxxing to valuemaxxing' — says the quiet part: selling tokens is a commodity business; owning the control plane between enterprise data and every model is the platform business. Snowflake reached the same conclusion by another route, signing its largest-ever AWS commitment at $6 billion, tied explicitly to agentic AI adoption with Marketplace go-to-market and migration incentives attached.
The connective tissue between the two stories is the thesis this publication exists to track: if the application is priced per outcome and the model is a swappable component behind a gateway, the durable layers are the data platform below and the measured workflow above. This week showed the distribution channel forming between them. Clay took its prospecting data and functions GA inside six assistants — Claude, ChatGPT, Claude Code, Codex, M365 Copilot, and Glean — via MCP, with admin permissioning, function allowlists, and per-rep credit budgets, and no MCP surcharge. Glean answered with the inverse play, pulling seven sales tools into its index. Atlassian conceded the coding-agent layer entirely — Jira work items are now assignable directly to Claude Code, Cursor, or GitHub Copilot — and claimed the context and governance layer instead, shipping AI cost-vs-output measurement alongside. The pattern is consistent: vendors are deciding, in public, which layer they can actually defend.
The governance clock is the urgent part. Microsoft message center notice MC1422074 confirms that OpenAI-operated GPT-5.6 becomes an M365 Copilot subprocessor that auto-enables for eligible tenants on July 24 unless admins set the control to 'No users.' That changes the data-processing chain under the DPA, and silence is consent. This is the second suite auto-enable event in three weeks — the W28 pattern of defaults replacing sales cycles, now applied to the data-processing chain itself. The same window produced the counter-current: OpenAI's GPT-Red disclosure claims an internal adversarial model hardened GPT-5.6 Sol to a 0.05% failure rate on OpenAI's own direct-injection attack corpus (vendor-reported; a corpus failure rate, not coverage of all possible attacks), making injection-resistance a published competitive axis for the first time — while SAP moved the opposite direction from everyone else, restricting direct third-party autonomous API access to SAP data and positioning Joule as the single mandatory window. One secondary source carries that report, so treat it as provisional — but if it holds, SAP is testing whether a closed gateway can survive in a week when $188 billion just priced the open one.
What to do with this week: CIOs have one hard deadline — audit the MC1422074 setting and set tenant posture before July 24, then fold injection-resistance evidence and subprocessor-chain disclosure into every AI vendor questionnaire. Function heads negotiating outcome-priced SKUs should contract the outcome definition, measurement rights, and audit access before signing — Salesforce's own carve-outs (no charge on escalation or dissatisfaction) show the definitions are negotiable now and will not be later. SaaS investors should reprice per-seat vertical SaaS against a world where the two data clouds just raised the cost of competing for the control plane, and should read the Databricks term sheet as directional but not closed. Vertical-AI founders: Clay just demonstrated the MCP distribution playbook — ship your function into every assistant your buyer already pays for, with admin controls as the enterprise unlock — and Sierra's Horizon shows the next defensible wedge is the long-horizon workflow, not the conversation.
Vertical movements
2026-07-14 · Anthropic · Other · Frontier Lab · Unknown
Claude for Teachers
District IT and curriculum leaders now have two frontier labs bidding for the teacher relationship — this answers ChatGPT for Teachers from November 2025 — so the evaluation window is now, while the product is free and the paid district tier is still 'coming soon.' The Detroit Public Schools evaluation pilot and the MCP connectors are the tells: the land-grab is for the workflow and the edtech integration surface, not the chat seat, and pricing leverage disappears once a district's curricula and skills libraries are built on one lab's stack.
Sources Anthropic
2026-07-16 · Google · Research · Frontier Lab · Usage Based
Grounding with Parallel Web Search (Gemini Enterprise Agent Platform)
This unbundles the context layer from the model: grounding is now a swappable, pay-per-grounded-query line item on the GCP invoice rather than a take-it-or-leave-it model feature. Architects building research and analyst agents should re-quote grounding costs as a separate procurement axis, and the zero-data-retention option gives regulated buyers a concrete requirement to demand from every other assistant vendor.
Sources Google Developers Blog
2026-07-14 · Microsoft · Operations · Incumbent Saas · Unknown
SharePoint Copilot Apps (SPFx 1.24 preview)
The Copilot chat canvas is becoming an application runtime, which means intranet and workflow teams should redirect net-new lightweight app builds toward the surface employees already have open rather than standalone SharePoint pages. The end-user rollout completes July 20 whether or not IT has a governance story for who can publish action-taking apps into the canvas — write that policy this week, and note the preview name may change before GA.
Sources Microsoft 365 Developer Blog
2026-07-14 · Microsoft · Security · Incumbent Saas · Unknown
M365 Copilot OpenAI-subprocessor setting (MC1422074)
This is the week's hardest deadline: the change alters the data-processing chain under the DPA, and inaction opts the tenant in — silence is consent. CISOs and privacy officers should set the admin control before July 24, notify their DPO of the subprocessor change, and treat this as the template for how suite vendors will introduce third-party model operators from now on: via message-center notice and default-on, not via contract amendment.
2026-07-15 · OpenAI · Security · Frontier Lab · Unknown
GPT-Red adversarial hardening
For the first time a frontier lab is marketing quantified injection resistance, which means CISOs can — and should — fold injection-resistance evidence into vendor security questionnaires and demand comparable disclosure from every agent vendor touching enterprise data. Two cautions: the figure is vendor-reported until the promised preprint enables independent scrutiny, and it is a failure rate on the vendor's own attack corpus — not coverage of the space of possible attacks. Treat the disclosure precedent as durable regardless: the labs just made agent security a spec-sheet line item.
Sources OpenAI
2026-07-15 · OpenAI · Engineering · Frontier Lab · Unknown
Codex Micro (with Work Louder)
The hardware is a curiosity; the workflow it productizes is not — OpenAI is betting the human-supervising-N-agents pattern is mainstream enough to merchandise, backed by Codex's claimed 5M+ weekly users with 1M+ outside software development. Engineering leaders should read this as a signal to formalize the agent-supervision role (queues, status visibility, per-agent cost) rather than leaving multi-agent oversight to individual developers' improvisation.
Sources Ars Technica
Incumbent responses
2026-07-17 · Salesforce
Agentforce Help Agent GA + pending Fin acquisition
The largest CRM vendor just made outcome pricing the default track for support, with vendor-reported proof (4.3M inquiries, 70% autonomously resolved on its own site) and a downmarket acquisition to cover the SMB flank. Buyers should negotiate the resolution definition, dissatisfaction criteria, tiering, and renewal protection now — the carve-outs are the contract, and they will harden once the model is proven.
Sources Salesforce; Salesforce Ben
2026-07-16 · Databricks
Unity AI Gateway (Inkling day-0) + $188B strategic round term sheet
The valuation event is a signed term sheet expected to close later in the summer — directional, not closed capital — but the earmark tells investors and buyers where the fight is: the gateway between enterprise data and every model, not any single model. The CEO's 'tokenmaxxing to valuemaxxing' framing is the thesis in four words; buyers standardizing on the gateway should price the lock-in that comes with the convenience.
Sources Databricks
2026-07-14 · Snowflake
$6B AWS strategic collaboration agreement
Both major data clouds now have this week's receipts that the platform war is being fought at their layer — but the deal also deepens two-vendor concentration for joint Snowflake-AWS customers. Buyers should harvest the migration incentives while they are rich and model the exit cost at the same time; the discounts are the price of the concentration.
Sources International Business Magazine
2026-07-15 · Atlassian
Jira agent assignment + Jira Coding Agent + DX AI cost management
Atlassian is publicly conceding the coding-agent layer to the labs and claiming the layer it can defend: context and governance via Teamwork Graph, with the ticket as the unit of agent orchestration and cost accountability. Engineering leaders get a free orchestration surface out of this; the strategic read is that the system of record survives by becoming the agent dispatcher — and every SaaS vendor without a defensible context layer should be nervous.
Sources Atlassian
2026-07-14 · SAP
Joule as the single gateway to SAP data
Caveat first: this rests on a single secondary source, so treat it as provisional until SAP confirms in writing. If it holds, SAP customers should get the API-access roadmap in writing before any renewal — a closed gateway means every agent vendor in the estate must route through Joule on SAP's terms — and the July 23 Q2 results are the first test of whether customers reward or punish the posture.
Sources Chosun Biz (single secondary source; unconfirmed by SAP)
Startup signals
2026-07-16 · Support · last raise $950M at $15B+ (May 2026)
Sierra
The best-funded CX startup just moved its outcome unit from the resolved conversation to the completed multi-week workflow — into finance and healthcare operations where the outcome is worth orders of magnitude more than a support ticket. Function heads in lending and healthcare ops should expect outcome-priced agents on their processes within two quarters; the vendor-reported LINEMO metrics (97% resolution, 93% CSAT) are the reference story, unverified.
Sources Sierra
2026-07-15 · Healthcare · $700M at ~$7B
Neko Health
A 4x step-up in eighteen months on 100k+ scans and 350k+ registered users says investors are pricing AI-native care delivery — owning the clinic, the scan hardware, and the longitudinal data — not health-tech software. Health-system strategists should treat the NYC launch as the start of direct-to-consumer preventive competition; investors should note the pattern rhymes with legal's Norm Ai: the AI-native services firm is the venture product.
Sources Neko Health
2026-07-14 · Sales
Clay
This is the cleanest demonstration yet of MCP as a vertical-SaaS distribution channel: ship the function into every assistant the buyer already pays for, and make admin controls the enterprise unlock. Sales-ops leaders get Clay inside the reps' existing surfaces without a new seat; founders should copy the playbook — and note that 'no MCP surcharge' is a pricing shot at everyone who planned to toll the protocol.
Sources Clay
2026-07-14 · Sales
Glean
The two architectures of assistant-era distribution launched head-to-head: Clay pushes functions out to every assistant, Glean pulls every tool into one index. Buyers should recognize they are choosing a topology, not a feature — the index-centric play concentrates governance in one place, the MCP play distributes it — and the choice determines where sales context lives for years.
Sources AP News (Business Wire release)
2026-07-14 · Operations · $60M Series B (~$100M total)
InstaLILY
Strategic capital from two major industrial operators validates the forward-deployed-engineer model as a product, not a services line — the same pattern W28 flagged when a hyperscale incumbent restructured sales around embedded delivery engineers. Operations leaders in industrial verticals should shortlist the category; the claimed 5x revenue growth is vendor-reported, but the customer-as-investor signal is not.
Sources SiliconANGLE
2026-07-13 · Legal
Filevine + Codes Health
The platform-embed pattern matters more than the partnership: vertical AI increasingly distributes by embedding into the system of record the buyer already runs, not by selling a new tool. Legal-ops leaders in personal injury and mass tort should pilot against the turnaround claim; founders should note that the legal vertical now rewards embeds into incumbents as much as standalone products.
Sources PR Newswire
Pricing shifts
2026-07-17 · Usage Based → Outcome Based
Salesforce
Agentforce Help Agent went GA at $2 per successful autonomous resolution — the unit price carried over from the per-conversation model, but the trigger changed: no charge on human escalation or customer dissatisfaction, sold in 1,000-resolution pack minimums. Buyers should negotiate the resolution definition, dissatisfaction criteria, volume tiering, and renewal protection now, while the vendor still needs reference customers to prove the model.
Sources UpperEdge; Salesforce Ben
2026-07-13 · Seat Based → Hybrid
OpenAI
ChatGPT Work now layers consumption credits atop subscriptions — agent-run pricing began July 6 — and Enterprise/Edu admins gained workspace-default, group, and individual spend controls plus a credit-request workflow. The seat is becoming the floor and agent consumption the variable bill; finance and IT should stand up credit-budget governance now, before agent fleets scale the variable line past the subscription line.
Sources OpenAI
Vertical scorecard
As of 2026-07-18
| Vertical | Leader | Challenger | Read |
|---|---|---|---|
| Support | Salesforce (Agentforce Help Agent) | Sierra (Horizon) | Leadership reshuffled on the week's evidence: Salesforce GA'd $2/resolution with the Fin acquisition pending downmarket, while Sierra moved the outcome unit to multi-week workflows. ServiceNow's answer comes at Q2 earnings Jul 22. |
| Commerce | Salesforce (Agentforce Commerce) | OpenAI (ChatGPT checkout) | Carried from W28; no in-window commerce-specific move. First ChatGPT checkout conversion data remains the catalyst to watch before peak season. |
| Engineering | OpenAI Codex | Cursor | Codex claims 5M+ weekly users and shipped supervision hardware (Codex Micro); Jira now dispatches work items to both. The SpaceX-Cursor all-stock close expected in Q3 is the vertical's pending structural event. |
| Legal | Harvey | Norm Ai | Harvey still leads on enterprise penetration; the Filevine-Codes Health embed shows the vertical's distribution shifting toward embedding AI into the system of record rather than selling standalone tools. |
| Research | Anthropic (Claude Science) | Google (Gemini Enterprise Agent Platform) | Challenger updated: Google unbundled web grounding into a marketplace-billed, zero-data-retention component — the first move to make the context layer separately procurable. |
| Marketing | Adobe | Salesforce | Carried from W28; no in-window marketing-specific move. Watch whether outcome pricing's support-side success pulls marketing SKUs toward priced outcomes next. |
| Operations | ServiceNow | Microsoft (SharePoint Copilot Apps) | Challenger updated: action-oriented apps in the Copilot canvas (rollout completes Jul 20) put approve/triage/checklist workflows inside the suite surface — plus InstaLILY's strategic-backed round funds the forward-deployed-engineer lane. |
| Finance | Microsoft Dynamics | Sierra (Horizon loan origination) | Challenger updated: Horizon's long-horizon agents target loan origination directly — outcome-priced workflow completion is entering finance operations from the CX side, not the ERP side. |
| Security | Injection-hardened frontier stacks (GPT-Red) | Agent evaluation vendors (Patronus, Noma) | Leader reframed: OpenAI made injection resistance a published competitive axis (0.05% corpus failure rate, vendor-reported), while MC1422074's Jul 24 auto-enable makes subprocessor governance the week's live control gap. |
| Other | Data-cloud gateways (Unity AI Gateway, Snowflake-AWS) | Suite-default harnesses (ChatGPT Work, Claude Cowork) | Reframed from W28: capital just priced the control plane — $188B term sheet earmarked for the gateway, $6B AWS commitment — so the platform question is now data-cloud gateway versus suite harness, with MCP as the connective tissue. |
Architecture watch
Outcome pricing crosses from experiment to default track
The largest CRM incumbent GA'd resolution pricing and the best-funded CX startup extended the outcome unit to multi-week workflows in the same week — the experiment phase is over, and the negotiating leverage is shifting to vendors as the model proves out. Buyers should demand three things in every outcome-priced contract while they still can: a precise outcome definition (what counts as 'resolved,' who decides 'dissatisfied'), measurement rights over the vendor's telemetry, and audit access to the resolution log. The carve-outs are the contract.
- Examples
- Salesforce Agentforce Help Agent ($2/resolution GA), Sierra Horizon (long-horizon outcome units), Zendesk / HubSpot / Coupa (prior weeks' resolution and savings SKUs)
Sources Salesforce; Sierra; UpperEdge
The data platform consolidates as the multi-model control plane
The week's capital events all priced the same layer: the gateway that sits between enterprise data and a rotating cast of models. Databricks earmarked the largest private valuation event of the year for its gateway, Snowflake bought distribution and migration incentives at $6B, and even the productivity suite now routes multiple model operators behind tenant-level governance controls. Buyers should treat gateway selection as the decade's platform decision — it determines model portability, governance surface, and where switching costs accumulate — and negotiate exit terms at entry.
- Examples
- Databricks Unity AI Gateway (Inkling day-0, $188B term sheet earmark), Snowflake $6B AWS agentic-AI commitment, M365 Copilot multi-model with tenant-level subprocessor controls
MCP becomes the vertical-SaaS distribution channel
Clay's GA is the template: ship the vertical function into every assistant the buyer already pays for, and make admin permissioning and credit budgets the enterprise unlock — no new seat, no new surface, no MCP toll. Glean's same-day connector wave is the counter-architecture (pull tools into one index rather than push functions into many assistants), and the choice between the two topologies determines where context and governance concentrate. Founders should treat MCP distribution as a launch requirement now; buyers should decide their topology deliberately rather than inheriting it one connector at a time.
- Examples
- Clay across six assistants (allowlists, per-rep credit budgets, no surcharge), Glean's inverse index-centric connector play, Atlassian Rovo MCP Server
Sources Clay; AP News; Atlassian
Ticketing and canvas layers become agent-orchestration runtimes
The systems of record are converting their coordination surfaces into agent dispatchers: the Jira ticket now routes work to whichever coding agent the team prefers, the Copilot canvas now hosts action-taking apps, and OpenAI is merchandising the human-supervising-N-agents workflow as hardware. The pattern predicts the durable incumbent position — own the queue, the context graph, and the cost accounting, and let the agents underneath commoditize. Engineering and IT leaders should instrument agent cost-versus-output now, because the dispatch layer is where that telemetry will live.
- Examples
- Jira work items assignable to Claude Code / Cursor / GitHub Copilot with DX cost-vs-output measurement, SharePoint Copilot Apps (action apps in the Copilot canvas), Codex Micro (multi-agent supervision hardware)
Sources Atlassian; Microsoft 365 Developer Blog; Ars Technica
Watchlist
Jul 20
SharePoint Copilot Apps worldwide end-user rollout completes
Action-taking apps reach every eligible tenant's Copilot canvas — the governance question of who can publish apps into the canvas becomes live for IT this week, ahead of GA later in 2026.
Jul 22
ServiceNow Q2 earnings
The first monetization read on Now Assist — and the incumbent-support answer to Salesforce's $2/resolution GA. Watch whether ServiceNow follows into outcome pricing or defends the platform-subscription line.
Jul 22 (per IR page; one secondary calendar says Jul 28 — flag the conflict)
Alphabet Q2 earnings
First read on Gemini Enterprise Agent Platform traction and whether the marketplace-billed grounding unbundle shows up in cloud revenue mix commentary.
Jul 23
SAP Q2/H1 results
First test of the closed-gateway Joule strategy: if the mid-2026 third-party API restriction is confirmed and customers push back, it lands here — get the API-access roadmap in writing before renewal either way.
Jul 24
M365 Copilot OpenAI-subprocessor auto-enable (MC1422074)
The hard governance deadline: OpenAI-operated GPT-5.6 auto-enables as a subprocessor for eligible tenants unless admins set 'No users' — the data-processing chain changes under the DPA, and silence is consent.
Jul 29
Microsoft FY26 Q4 earnings
Copilot multi-model economics on display — the first print since the OpenAI-subprocessor change and the SharePoint Copilot Apps rollout, showing whether the multi-model suite is a margin or adoption story.
Q3 2026
SpaceX-Cursor $60B all-stock close expected; Grok 4.5 EU availability pending
The engineering vertical's structural event: post-close model availability on Cursor is the live vendor-risk test, and the EU availability question determines whether the combined entity can sell into European enterprises this year.
Changelog
- Scorecard refresh: support leadership reshuffled to Salesforce (Agentforce Help Agent GA) vs Sierra (Horizon); finance challenger updated to Sierra's loan-origination push; research challenger updated for Google's grounding unbundle; security leader reframed around injection-hardened stacks (GPT-Red); the 'other' row reframed from suite harnesses to data-cloud gateways after the week's capital events.
- W28's outcome-pricing pattern promoted from 'spreading' to 'default track': Salesforce's $2/resolution GA and Sierra's long-horizon outcome units close the experiment phase — the section's standing guidance shifts from 'watch for outcome SKUs' to 'negotiate outcome definitions, measurement rights, and audit access.'
- New coverage thread: MCP as the vertical-SaaS distribution channel (Clay's six-assistant GA vs Glean's index-centric counter), tracked alongside the data-cloud control-plane consolidation (Databricks term sheet earmark, Snowflake-AWS $6B).
- Editorial note: the SAP Joule closed-gateway report rests on a single secondary source and is labeled provisional pending SAP confirmation; the Databricks valuation is a signed term sheet, not closed capital.