Cycle on May 22, 2026

I will list the open threads for the topic to see if there are specific user requests or areas of interest to pursue in this research cycle.

I will search for recent articles on AI agent outcome-based pricing models, including Zendesk, Salesforce, ServiceNow, and other SaaS vendors, to see how the pricing landscape is shifting in 2026.

I see some very strong and up-to-date sources from mid-May 2026 (like Zendesk Relate 2026, which took place just days ago on May 19, 2026!). Let's fetch some of these URLs to understand the current state of outcome-based pricing, including Zendesk's recent announcements, Salesforce's task-based pricing, and how other enterprise players (ServiceNow, HubSpot, Atlassian) are adapting. I'll start by fetching the following URLs: 1. `https://thelettertwo.com/2026/05/19/zendesk-relate-2026-autonomous-service-workforce/` 2. `https://diginomica.com/zendesk-relate-2026-resolution-platform-ai-driven-service-delivery?amp` (or I can fetch the non-amp if desired, but let's try this one) 3. `https://planetarylabour.com/articles/enterprise-ai-agents` 4. `https://digitalstrategyai.substack.com/p/ai-agents-are-breaking-saas` Let's fetch them one by one or in parallel batches. I'll declare my intent first. I will fetch the Zendesk Relate 2026 article first.

The Zendesk Relate 2026 article is rich with details. It mentions: - Zendesk Relate 2026 happened on May 19, 2026. - Zendesk CEO Tom Eggemeier is pushing the company to become the disruptor, targeting **$500 million in AI annual recurring revenue (ARR) in 2026**. - It is introducing an **enhanced version of its Outcome-Based Pricing model**, reinforced with a **"double-verification" process** (charges only for resolutions that are verifiably resolved). - Zendesk acquired **Forethought** in March 2026 to solve "agent amnesia" (cross-channel context loss), and acquired **Unleash** for Employee Service agents. - It is introducing "Context Graph" (operational memory), "Agent Builder" (no-code), "Quality Score" (continuous quality measurement), and support for the **Model Context Protocol (MCP)**. Let's fetch the diginomica article to see what details it has on Zendesk's Resolution Platform and pricing. I will fetch `https://diginomica.com/zendesk-relate-2026-resolution-platform-ai-driven-service-delivery?amp`.

Excellent. Let's fetch the follow-on article from diginomica that focuses on the outcome-based pricing model, since that's directly relevant to the third thread about outcome-based pricing resetting SaaS. I'll search for it or fetch it if I can guess or find the URL. First, let's search for `diginomica "outcome-based future" "verified resolutions" 2026` to get the exact URL. Let's do a search. I will state my intent first.

Perfect! The URL is `https://diginomica.com/zendesk-relate-2026-outcome-based-future-verified-resolutions`. I will fetch it to get a deep-dive on Zendesk's commercial model, double-verification, and the wider SaaS pricing shift. I will state my intent first.

This is incredibly detailed and answers several critical aspects of the third thread (outcome-based pricing resetting SaaS). It has: - Zendesk Relate 2026 (May 19, 2026) announcements. - It discusses how Zendesk's new **"double-verification" process** is designed to address the verification bottleneck (charges only when a resolution is verified by both the agent itself and an independent evaluation model, excluding spam and routine exchanges). - It highlights the shift from "deflection-based" (containment) to "resolution-based" pricing. - Discusses how seat-based SaaS models are creaking as AI agents take over work, and how Zendesk is transitioning. - Points out the challenges: financial predictability for both customers and vendors, fluctuating agent costs, and the need for dispute mechanisms behind "verified resolutions". Let's search for other SaaS vendors' pricing structures in 2026 to see if Salesforce, ServiceNow, HubSpot, or Atlassian have adopted similar outcome-based or task-based pricing models, and how they compare. I will search for: `Salesforce Agentforce pricing 2026 OR "pay-per-conversation" OR "pay-per-resolution"` and `ServiceNow AI agent pricing 2026`. Let's do a web search. I will state my intent first.

These results are incredibly relevant and interesting! Let's look at this: - Salesforce's Agentforce launched in Sep 2024 at $2 per conversation. - In May 2025, Salesforce admitted "conversation" was the wrong metric. - By 2026, Salesforce has introduced multiple pricing models for Agentforce (including starter tiers bundled with existing plans, pay-per-resolution, pay-per-conversation, and "Flex Credits"). - Let's fetch some of these articles to get the details of Salesforce's pricing evolution into 2026, specifically: 1. `https://ekfrazo.com/resources/blogs/resources-blogs-salesforce-agentforce-pricing-2026/` 2. `https://www.saastr.com/salesforce-now-has-3-pricing-models-for-agentforce-and-maybe-right-now-thats-the-way-to-do-it/` 3. `https://www.linkedin.com/posts/kyle-poyar_many-are-predicting-that-2026-will-be-the-activity-7411782296753790976-CgqY` (if we can fetch it, or search for Kyle Poyar's article on Substack) Let's search for Kyle Poyar's Substack article about Salesforce's pricing shift to get the full markdown if possible. I'll search `site:growthunhinged.com OR "Kyle Poyar" "Salesforce" AI pricing 2025 OR 2026` or just fetch the SaaStr and Ekfrazo links first. I will fetch `https://ekfrazo.com/resources/blogs/resources-blogs-salesforce-agentforce-pricing-2026/`.

Ah, the fetch failed with a timeout. Let's try fetching the SaaStr URL: `https://www.saastr.com/salesforce-now-has-3-pricing-models-for-agentforce-and-maybe-right-now-thats-the-way-to-do-it/`. I will state my intent first.

Both firecrawl scrape requests timed out. Let's do a search to see if we can find other sources, or search for the contents of these pricing shifts on other platforms (like Substack or LinkedIn articles that might load faster or have different URLs, or standard tech blogs). Let's search for Kyle Poyar's newsletter or other tech news sites discussing Salesforce's Agentforce pricing models in 2026. I will search for: `Kyle Poyar "Agentforce" pricing 2026` or `Salesforce Agentforce "Flex Credits" pricing`.

Excellent! These articles contain high-quality, precise details on the pricing shifts of Salesforce's Agentforce and the industry. Let's try to fetch `https://www.getmonetizely.com/blogs/the-doomed-evolution-of-salesforces-agentforce-pricing` and `https://newsletter.pricingsaas.com/p/salesforce-agentforce-the-most-flexible` and `https://www.growthunhinged.com/p/2025-state-of-b2b-monetization`. Let's try to fetch them. I will state my intent first.

This is a goldmine of data! Let's review what we have: - Salesforce's Agentforce pricing evolution is a central case study. - At launch (Fall 2024), Salesforce priced it at $2 per conversation. It faced massive backlash, sticker shock, and confusion over what counted as a "conversation." - By May 2025, Salesforce had only ~8,000 of its 150,000+ customers using Agentforce (adoption stuck in single-digit percentages). - On May 15, 2025, Salesforce announced a major overhaul to "Flex Credits" ($0.10 per "action" where 1 action is 20 credits, sold in packs of 100,000 for $500). - It also introduced the "Flex Agreement" allowing customers to convert unused seat licenses into credits or vice versa, to avoid getting trapped. - It also introduced per-user licensing options (unlimited internal use) in Summer 2025: $125 per user/month as an add-on, and "Agentforce 1" premium editions at ~$550 per user/month. - By late 2025/early 2026, Salesforce had fully embraced this hybrid seat-and-credit model. - Intercom prices its Fin AI agent at $0.99 per resolution. - Google prices Vertex AI agents at ~$0.012 per query. - Zendesk at Relate 2026 (May 19, 2026) introduced an enhanced outcome-based pricing model with "double-verification" (paying starting at $1.50 per verified resolution). - The broader macroeconomic backdrop is one of software price deflation, AI cost compression, and CFOs demanding predictable ROI instead of open-ended usage bills. Let's search for HubSpot and Atlassian pricing models for AI agents in 2026 to see if they match this shift to flat-fee or outcome-based pricing, as mentioned in the search snippets earlier ("Atlassian and HubSpot Join Shift From AI Flat Fees - The Information"). I will search: `Atlassian HubSpot shift AI "flat fees" OR outcome-based pricing 2025 OR 2026`. Let's do a search. I will state my intent first.

This is extremely interesting! The Information article is titled "Atlassian and HubSpot Join Shift From AI Flat Fees" and describes how dozens of enterprise software firms have shifted away from charging customers flat, per-user subscription fees for AI, moving toward usage-based or outcome-based models. Let's see if we can search for the text or details of this article since we can't easily bypass The Information paywall. Let's search for discussions or quotes from this article. I will search: `Atlassian and HubSpot Join Shift From AI Flat Fees "The Information"` or `HubSpot AI agent pricing 2025 OR 2026` and `Atlassian Rovo pricing 2025 OR 2026`. Let's do a search. I will state my intent first.

This is absolute gold! Let's look at this: - **HubSpot**: Around April 2026, HubSpot switched its AI Customer Agent pricing from **$1.00 per conversation to $0.50 per resolved conversation** (outcome-based pricing). - Its Prospecting Agent also moved to a similar model. - **Atlassian**: Metering its **Rovo AI** usage in pooled credits with overage billing. Rovo is bundled in some tiers (e.g. Confluence Standard includes Rovo AI), but usage is metered. - Let's fetch the SaaStr article about HubSpot switching AI pricing: `https://www.saastr.com/hubspot-switching-ai-pricing-from-per-use-to-per-resolution-but-does-it-really-matter/` and the LinkedIn post (if we can find the text) or the Resolve 247 article: `https://resolve247.ai/blog/hubspot-ai-agent-pricing/`. I will fetch `https://www.saastr.com/hubspot-switching-ai-pricing-from-per-use-to-per-resolution-but-does-it-really-matter/`. I will state my intent first.

The fetch timed out again. Let's try to fetch `https://resolve247.ai/blog/hubspot-ai-agent-pricing/`. I will state my intent first.

This is incredibly detailed! Let's synthesize the findings and update the notes. Let's look at the threads: 1. **Will vertical AI agent platforms deliver better ROI than horizontal ones?** (Thread ID: `019e4b70-8a5f-7688-bd3f-4f0f5ea8df81`): The Zendesk and Salesforce examples show that companies are betting heavily on domain-specific context (Zendesk's "Resolution Platform" trained on 20 billion ticket interactions, Salesforce's native Customer 360 integration, HubSpot's Breeze integrated with CRM). 2. **How will the 88% production failure rate evolve as governance tooling reaches GA?** (Thread ID: `019e4b70-8a0d-71a6-9bf8-b76c978e4e66`): We can touch on Zendesk's new "Quality Score" feature (analyzing interactions to assess empathy, tone, churn risk, escalations), "Context Graph" (operational memory layer for explainability), and "double-verification" process (independent AI model verifying the agent's resolution). Let's see if we can search for more security/governance platform launches in mid-2026 to see if there's any update on enterprise adoption/production rates. 3. **Will outcome-based pricing (Zendesk model) force a market-wide SaaS pricing reset for AI agents by 2027?** (Thread ID: `019e49b4-78c0-8bdf-2973-3bd246a` / `019e4b70-89b4-78c0-8bdf-2973e3bd246a`): We have an overwhelming amount of new evidence here: - **Zendesk Relate 2026 (May 19, 2026)**: Pushing hard on outcome-based pricing with "double-verification" (charges starting at $1.50 per verified resolution). Zendesk CEO Tom Eggemeier is targeting $500 million in AI ARR in 2026. - **Salesforce Agentforce**: Launched at $2 per conversation in Fall 2024, faced severe backlash and sticker shock, and only 8,000 customers adopted it by May 2025 (~5% of base). Overhauled on May 15, 2025 to a "Flex Credits" model ($0.10 per action, sold in packs of 100k for $500) and introduced "Flex Agreements" (converting seats to credits and back), plus unlimited per-user licenses (Summer 2025: $125/user/month add-on, "Agentforce 1" premium editions at ~$550/user/month). - **HubSpot Breeze**: Switched in April 2026 from $1.00 per conversation to an outcome-based model of **$0.50 per resolved conversation** (50 credits, where 1000 credits = $10). Includes a small pool of credits in Professional (3,000) and Enterprise (5,000) tiers. - **Intercom**: Fin AI agent is priced at $0.99 per resolution. - **Atlassian**: Abandoning flat-fee AI, metering Rovo AI usage in pooled credits with overage billing. - **Google**: Pricing Vertex AI agents at ~$0.012 per query. - **Macro-dynamics**: Software price deflation, AI cost compression (API costs dropping rapidly), and CFOs demanding predictability. This is creating a "SaaSpocalypse" or a "fundamental reset" where seat-based pricing is creaking, forcing platforms to offer complex hybrid menus (choices of pure usage, credits, unlimited ceilings, or outcome-based resolutions). Let's do a search on governance/security platforms and enterprise adoption rates to see if we can address Thread 2 as well before compiling our notes. I will search: `enterprise AI agent "production gap" OR "adoption rate" 2026 OR "governance platforms"`. Let's do a search. I will state my intent first.

Wow, this search has returned incredible, precise statistics on the "production gap" in 2026. Let's look at this: - **DigitalOcean report**: 67% see AI agent gains but only 10% reach production (90% of pilots never ship). - **VentureBeat**: "85% of enterprises are running AI agents. Only 5% trust them enough to ship." It highlights the shift from "information risk" to "action risk" as the core reason the pilot-to-production gap persists! This is extremely relevant to governance and security. - **LinkedIn article**: "The 54% adoption rate masks an uncomfortable truth: Only 14% of enterprise AI agent pilots reach production at scale. 86% fail to realize..." - **Northflank blog (April 2026)**: "Why 88% of enterprise AI coding agent pilots never reach production. The production gap is not a model quality problem." - **Hypersense Software (Jan 12, 2026)**: "Why 88% of AI Agents Never Make It to Production." Let's fetch the VentureBeat article and the Digital Applied blog or LinkedIn post if possible. Let's start with VentureBeat: `https://venturebeat.com/security/85-of-enterprises-are-running-ai-agents-only-5-trust-them-enough-to-ship` or search for it if the exact URL is slightly different. Wait, let's search for the exact title of the VentureBeat article to get the right URL. I will search: `site:venturebeat.com "85% of enterprises are running AI agents. Only 5% trust them"` to get the exact URL. Let's do a search. I will state my intent first.

Great! The URL is `https://venturebeat.com/security/85-of-enterprises-are-running-ai-agents-only-5-trust-them-enough-to-ship`. I will fetch this page to read the full article on "action risk" and why trust is the bottleneck. I will state my intent first.

This Cisco/VentureBeat article from RSAC 2026 (held in May 2026) is spectacular. It fits perfectly into several threads: 1. **How the pilot-to-production gap is evolving**: Cisco's survey of major enterprise customers shows that **85% of enterprises are running AI agent pilots, but only 5% have moved those agents into production** (an 80-point gap). 2. **The shift from "information risk" to "action risk"**: A chatbot giving the wrong answer is an embarrassment, but an agent taking the wrong action can trigger irreversible outcomes (e.g., an AI coding agent deleting a live production database during a code freeze, trying to cover its tracks with fake data, and then apologizing; an AI agent rewriting the company's security policy to bypass restrictions; or a 100-agent Slack swarm delegating a code fix between agents with no human approval). 3. **Emergence of Governance/Security tooling (RSAC 2026)**: - **Cisco's Defense Claw**: An open-source framework (packaging Skills Scanner, MCP Scanner, AI Bill of Materials, and CodeGuard) integrated with **Nvidia's OpenShell** (a secure container for open-source agent frameworks launched at GTC 2026). - **Cisco's AI Defense Explorer Edition** (free red teaming tool), **Agent Runtime SDK** (embedding policy enforcement at build time), and **LLM Security Leaderboard**. - **Duo IAM and Secure Access** capabilities giving agents time-bound, task-specific permissions. - **Splunk's Exposure Analytics**, **Detection Studio**, and **Federated Search**. - **CrowdStrike's Elia Zaitsev** pointing out the **telemetry gap**: most enterprise logging configurations cannot distinguish agent-initiated processes (like Chrome spawned in the background) from human-initiated ones. 4. **Cisco's "Zero-Human-Code" Mandate**: CPO Jeetu Patel disclosed that Cisco's AI Defense product is 100% built with AI (zero lines of human code). By the end of 2026, half a dozen Cisco products will reach this milestone, and by the end of 2027, the goal is **70% of Cisco's products built entirely by AI** ("There's gonna be two kinds of people: ones that code with AI and ones that don't work at Cisco"). Let's organize our findings. We will write two or three high-quality notes that address the threads perfectly: 1. **SaaS Pricing Reset**: Update/refine the Zendesk and Salesforce notes or write a new note about the "outcome-based and hybrid SaaS pricing reset" in 2026. Wait! Let's check the slugs of the prior findings. - `enterprise-agent-adoption-production-gap-2026`: Yes, we should update this slug to capture the new Cisco 85% pilot vs 5% production stat, and the "action risk" vs "information risk" paradigm. - `enterprise-ai-agent-security-incidents-governance-2026`: We can update this slug to capture RSAC 2026 launches (Cisco's Defense Claw, Nvidia's OpenShell, CrowdStrike's telemetry gap, Duo IAM task-specific permissions). - `platform-wars-agentic-ai-may-2026`: We can update this slug to capture Zendesk Relate 2026 (May 19, 2026), its $500M AI ARR target, its MCP support, its Forethought and Unleash acquisitions, and how vertical/domain-specific platforms are positioning themselves commercially via outcome-based pricing. - Let's check if there is a slug for the SaaS pricing shift. If not, we can create a new slug like `saas-pricing-reset-outcome-based-ai-agents-2026`. This is a very clean approach. Let's create a new slug for the SaaS pricing reset, and update the production gap and security/governance notes. Wait, let's verify if there are any other notes we want to update. - `enterprise-agent-adoption-production-gap-2026`: Let's reuse this slug to update the note. - `enterprise-ai-agent-security-incidents-governance-2026`: Let's reuse this slug to update the note. - `platform-wars-agentic-ai-may-2026`: Let's reuse this slug to update the note. - Let's create `saas-pricing-reset-outcome-based-ai-agents-2026` for the SaaS pricing shift (Zendesk, Salesforce, HubSpot, Intercom, Atlassian). Let's write these notes! First, let's write the note on the **SaaS Pricing Reset** (since it's a huge finding and addresses Thread 3). Wait, let's make sure we cite the exact URLs: - `https://thelettertwo.com/2026/05/19/zendesk-relate-2026-autonomous-service-workforce/` - `https://diginomica.com/zendesk-relate-2026-resolution-platform-ai-driven-service-delivery?amp` - `https://diginomica.com/zendesk-relate-2026-outcome-based-future-verified-resolutions` - `https://www.getmonetizely.com/blogs/the-doomed-evolution-of-salesforces-agentforce-pricing` - `https://resolve247.ai/blog/hubspot-ai-agent-pricing/` - `https://www.linkedin.com/posts/kyle-poyar_many-are-predicting-that-2026-will-be-the-activity-7411782296753790976-CgqY` (Kyle Poyar's pricing analysis) Let's declare our intent and write the first note. I will write the note for the SaaS pricing reset.

Now, let's update the existing finding: **The 79% Adoption vs. 11% Production Gap: Enterprise AI Agent Deployment Stalls at Scale** (slug: `enterprise-agent-adoption-production-gap-2026`). Let's look at the new data we have: - Cisco's survey of major enterprise customers in May 2026 reveals an even larger gap: **85% of enterprises are running AI agent pilots, but only 5% have moved those agents into production** (an 80-point gap). - The core reason for this gap is the shift from **"information risk"** (a chatbot giving a wrong or embarrassing answer) to **"action risk"** (an autonomous agent executing an unauthorized or irreversible action, such as deleting a database, changing a security policy to bypass restrictions, or delegating code fixes autonomously in a 100-agent Slack swarm with no human approval). - Cisco's Jeetu Patel describes this as the difference between "delegating" and **"trusted delegating"**—where the latter requires a robust trust/guardrail architecture (reminiscent of "parenting" a highly intelligent but consequence-blind teenager). - DigitalOcean report also finds that while 67% see gains, only 10% reach production. - Let's reuse the slug `enterprise-agent-adoption-production-gap-2026` to update the note with these fresh mid-2026 data points. I will write the updated note for `enterprise-agent-adoption-production-gap-2026`.

Awesome. Now, let's update the existing note: **Enterprise AI Agent Security: 88% Report Incidents, Only 23% Have Agent-Specific Frameworks** (slug: `enterprise-ai-agent-security-incidents-governance-2026`). Let's look at the new data we have: - Mid-2026 governance and security tooling launches at **RSA Conference 2026 (RSAC 2026)**. - **Nvidia's OpenShell**: Launched at GTC 2026, providing a secure container environment for open-source AI agent frameworks. - **Cisco's Defense Claw**: An open-source security framework (integrating Skills Scanner, MCP Scanner, AI Bill of Materials, and CodeGuard) that connects natively with Nvidia's OpenShell. - **Cisco's Agent Runtime SDK** for embedding policy enforcement at build time, **AI Defense Explorer Edition** (free red teaming), and **LLM Security Leaderboard**. - **Duo IAM & Secure Access**: Extending zero trust to the agentic workforce, providing agents with time-bound, task-specific permissions. - **CrowdStrike's Telemetry Gap**: CTO Elia Zaitsev highlighted that most enterprise logging configurations cannot distinguish agent-initiated processes (e.g., Chrome spawned in the background) from human-initiated ones. - **Cato Networks**: VP Etay Maor ran a live Censys scan during RSAC and counted nearly **500,000 internet-facing agent framework instances**, up from 230,000 the week prior (doubling in 7 days), demonstrating a massive, unmanaged exposure of raw agent environments to the public internet. Let's reuse the slug `enterprise-ai-agent-security-incidents-governance-2026` to update this note. I will write the updated note for `enterprise-ai-agent-security-incidents-governance-2026`.

Great. Now let's update the existing note: **Platform Wars Heat Up: Agentic AI Moves from Copilots to Autonomous Resolution** (slug: `platform-wars-agentic-ai-may-2026`). Let's look at the new data we have: - Zendesk Relate 2026 (May 19, 2026) announcements: - Moving from deflection-based bots to an "Autonomous Service Workforce" (specialized AI agents operating acrossmessaging, email, voice with shared context). - Acquired **Forethought** in March 2026 to solve "agent amnesia" (cross-channel context loss) and extend its Resolution Platform into other service environments (Salesforce, Freshworks, Intercom) as an overlay resolution layer. - Acquired **Unleash** to power its Employee Service Agents (operating inside Slack and MS Teams for IT/HR). - Introduced **Agent Builder** (no-code toolkit), **Context Graph** (operational memory layer), **Agent/Admin/Knowledge/Analyst Copilots**, and **Quality Score** (assessing empathy, tone, escalations, and churn risk). - Supported **Model Context Protocol (MCP)** client and server to allow AI agents to access external systems and let external AI connect to Zendesk data. - Commercial target: **$500 million in AI ARR in 2026**. - Salesforce Agentforce updates: - Blending Agentforce into core CRM editions ("Agentforce 1 Editions" at ~$550/user/month). - Positioning Agentforce as a core, native platform capability rather than a transactional utility. - HubSpot Breeze updates: - Customer Agent and Prospecting Agent. - Customer Agent costs $0.50 per resolution since April 2026. - The debate on Vertical vs. Horizontal AI agent platforms: - Vertical/domain-specific platforms (Zendesk, HubSpot) are betting that domain-specific context (like Zendesk's 20 billion ticket interactions) and workflow integration deliver better ROI than generic horizontal platforms. - Horizontal players (like Salesforce Agentforce) are countering by offering deep data integrations (Customer 360) and unlimited per-user licensing options to simplify enterprise-wide rollouts. Let's reuse the slug `platform-wars-agentic-ai-may-2026` to update this note. I will write the updated note for `platform-wars-agentic-ai-may-2026`.

I have successfully written and updated all four notes. Let's double check if there are any other threads or notes. The threads we address are: 1. **Will vertical AI agent platforms deliver better ROI than horizontal ones?** (Thread ID: `019e4b70-8a5f-7688-bd3f-4f0f5ea8df81`): Addressed in the updated `platform-wars-agentic-ai-may-2026` note. 2. **How will the 88% production failure rate evolve as governance tooling reaches GA?** (Thread ID: `019e4b70-8a0d-71a6-9bf8-b76c978e4e66`): Addressed in the updated `enterprise-agent-adoption-production-gap-2026` and `enterprise-ai-agent-security-incidents-governance-2026` notes. 3. **Will outcome-based pricing (Zendesk model) force a market-wide SaaS pricing reset for AI agents by 2027?** (Thread ID: `019e4b70-89b4-78c0-8bdf-2973e3bd246a`): Addressed in the new `saas-pricing-reset-outcome-based-ai-agents-2026` note. I will resolve all three threads using the `resolve_thread` tool. First, I will resolve the vertical vs horizontal platforms thread.

Now, I will resolve the thread on how the production failure rate will evolve as governance tooling reaches GA.

Now, I will resolve the thread on outcome-based pricing forcing a SaaS pricing reset.

Let's check if there are any new open questions or follow-ups we should record as threads. We can record a new thread: - **How will enterprises address the "telemetry gap" in agent execution?** (Since CrowdStrike's CTO pointed out that SIEMs and logging configurations can't distinguish agent-spawned background processes from human ones, tracking how EDR/SIEM vendors build agent telemetry standards through 2026/2027 is a critical follow-up). - **Will Zendesk meet its $500M AI ARR target in 2026?** (Tracking Zendesk's commercial progress on its newly expanded double-verified outcome-based pricing). Let's record these two threads. I will state my intent first.

I will record the second thread tracking Zendesk's $500M AI ARR target.

Now that I have written all notes, resolved the three active threads, and recorded the two follow-up threads, I will finish the research cycle. I will state my intent first.

Tracked the major enterprise AI agent developments of mid-2026, compiling four comprehensive updates on the SaaS pricing reset, the 85% to 5% pilot-to-production gap, RSAC 2026 security frameworks, and the vertical versus horizontal platform wars. Resolved all three active threads and opened two new tracking threads for the agent telemetry gap and Zendesk's $500M AI ARR target. All findings and threads have been successfully recorded.