Five releases on July 28, 2026 addressed the same layer of marketing work: what happens after an AI system produces output. Knak surveyed more than 300 enterprise marketing leaders and found that 70% have deployed AI in production while 88% say the output still requires moderate to substantial human editing before use. In the same population, 85% missed at least one planned campaign launch date in the past twelve months, and the causes they named were operational: approvals and sign-off (47%), design and creative production (38%), and cross-team coordination (36%). Algolia responded to the same category of problem on the commerce side by adding guardrails, request limits, and token ceilings to its agent product. Cyabra shipped an agent that returns a verdict with an evidence trail instead of a dashboard. StackAdapt moved planning, forecasting, and execution into a conversational workspace. WPP Media contracted Kily to let agents manage listings, pricing, and advertising across Indian marketplaces.
The Knak figures give marketing leaders a cost line to work with. Sixty percent of surveyed teams involve four or more people in producing a single email, 69% require two to three rounds of revision, and 54% move that work across three to five separate tools. Knak prices the internal labor at roughly $300 per send using Bureau of Labor Statistics pay data. Half of these teams still run approvals through email, Slack, or Teams threads. AI use concentrates where the work is easiest to hand off, with 64% using it for first-draft copy and 56% for image generation, while only 25% apply it to building or coding the emails and landing pages that consume the most production time. A CMO who reads high AI adoption rates as evidence of recovered capacity is measuring drafts rather than launches.
Algolia’s release names the second cost that appears in year two of agent deployment. The Agent Studio update lets retailers define disallowed content categories, apply them to shopper inputs and agent outputs separately, and configure fallback messages when content gets blocked. It also adds global and per-IP request limits, maximum tokens per response, conversation-depth limits, maximum steps per completion for tool-using agents, and approved-domain restrictions. IDC’s Heather Hershey framed these as the conditions for moving from pilots to production. The features describe a specific failure mode: agents that answer shopper questions accurately and generate unbounded inference spend while doing it. Retailers evaluating agentic commerce protocols now need a token budget alongside a conversion target, and the controls that enforce it belong in the procurement conversation rather than the post-launch one.
Cyabra and StackAdapt both moved interpretation work from the operator to the system. Cyabra’s Coordinated Activity Detection runs automatically on every scan, establishes a baseline suspicion level from account authenticity, screens clusters for concentrated inauthentic behavior, examines posting time and profile signals, and returns one of four states with a confidence score and a downloadable evidentiary report, typically in under 30 seconds. Every signal ties back to the data point that triggered it. StackAdapt’s Ivy Studio applies the same pattern to media, letting marketers state an outcome and having agents surface the supporting context and recommended next steps. StackAdapt reports that its own employees run more than 15,000 AI-assisted workflows per week and that roughly 70% of Ivy Studio usage concentrates in its top 20 workflows. Both products keep the approval with the operator, which puts human-in-the-loop design at the center of the buying decision rather than at the edge of it. The question for a CMO is who signs off on an agent’s verdict and what happens when that verdict turns out to be wrong in front of a client or a regulator.
The WPP Media and Kily partnership shows where the same logic lands in commerce operations. Kily builds agents that manage listing, pricing, and advertising decisions across marketplaces and quick commerce, and WPP will connect that capability to WPP Open and an Integrated Commerce Planner for its India Commerce Practice. Handing pricing and listing decisions to a system moves margin control into software, which makes agent behavior a P&L variable rather than a productivity feature. Brands operating under a business-to-agent-to-consumer model already accept that machine intermediaries shape discovery. Extending that to pricing changes the review cadence, the escalation path, and the definition of an acceptable error. The same governance question follows any brand implementing the Agentic Commerce Protocol (ACP) endpoints that decide which agent-initiated transactions get accepted.
The consistent trade-off across all five releases sits between speed and accountability. Each product removes a step a person used to perform and replaces it with a control surface, an audit trail, or a limit. Buying the control surface without staffing the review it assumes reproduces the Knak finding at a larger scale: more output, the same launch dates. Marketing organizations that have not defined who approves agent decisions, what the token budget is, and how errors get escalated will pay for governance twice, once in software and once in the incident that follows. An AI governance board gives those decisions a standing owner, and the count of agents already running determines how much that structure costs.
Strategic priorities for marketing leaders
- Measure AI adoption by launches shipped on time, not by drafts produced. Knak’s 70% deployment rate against an 88% heavy-editing rate shows the two numbers move independently.
- Audit the production layer before buying another generation tool. Count the people, tools, and revision rounds required per email, and price the internal labor.
- Add token and request limits to agent procurement criteria. Ask vendors for per-agent cost controls, input and output guardrails, and approved-domain restrictions before signing.
- Name the person who approves agent verdicts and pricing changes. Document the escalation path and the acceptable error rate for each agent in production.
- Move approvals off email and chat threads. Half of surveyed enterprise teams still route sign-off through Slack or Teams, which is where the 47% approval bottleneck originates.
- Treat agent-managed pricing and listings as a margin decision. Assign finance a review cadence when commerce agents gain write access to price and inventory.
Here’s The News:
Knak Report Finds 85% of Marketing Teams Missed a Campaign Launch Date in the Past Year
Knak, the marketing production platform for enterprise teams, released Marketing Production in the Age of AI, based on a survey of more than 300 enterprise marketing leaders. The report finds that 85% of teams missed at least one planned campaign launch date in the past twelve months due to workflow constraints, and one in ten missed launches more than five times per year. Respondents attributed delays to approvals and sign-off (47%), design and creative production (38%), and cross-team coordination (36%). Producing a single email involves four or more people at 60% of surveyed companies, requires two to three revision rounds at 69%, and moves through three to five separate tools at 54%, which Knak prices at more than $300 in internal labor per send using BLS pay data. Seventy percent of teams have deployed AI in production, and 88% say the output still needs moderate to substantial human editing. AI use concentrates on first-draft copy (64%) and image generation (56%), while 25% apply it to building or coding emails and landing pages. Eighty-eight percent of teams produce assets for email, and 52% plan to add or significantly increase email spend in the next twelve months. Knak CMO Jennifer Delevante described the constraint as the production layer between an approved idea and a launched campaign. Knak counts Google, Amazon, Uber, Meta, and OpenAI among its customers. Read the full press release (Published: July 28, 2026 | Source: PR Newswire)
StackAdapt Introduces Ivy Studio, an AI-First Advertising Hub
StackAdapt announced Ivy Studio, a workspace inside its platform that combines planning, forecasting, analysis, optimization, and execution. Marketers describe an intended outcome in natural language, and AI agents grounded in StackAdapt’s campaign, audience, creative, and performance data surface supporting context, identify opportunities, and recommend next steps while the marketer retains approval. Ivy Studio runs on Ivy, StackAdapt’s AI engine, and supports programmable skills that let teams pull routine metric reports across multiple campaigns into unified outputs. StackAdapt reports that its own employees now run more than 15,000 AI-assisted workflows per week across campaign management, reporting, troubleshooting, and optimization, and that nearly 70% of Ivy Studio usage concentrates in the top 20 workflows. Co-founder and CEO Vitaly Pecherskiy said the product lets teams complete in minutes what previously took hours or coordination across multiple teams. StackAdapt is also using OpenAI language models to build advertising-specific capabilities that extend Ivy Studio. Read the full press release (Published: July 28, 2026 | Source: ExchangeWire PressBox)
Algolia Adds Guardrails and Cost Controls to Agent Studio
Algolia announced enhancements to Agent Studio, its application for building AI agents and onsite agentic commerce experiences. The update adds custom guardrails that let teams define disallowed content categories, apply them to shopper inputs, agent outputs, or both, and configure fallback messages when content gets blocked. Input guardrails evaluate shopper messages before they reach the model, and output guardrails review generated responses before display. Retailers can configure global request limits, per-IP limits, maximum tokens per response, conversation-depth limits, maximum steps per completion for tool-using agents, and approved domains. The release also brings AI mode into the search bar with contextual prompt suggestions that open a chat experience with the selected prompt populated, and extends both capabilities to mobile. Heather Hershey, Senior Research Director for Digital Commerce and Agentic Commerce Strategies at IDC, said retailers need guardrails that make agents safe, governed, and economically manageable in live shopping environments. Algolia CEO Stephen Lynch framed the update around governance and control across the shopping journey. Guardrails and cost controls are available now to Algolia customers. Algolia reports 1.75 trillion searches a year across more than 18,000 businesses. Read the full press release (Published: July 28, 2026 | Source: Algolia via BigDATAwire)
Cyabra Launches an AI Agent That Returns a Verdict on Coordinated Online Activity
Cyabra, Inc. (Nasdaq: CYAB) launched Coordinated Activity Detection, an AI agent that automates the company’s analyst methodology and returns a single evidence-backed determination on every scan. The agent establishes a baseline suspicion level from account authenticity, identifies peak activity windows, screens every cluster in a scan for concentrated inauthentic behavior, and examines posting time, content, and profile signals for evidence of manipulation. When evidence supports it, the agent traces which narratives are spreading, who introduced them, and how they propagate. Each phase streams findings as they are generated, and every signal ties to the specific data point that triggered it. The agent returns one of four states, Confirmed Coordination, Likely Coordination, Insufficient Evidence, or No Coordination Detected, with a confidence score and a downloadable evidentiary report, typically in under 30 seconds. The capability runs automatically on every scan and surfaces within the platform’s Authenticity and Narrative tabs. Cyabra targets enterprise brand and communications teams, public sector and government teams, and its own analysts. CEO Dan Brahmy said the product gives customers a direct answer in place of another dashboard to interpret. CMO David Low said the agent assembles an explanation teams can share and defend. The launch follows Cyabra’s recognition as a Market Shaper in the inaugural June 2026 Gartner Emerging Market Quadrant for Narrative Intelligence, Startup Vendors. Read the full press release (Published: July 28, 2026 | Source: GlobeNewswire)
WPP Media Partners with Kily to Run Agentic Commerce Operations in India
WPP Media announced a technology partnership with agentic AI startup Kily to expand its commerce offering across India’s marketplace and quick commerce ecosystem. Kily builds AI agents that autonomously manage listing, pricing, and advertising decisions across multiple commerce platforms, unifying signals across platforms, categories, and consumer touchpoints to support both intelligence and execution. The partnership connects Kily’s technology to WPP Open, WPP’s agentic marketing platform, and introduces an Integrated Commerce Planner that combines data across commerce platforms so brands can plan, activate, and measure investment in one system. Kily was founded by former Flipkart and Affle executives Sankalp Mehrotra and Anurag Singh. Mehrotra said India’s fragmented multi-marketplace landscape makes autonomous commerce management necessary for judgment and decision-making at scale. Ashwin Padmanabhan, COO of WPP Media South Asia, said the Integrated Commerce Planner connects planning, activation, and measurement so brands can align demand, inventory, and media. The partnership advances WPP Media’s stated focus on autonomous advertising. Read the full article (Published: July 28, 2026 | Source: WPP Media via MediaNews4U)
Progress Software: Introduces AI Agents Built into CMS Workflows to Reduce Content Bottlenecks
Progress Software (Nasdaq: PRGS) announced agentic AI capabilities embedded directly into its Sitefinity Generative CMS platform. Unlike AI assistants that operate outside the publishing process, Sitefinity AI agents operate within content workflows — executing content analysis, optimization, editorial review, and deployment tasks as part of the content lifecycle itself. The distinction matters: agents reduce bottlenecks and shorten production cycles by eliminating the handoff between AI tools and the CMS, rather than adding another disconnected tool to the stack. EVP Loren Jarrett described the shift as enterprise AI entering its next phase — from AI-assisted engagement to AI-powered digital operations. The announcement is significant for marketing operations leaders because it represents the architectural model that addresses the production bottleneck identified in the Knak report: AI embedded in the workflow, not layered on top of it. Progress Software serves hundreds of thousands of businesses and tens of millions of professionals worldwide. (Published: July 28, 2026 | Source: Progress Software / Globe Newswire | Read the full press release)
