Yesterday’s announcements from RTB House, Upwork, Viral Nation, accessiBe and Fusion Connect each named a specific point where a person intervenes in an AI-run process. RTB House, the Warsaw-founded performance marketing company, released its global consumer study “Who’s Buying? Consumer Trust in the Age of Agentic AI” on August 10, 2026, based on 1,840 respondents surveyed across the United States and a combined non-US group spanning the United Kingdom, France and Japan during June and July 2026 in partnership with survey sampling platform Cint. The study found that approval before checkout ranks as the single most requested safeguard globally among consumers evaluating AI-executed purchases.
That finding sets the frame for the other four releases. Each vendor shipped a product decision about where a human sits in an automated sequence, and each attached a price to that position.
RTB House reported that leading AI tools including Google AI Overviews and ChatGPT now score higher on shopping trustworthiness than TikTok, Instagram and Facebook among US and international shoppers, and higher than newspapers and major media outlets. RTB House also found that trust splits unevenly across AI platforms, with Google AI Overviews and ChatGPT leading while Claude and Grok trail in both the US and globally. The behavioral consequence runs against the compression story vendors have been selling. RTB House found that 42% of US consumers and 33% of non-US consumers say AI tools increase the time they need to reach a final purchase decision by surfacing more brands, products and alternatives, with the effect strongest among US Gen Z at 48%. On discovery, RTB House reported that 59% of US consumers and 52% of non-US consumers say AI surfaces brands they did not previously know, with US Gen X at 61% and Baby Boomers at 62% among the most receptive groups.
Financial delegation in the RTB House data arrives with conditions attached. RTB House found that 42% of US Millennials are comfortable giving an AI agent a budget of up to $250 to make a purchase on their behalf when a seven-day return window is guaranteed, compared with 27% of non-US Millennials. Removing the return protection drops those figures to 34% of US Millennials and 20% of non-US Millennials. Digital Transactions, reporting on the study on August 10, 2026, put the share of consumers globally who want human approval before an AI completes a purchase at 35%.
Work through what that means for a specific team. A consumer electronics brand with a $180 average order value sits below the $250 delegation ceiling, so its category qualifies for agent-executed purchase among the most willing cohort. The seven-day return window carries an eight-percentage-point swing among US Millennials, which a merchandising lead can price directly: the incremental return-processing cost against the incremental conversion from a guaranteed window. A furniture brand with a $1,400 average order value has no such calculation available, because its price point sits far outside the range consumers currently delegate. The RTB House data gives a category-level test. Marketing leaders in high-ticket categories should read the extended consideration window as the operative finding for their planning.
Upwork Inc. (NASDAQ: UPWK) built the approval point into the product. Upwork announced the Upwork MCP server on August 10, 2026, connecting its marketplace to Claude, ChatGPT, Cursor and any product compatible with the Model Context Protocol (MCP). A client’s agent can post a job, surface talent, prepare offers and summarize proposals as they arrive, and a freelancer’s agent can flag matching opportunities, draft proposals, surface client messages and submit completed milestone work. Peter Sanborn, chief business officer at Upwork, stated that the company began seeing AI agents attempting to log into Upwork on their users’ behalf during 2026. Rohit Singh, founder of Populosof and an Upwork client quoted in the release, described the constraint that makes the integration usable: every action that changes anything is previewed and requires his explicit confirmation. That design places human-in-the-loop (HITL) at the write boundary, and it is the load-bearing detail in the announcement. Upwork stated that every connection runs through an authenticated Upwork account and carries the platform’s existing identity verification, secure payments, escrow and dispute protections, and that the MCP server is available to all Upwork clients and freelancers at no additional cost. Upwork reported it has facilitated more than $30 billion in total transactions and services since its founding.
Marketing operations leaders should read the Upwork release as a change in where contractor spend originates. A job posted from inside a creative director’s AI session bypasses the intake form, the procurement queue and the agency-of-record review. Someone in finance reconciles that spend afterward. Teams that route freelance creative through a managed vendor list now need a policy covering agent-initiated hiring, because the confirmation step Upwork built protects the individual user from an unintended action and does nothing about whether that user held budget authority in the first place.
accessiBe applied the same layering logic to the front end. accessiBe, an end-to-end web accessibility platform, announced its AI accessibility assistant on August 10, 2026, a conversational interface built into its accessWidget product. Site visitors describe what they need in ordinary language, using phrasing such as “the text is too small” or “take me to the footer,” and the assistant adjusts the page, answers questions about page content or navigates the visitor to a destination. Robert Lopez, CEO of accessiBe, stated that the accessibility adjustments already existed and that conversational AI delivers them in a more natural way. accessiBe kept the existing accessWidget settings menu fully available alongside the assistant, and the company stated the assistant is live for accessWidget customers on Growth plans and higher with a phased rollout to additional customers planned. accessiBe developed the assistant with input from its Inclusive Product Advisory Board and accessLabs, groups composed of people with disabilities.
The accessiBe decision to retain the manual settings menu carries more operational weight than the conversational layer itself. A visitor who knows which control to change reaches it in one click. A visitor who does not know accessibility terminology describes the problem instead. Web teams evaluating conversational overlays should require the same dual path, because replacing a deterministic control surface with a probabilistic one removes a guarantee that a compliance audit depends on.
Fusion Connect addressed the cost side of the same question. Fusion Connect, a US provider of managed communications, connectivity and cloud solutions, announced general availability of its AI-powered Contact Center as a Service platform on August 10, 2026 alongside a flat-fee pricing model. Fusion Connect stated that most AI solutions carry usage-based pricing keyed to token consumption rate cards, and that organizations struggle to estimate costs before deployment because pricing varies with which capabilities run and how often. Under the announced model, core capabilities including Agent AI Assistance and CX AI Analytics carry a monthly per-seat price, and high-volume automation services including Autopilot Agentic AI and AI-powered call routing carry monthly packages priced against anticipated call volumes. John Nee, Chief Marketing Officer of Fusion Connect, stated that organizations should not need a data scientist or a financial model to understand what an AI investment will cost. Frederic Dickey, Senior Vice President of Product Management, stated that the pricing model removes uncertainty from broader AI adoption. The platform’s four AI capabilities cover real-time agent guidance, natural-language intent routing that replaces menu-driven interactive voice response, autonomous handling of routine interactions, and automated conversation analysis. Availability is immediate.
A Chief Marketing Officer (CMO) who owns a contact center line item can price this trade directly. Per-seat pricing caps exposure and charges for capacity a team may not consume, so a contact center running below its licensed seat count pays for headroom. Consumption pricing tied to LLM tokens charges only for work performed and puts a variable, hard-to-forecast number into a budget that finance reviews quarterly. The flat fee buys forecastability and the buyer pays a premium for it. Teams with stable call volumes and a completed AI deployment capture less value from that premium than teams still sizing their volumes.
Viral Nation released the item that inverts the pattern. Viral Nation, a social-first agency and technology company, launched AI Discovery on August 10, 2026, a generative engine optimization (GEO) offering built to shape the signals, authority, narrative and intelligence that influence how AI systems describe and recommend brands. Viral Nation stated that AI systems learn from a broader ecosystem than websites alone, naming owned content, creator content, community discussion, brand authority, executive visibility, earned media, third-party validation and cultural relevance as inputs, and it combined technical optimization, content strategy, creator ecosystems, community engagement, authority building, social intelligence and measurement into a single operating model. Brands buying answer engine optimization (AEO) and GEO services get no approval step over the output. A brand shapes inputs and reads results, and the intervening generation happens inside a system the brand cannot gate.
That asymmetry defines the strategic question the five releases raise together. Consumers get an approval step and ask for it. Upwork clients get an approval step and cite it as the reason they adopted. accessiBe visitors keep the old control surface. Fusion Connect buyers convert an unpredictable variable into a fixed line. Brands operating in agentic commerce discovery get none of the above, and the RTB House trust rankings show consumers already treating those ungated outputs as more reliable than social feeds and newspapers. Marketing leaders should audit which AI-mediated processes touching their revenue carry a human gate and which do not, then spend against the ungated ones first, because those are the processes where a wrong answer propagates without anyone on the team seeing it.
Strategic priorities for marketing leaders
- Test the $250 delegation ceiling against your category. RTB House found 42% of US Millennials comfortable delegating purchases up to $250 with a seven-day return window. Brands with average order values above that threshold should plan against the extended consideration window instead of agent-executed checkout.
- Price the return window as a conversion lever. RTB House measured an eight-percentage-point difference among US Millennials between delegation with and without a guaranteed seven-day return, which merchandising teams can weigh against incremental return-processing cost.
- Write a policy for agent-initiated spend. The Upwork MCP server lets a client’s AI agent post jobs and prepare offers inside Claude, ChatGPT or Cursor, and the built-in confirmation step verifies user intent without verifying budget authority.
- Require a manual control path alongside any conversational overlay. accessiBe retained the accessWidget settings menu when it added its AI assistant, which preserves the deterministic behavior a compliance audit tests against.
- Model per-seat against consumption pricing before signing. Fusion Connect’s flat monthly per-seat and volume-package pricing removes forecasting uncertainty, and teams with stable, known call volumes pay a premium for a predictability they already have.
- Rank your AI-mediated processes by whether a person can gate the output. Viral Nation’s AI Discovery addresses a layer where brands shape inputs and read results with no approval step over what an AI system says about them.
Press Release Roundup: August 10, 2026
Shoppers Trust AI More Than Social Feeds but Still Want Human Control, Reveals RTB House
RTB House, a global performance marketing company founded in 2012, released its global consumer study “Who’s Buying? Consumer Trust in the Age of Agentic AI” on August 10, 2026. The research surveyed 1,840 respondents across the United States and a combined non-US group spanning the United Kingdom, France and Japan during June and July 2026 in partnership with survey sampling platform Cint. RTB House found that leading AI tools now outrank TikTok, Instagram, Facebook, newspapers and major media outlets on shopping trustworthiness among US and international shoppers, with Google AI Overviews and ChatGPT leading and Claude and Grok trailing. RTB House reported that 42% of US consumers and 33% of non-US consumers say AI tools increase the time needed to reach a final purchase decision, rising to 48% among US Gen Z, and that 59% of US consumers and 52% of non-US consumers say AI surfaces brands they did not previously know, with US Gen X at 61% and Baby Boomers at 62%. On financial delegation, RTB House found 42% of US Millennials comfortable giving an AI agent a budget of up to $250 with a guaranteed seven-day return window, against 27% of non-US Millennials, and 34% of US Millennials and 20% of non-US Millennials comfortable without return protection. Jaysen Gillespie, VP of Product Marketing & Analytics at RTB House, stated that consumers use AI as an analytical tool to evaluate and discover brands, compare prices and research before buying, and that the longer research window gives marketers more room to engage. Read the full press release (Published: August 10, 2026 | Source: Business Wire)
Upwork Talent Is Now Everywhere AI Works
Upwork Inc. (NASDAQ: UPWK) announced the Upwork MCP server on August 10, 2026, connecting its marketplace to Claude, ChatGPT, Cursor and any product compatible with the Model Context Protocol. Clients can direct an AI agent to post a job, surface qualified talent, prepare offers and summarize proposals without leaving the AI tool, and freelancers can direct an agent to flag matching opportunities, draft tailored proposals, surface client messages and submit completed milestone work. Peter Sanborn, chief business officer at Upwork, stated that the company began seeing AI agents attempting to log into Upwork on their users’ behalf during 2026. Upwork client Rohit Singh, founder of Populosof, stated that every action changing anything is previewed and requires his explicit confirmation. Upwork stated that every connection runs through an authenticated Upwork account carrying the platform’s identity verification, secure payments, escrow and dispute protections, and that the MCP server is available to all Upwork clients and freelancers at no additional cost. Upwork reported facilitating more than $30 billion in total transactions and services since its founding. Read the full press release (Published: August 10, 2026 | Source: GlobeNewswire)
Viral Nation Launches AI Discovery, a Social-First GEO Offering
Viral Nation, a social-first agency and technology company, launched AI Discovery on August 10, 2026, a generative engine optimization offering built to shape the signals, authority, narrative and intelligence that influence how AI systems describe and recommend brands. Viral Nation stated that AI systems learn from a broader ecosystem than websites alone, naming owned content, creator content, community discussion, brand authority, executive visibility, earned media, third-party validation and cultural relevance as the inputs those systems draw on. The offering combines technical optimization, entity clarity, content strategy, creator ecosystems, community engagement, authority building, social intelligence and measurement into one operating model, and Viral Nation positioned it against a discovery environment where consumers ask AI platforms which brands to consider and how competing options compare. Read the full press release (Published: August 10, 2026 | Source: PR Newswire)
accessiBe Introduces Conversational AI Assistant Built Into Its Accessibility Widget
accessiBe, an end-to-end web accessibility platform, announced its AI accessibility assistant on August 10, 2026, a conversational interface built into the company’s accessWidget product. Site visitors describe what they need in ordinary language, using phrasing such as “the text is too small,” “take me to the footer” or “what’s this page about?”, and the assistant applies the relevant accessibility adjustment, answers questions about page content or guides the visitor to a destination. Robert Lopez, CEO of accessiBe, stated that the accessibility adjustments already existed in the product and that conversational AI reduces the distance between a person and the experience they need. accessiBe kept the existing accessWidget settings menu fully available for visitors who prefer to browse and select adjustments directly. The assistant is live for accessWidget customers on Growth plans and higher, with a phased rollout to additional customers and partners planned, and accessiBe developed it with input from its Inclusive Product Advisory Board and accessLabs, groups composed of people with disabilities. Read the full press release (Published: August 10, 2026 | Source: PR Newswire)
Fusion Connect Simplifies Enterprise AI Adoption with Predictable Flat-Fee Pricing for AI-Powered Contact Centers
Fusion Connect, a US provider of managed communications, connectivity and cloud solutions, announced general availability of its AI-powered Contact Center as a Service platform on August 10, 2026 alongside a flat-fee pricing model. Fusion Connect stated that most AI solutions price against token consumption rate cards and that costs vary with which capabilities run and how often, which leaves organizations unable to estimate spend before deployment. Under the announced structure, core capabilities including Agent AI Assistance and CX AI Analytics carry a monthly per-seat price, and high-volume automation services including Autopilot Agentic AI and AI-powered call routing carry monthly packages priced against anticipated call volumes. The platform’s capabilities cover real-time agent guidance, natural-language intent understanding and routing that replaces menu-driven interactive voice response, autonomous handling of routine interactions without a live agent, and automated analysis of customer conversations for trends, coaching insights and quality issues. John Nee, Chief Marketing Officer of Fusion Connect, stated that organizations should not need a data scientist or a financial model to understand what an AI investment will cost. Frederic Dickey, Senior Vice President of Product Management, stated that the model lets customers focus on business value instead of estimating token consumption. Fusion Connect stated the solutions and pricing model are available immediately. Read the full press release (Published: August 10, 2026 | Source: PR Newswire)
Frequently asked questions
What marketing technology and AI news was announced on August 10, 2026?
Yesterday’s announcements covered consumer AI trust and marketing technology products. RTB House released its global consumer study “Who’s Buying? Consumer Trust in the Age of Agentic AI,” Upwork Inc. (NASDAQ: UPWK) announced the Upwork MCP server, Viral Nation launched its AI Discovery generative engine optimization offering, accessiBe introduced a conversational AI assistant inside its accessWidget product, and Fusion Connect announced flat-fee pricing for its AI-powered Contact Center as a Service platform.
What did the RTB House study find about consumer trust in AI shopping tools?
The RTB House study “Who’s Buying? Consumer Trust in the Age of Agentic AI,” released August 10, 2026, found that leading AI tools including Google AI Overviews and ChatGPT now outrank TikTok, Instagram, Facebook, newspapers and major media outlets on shopping trustworthiness among US and international shoppers. RTB House surveyed 1,840 respondents across the United States and a combined non-US group spanning the United Kingdom, France and Japan during June and July 2026 in partnership with Cint. Trust splits unevenly across AI platforms, with Claude and Grok trailing Google AI Overviews and ChatGPT in both the US and globally.
Does AI shorten or lengthen the path to purchase?
RTB House found on August 10, 2026 that AI lengthens it. In the study, 42% of US consumers and 33% of non-US consumers said AI tools increase the time they need to reach a final purchase decision by surfacing more brands, products and alternatives to consider, with the effect strongest among US Gen Z at 48%. RTB House also found that 59% of US consumers and 52% of non-US consumers say AI surfaces brands they did not previously know, with US Gen X at 61% and Baby Boomers at 62%.
How much money will consumers let an AI agent spend on their behalf?
RTB House reported on August 10, 2026 that 42% of US Millennials are comfortable giving an AI agent a budget of up to $250 to make a purchase on their behalf when a seven-day return window is guaranteed, compared with 27% of non-US Millennials. Without return protection, those figures fall to 34% of US Millennials and 20% of non-US Millennials. RTB House found that approval before checkout is the single most requested safeguard globally.
What is the Upwork MCP server and what can an AI agent do with it?
Upwork Inc. (NASDAQ: UPWK) announced the Upwork MCP server on August 10, 2026, connecting the Upwork marketplace to Claude, ChatGPT, Cursor and any product compatible with the Model Context Protocol. A client’s AI agent can post a job, surface qualified talent, prepare offers and summarize proposals, and a freelancer’s agent can flag matching opportunities, draft proposals, surface client messages and submit milestone work. Upwork stated that every connection runs through an authenticated Upwork account with existing identity verification, secure payments, escrow and dispute protections, and that the server is available to all clients and freelancers at no additional cost.
Why does Fusion Connect’s flat-fee AI pricing matter to marketing leaders?
Fusion Connect announced flat-fee pricing for its AI-powered Contact Center as a Service platform on August 10, 2026, stating that most AI solutions price against token consumption rate cards that leave buyers unable to estimate costs before deployment. Core capabilities including Agent AI Assistance and CX AI Analytics carry a monthly per-seat price, and high-volume automation including Autopilot Agentic AI and AI-powered call routing carries monthly packages priced against anticipated call volumes. Marketing leaders trade a variable consumption cost for a fixed line item and pay a premium for that forecastability.
