Agent Fever

March 3 - March 10, 2025

The past week has seen an explosion of new AI “agents” – autonomous AI tools and platforms designed to take on complex tasks with minimal human guidance. Unlike traditional single-purpose chatbots, these AI agents can act with a degree of autonomy, chaining together reasoning steps and even invoking other tools to get jobs done.

This trend matters for marketing, sales, and customer success teams because it promises 24/7 team members who can handle everything from qualifying leads to drafting campaign content without constant hand-holding. In short, AI agents are emerging as tireless colleagues who could dramatically boost productivity and responsiveness.

Behind the frenzy is a convergence of factors: breakthroughs in AI model capabilities, big tech investments, and competitive pressure across continents. In just days, multiple high-profile agent platforms launched or hit milestones, each vying to automate more of the knowledge work pipeline.

The takeaway is clear – the AI agent era has arrived. Those in marketing, sales, and creative fields should pay attention because these tools are quickly moving from hype to practical application, streamlining workflows and unlocking new strategies for engagement. Below, we break down the key developments driving this Agent Fever and what they mean for your business.

Peter & Torsten

The Manus Madness

One of the most buzzed-about launches was Manus AI, a new autonomous agent from China that generated enormous hype on social media. Officially launched on March 5, Manus is billed as the world’s first fully autonomous general-purpose AI agent—essentially an AI “intern” that never sleeps.

Early access has been so coveted that invite codes for its beta were resold for over 100,000 yuan (≈$14,000) amid a waitlist topping one million users. This frenzy underscores the surging demand for AI agents that can handle real work. Manus’s creators describe it as combining “mind and hand,” blending cognitive intelligence with action-taking abilities. In demos, it has surprised observers by completing tasks in seconds that typically require days of human effort.

What can Manus do?

Quite a lot. It boasts autonomous task execution across various activities: writing reports, analyzing data, generating content, and planning travel itineraries. It’s a multimodal agent, meaning it handles text, images, and even code, and can plug into external tools like web browsers or spreadsheets to get information.

For example, Manus can plan a detailed trip abroad, find you a new home overseas, or crunch financial statements with minimal prompts. It continuously learns and adapts to user preferences over time, aiming to become more personalized and efficient with use. In benchmark tests, Manus reportedly outperformed some leading AI models in solving real-world problems, hinting at its advanced reasoning skills.

Why it matters for marketers

Manus positions itself as a versatile digital worker who can handle routine or research-heavy tasks. Imagine having a virtual marketing assistant that can draft campaign reports, pull industry stats, or even do sales scouting (see a demo here).

Early use cases span industries, from automating data analysis (e.g., parsing sales reports) to content creation (e.g., building a quick promotional webpage or comparing product pricing).

For marketing teams, an agent like Manus could offload grunt work such as compiling analytics dashboards or creating multiple ad variants, freeing humans to focus on strategy and creativity. Sales and customer success teams might leverage it to automatically summarize CRM data, screen inbound resumes or leads, and handle first drafts of client outreach.

While Manus is currently in limited beta, its instant popularity signals that businesses are eager for AI that “thinks and acts” in tandem, not just chats. As more companies experiment with Manus, we’ll see whether an AI agent truly can function as an “employee” that reliably delivers – but the initial excitement indicates high expectations for productivity gains.

The path is obvious: it is a matter of time before competent AI bots (agents as we call them now) fully replace parts of the human workforce while also augmenting entire workflows, giving workers a significant productivity boost. AI will not just be part of teams but will become team members.

OpenAI’s Answer

Not to be outdone, OpenAI rolled out new tools this week aimed squarely at empowering AI agents on its platform. These releases mark a pivot for OpenAI, from providing AI models to enabling automation and action.

Compared to Manus, OpenAI’s approach seems more modular and enterprise-focused. Rather than a single monolithic AI assistant, OpenAI is providing building blocks so organizations can create custom agents tailored to their needs.

For example, a sales team could use the new toolkit to build an agent that automatically scans incoming emails, identifies high-priority customer inquiries, and drafts suggested responses for reps. Another agent might continuously research market news and update an internal report. Previously, stitching together web search or file access with an AI model required custom coding; now OpenAI’s API can handle those actions natively. This lowers the technical barrier for companies to experiment with AI-driven automation in their workflows.

Not quite ready yet, but watch this space

For marketing and customer-facing professionals, OpenAI’s expanded toolkit is promising. We’ll soon see more specialized AI assistants embedded in everyday software. Think of marketing tools that generate content ideas, auto-publish posts at optimal times, and analyze the results—all coordinated by an AI agent behind the scenes. Or a customer success dashboard with an integrated GPT-based agent that proactively pulls up relevant knowledge base articles and drafts follow-ups after a client call.

OpenAI enables a new wave of AI automation beyond one-off chat interactions by baking agent capabilities into its platform. The practical upshot: expect your existing software and SaaS vendors to start offering “smarter” features powered by these agents. As this technology matures, professionals should be ready to collaborate with AI helpers who can take on complex tasks, not just answer questions.

Your New Team Member: OpenAI’s Premium AI Agent

On the heels of new tools, buzz is also building around OpenAI’s rumored premium AI agents for enterprise – with eye-popping price tags. Multiple reports suggest OpenAI is plotting tiered offerings of advanced AI “professionals” to target business users . The talk is that companies could subscribe to a “PhD-level AI researcher” agent for $20,000 per month, or opt for a less costly “AI knowledge worker” at around $2,000/month, with a mid-tier AI software engineer at $10,000/month. These hypothetical agents would be highly specialized, top-of-the-line AI systems tuned for expert-level work in their domain. To put it in perspective, $20k a month rivals the all-in cost of a human employee with an advanced degree. It’s a bold gambit – essentially asking, what’s the ROI of an AI that could potentially replace a high-end hire?

If OpenAI follows through, it signals that enterprise adoption of AI is entering a new phase. Companies might be willing to invest heavily in AI if it performs at an elite level. For instance, a $2k/month “knowledge worker” AI could assist a consulting firm by digesting massive reports and producing insights, acting like a tireless analyst. A $10k/month coding agent might handle large portions of software QA and bug fixes overnight. In theory, the top-tier $20k researcher could design experiments or crunch complex data faster than a whole team. OpenAI’s CEO Sam Altman has hinted that these agents are envisioned as “expert assistants” rather than mere chatbots, aimed at tasks that normally require graduate-level expertise. Such capabilities, if realized, could radically alter knowledge work – imagine R&D teams or marketing analytics groups with essentially an AI PhD on staff to brainstorm ideas or run simulations.

However, these rumors also come with healthy skepticism. Analysts and industry voices have questioned whether today’s AI is truly reliable enough to justify that cost. While GPT-4 and its peers are powerful, trusting a $20k/month agent to operate consistently at a PhD researcher’s caliber is a high bar—especially without human oversight.

But a $2k/month junior marketer? Now we are talking.

In practical terms, it underscores the importance of developing AI literacy on your team. So you know how to effectively utilize such an AI specialist if and when it becomes available. It also suggests budgeting for AI in the future might include not just software licenses, but potentially line items for AI talent, just as you budget for human talent.

Salesforce’s AgentForce 2DX Launch

Salesforce jumped into the fray with a major announcement: AgentForce 2DX, an upgraded platform for building and deploying AI agents across sales, marketing, and customer service workflows. Unveiled during the company’s developer conference, this launch is a significant leap for AI-driven CRM.

AgentForce 2DX is designed to let enterprise customers create custom AI agents that operate within Salesforce and beyond – from automating lead follow-ups to handling support tickets – with far less effort than before. A suite of new tools accompanies the platform: a no-cost AgentForce Developer Edition for prototyping agents, an AI-assisted Agent Builder to configure agent behavior with clicks instead of code, and even a Testing Center to simulate and validate how agents perform at scale. In short, Salesforce is lowering the technical barrier so businesses can use their autonomous agents tailored to customer data and processes.

One of the most intriguing aspects of AgentForce 2DX is Salesforce’s vision of a “multi-agent ecosystem.” The idea is that shortly, your personal AI assistant (say, on your phone or in Slack) could directly negotiate or interact with a company’s AI agent on Salesforce. For example, a customer might tell their personal AI, “I need to reschedule my flight,” and that AI liaises with the airline’s booking agent built on Salesforce to handle the entire transaction. This hints at a world where AI agents handle the grunt work of B2C interactions by talking to each other, only looping humans in for final approval.

While this scenario is emerging, Salesforce lays the groundwork by ensuring AgentForce agents can connect through APIs and even embed in channels like Slack and WhatsApp. They’ve launched an AgentExchange marketplace with over 200 partner-built agent components and templates. Like an app store, businesses can pick pre-built agent skills (for marketing, service, etc.) and plug them into their workflows – a massive accelerator for adoption.

Agents will be part of YOUR marketing stack, this year

AgentForce 2DX’s launch strongly signals sales, marketing, and support leaders that autonomous AI is ready for mainstream enterprise use.

Early adopters across industries have reported impressive results: one travel company using AgentForce to automate customer service saved nearly $1.9 million annually, and a recruiting firm drastically reduced time spent screening resumes. The practical benefit for a sales team might be an AI agent that automatically updates CRM entries, drafts follow-up emails after sales calls, and even alerts reps to the best next action with a prospect – all without manual input. For marketers, AgentForce could power an AI that analyzes campaign data from Marketing Cloud and autonomously reallocates budget to the best-performing channels in real time. And for customer success or service, having AI agents that resolve common inquiries or triage issues means faster response times and happier customers.

Salesforce’s move also reflects trust: By building testing and oversight into the platform, they acknowledge companies' need to validate AI actions and maintain control, easing fears of handing keys to a rogue AI. AgentForce 2DX’s debut makes it much easier for businesses to start small with AI agents and scale up success once proven—a notable advancement in making “autonomous CRM” a reality.

Gong’s AI-Powered Revenue Solutions Milestone

AI agents aren’t just coming from new products – they’re already driving success in existing platforms, as evidenced by Gong’s latest milestone. Gong, a revenue intelligence platform that uses AI to analyze sales calls and customer interactions, announced it surpassed $300 million in annual recurring revenue (ARR), fueled by surging demand for its AI-driven solutions. This milestone is telling: it shows that sales organizations are embracing AI not just in pilots, but at scale, to hit their numbers.

Gong’s platform, often dubbed a “virtual sales coach,” listens to calls, transcribes and analyzes them for insights, and guides reps on next steps to close deals. Over the past year, usage of Gong’s AI features has jumped nearly 50%, and one of its newest capabilities – an “Ask Anything” AI Q&A that lets managers query sales data in natural language –saw over 400% year-over-year growth in adoption. That means sales managers increasingly rely on AI to get instant answers (like “Which deals are at risk this quarter and why?”) rather than manually sifting through CRM reports.

What does this mean for sales and marketing teams?

In short, AI-powered revenue enablement has moved from nice-to-have to must-have. Gong’s success suggests that teams using AI for guidance and automation outperform those that don’t. By analyzing thousands of customer interactions, AI can spot patterns – phrases or questions reliably indicating a hot lead or disengaged customer. It can then prompt your reps or account managers with real-time coaching. Gong and similar platforms can even auto-generate call summaries and update CRM fields, eliminating tedious admin work for salespeople. The $300M ARR landmark, alongside a valuation reportedly north of $7B, indicates confidence that this kind of AI delivers tangible ROI. Gong’s customers now span sales and customer success, support, and professional services teams that tap into conversation insights.

For a CMO or CSO, it’s an affirmation that investing in AI for revenue operations (whether through Gong or comparable tools) can yield better forecasting accuracy, more consistent messaging to customers, and ultimately more closed deals. If “Agent Fever” has a proof point, Gong is living it: their AI acts like an ever-vigilant analyst, pinpointing what moves the needle in deals and ensuring nothing falls through the cracks. As competitors and new startups race to add similar AI capabilities, professionals in sales and marketing should be prepared to lean on data-driven coaching – your next sales strategy session might include insights generated by AI parsing millions of datapoints that no human could process on their own.

Conversational Video AI from Tavus

Personalization is a core goal in marketing and customer engagement, and this week showcased a futuristic leap in that direction with Tavus’s conversational video AI. Tavus, a generative video startup, introduced a new system that enables truly interactive, face-to-face style AI videos – essentially video avatars that can see, hear, and converse with users in real time.

This isn’t your average deepfake or prerecorded avatar; Tavus’s technology lets you create a digital twin (a lifelike virtual spokesperson) to carry on a conversation as naturally as a human on a Zoom call.

Under the hood, Tavus rolled out three AI models working in tandem: Phoenix-3, which renders highly realistic human facial expressions (capturing even micro-expressions); Raven-0, which gives the AI the ability to “see” and interpret the user’s visual cues and environment; and Sparrow-0, which manages the timing and flow of conversation so the AI knows when to listen vs. speak. Together, these components allow an AI avatar to look at your facial expressions, understand if you’re confused or interested, and respond with appropriate expressions and answers – almost like a human would.

Conversational video is maturing fast

For marketers and creative professionals, Tavus’s breakthrough opens up an intriguing new channel: conversational marketing via AI video. Imagine cloning your best salesperson or CEO into a virtual avatar that can personally greet each website visitor by name, answer questions about products, and adapt its pitch based on the customer’s reactions – all at once, for thousands of visitors. That’s the kind of scenario conversational video AI can enable. It merges the scalability of AI with the impact of face-to-face interaction.

Early uses of Tavus’s tech could include things like interactive product demos, where a potential buyer can ask a virtual rep (appearing on camera) to show different features, or personalized onboarding in customer success, where new users get a guided tour from a friendly AI persona that responds to their spoken questions.

Importantly, because Tavus’s AI can detect emotions (via Raven-0), it might adjust its approach if it senses a user is frustrated, offering to clarify or switching to a more empathetic tone. This level of personalization and responsiveness is far beyond standard chatbots and could significantly boost engagement and conversion rates in marketing contexts.

Another area of impact is automated content creation for creative teams. Brands could leverage Tavus to generate customized video messages at scale – for instance, a thousand personalized thank-you videos from a founder to top customers, each one unique and addressing the customer by name and interests, without the founder actually recording anything. This kind of one-to-one marketing, previously impossible to do broadly, becomes feasible with conversational video AI. It also raises new creative possibilities: interactive training videos, virtual event hosts, and AI influencers that audiences can converse with.

While it’s early days for this tech, the direction is clear. For professionals, the advice is to start exploring how video-based agents might fit into your customer journey. As tools like Tavus make AI-driven video more lifelike, companies experimenting with these humanized AI interactions will stand out by delivering more engaging, personalized experiences.

Of course, there’s a learning curve to using such tools effectively (and responsibly, given potential uncanny valley effects), but the potential upside in marketing and customer success—where building personal connections is key—is enormous.

Google’s AI Mode in Search

Meanwhile, in digital marketing and SEO, Google pushes the envelope by weaving AI agents directly into search.

This week Google began testing a new “AI Mode” in Google Search, a lab experiment that could foreshadow the future of how consumers find information. AI Mode essentially turns the search experience into a more interactive, conversational process. When enabled, users can ask complex, multi-part questions and get a synthesized answer from Google’s AI (powered by its advanced Gemini 2.0 model), complete with cited links for more info.

Unlike the current Search Generative Experience, which gives short AI snapshots on certain queries, AI Mode goes further – it can perform “reasoning” and comparisons across multiple sources, break a tough question into sub-queries, and then present a detailed response that draws on up-to-date information from the web.

In other words, Google is building an AI agent that acts as a super smart research assistant on top of search, doing the heavy lifting of sifting through results and even letting you conversationally ask follow-up questions.

SEO is dead, long live SEO

The implications for brand discoverability and organic reach are significant. On one hand, Google’s AI Mode still prioritizes showing sources and links – the AI-generated answers are accompanied by website references, meaning content creators can still benefit from being featured.

In fact, Google emphasizes using this approach to bring more depth and breadth than a traditional search by querying many subtopics and pulling in diverse content. This could surface niche blog posts or detailed guides (possibly your brand’s content) that might not have ranked at the top in classic search but are relevant to the nuanced question.

On the other hand, as the AI takes on more of the aggregation and explanation, users might click fewer individual links. If the AI gives a comprehensive answer, users might need to visit multiple websites less. For marketers and SEO professionals, this means strategies must adapt: getting your content picked up in AI summaries could become just as important as traditional first-page rankings. Ensuring your site’s content is well-structured, authoritative, and friendly to AI parsing (e.g. clearly answering common complex questions in your domain) will be key to maintaining visibility.

For creatives and content marketers, Google’s AI mode also presents new opportunities to shine – and new challenges. High-quality, informative content that an AI deems worthy to quote or link will become part of Google’s answer engine. Brands that publish thought leadership or in-depth resources could gain greater exposure if the AI frequently pulls from their material. Conversely, simply churning out SEO keyword pages will no longer cut it; the AI will favor content that truly educates or solves problems.

There’s also a potential shift in the user journey: instead of clicking through to read a full article, users might rely on the AI’s summary and only click if they need more. This makes having a strong brand presence (so that if your name appears as a source, it’s trusted and enticing) more critical than ever. Marketers should monitor this experiment as it unfolds – currently AI Mode is limited to Search Labs and certain subscribers – but the writing is on the wall. Google is embracing an AI-assisted search future, and staying ahead will mean optimizing not just for humans but also for how AI algorithms will consume and relay your content.

The bottom line: continue producing valuable content, but be prepared to adjust SEO tactics to ensure your brand remains discoverable when an AI intermediary is answering the customer’s query.

That is it for this week - what a busy week it was!

If you have any questions, hit REPLY to this email.

If you need help with AI adoption or want to sponsor this newsletter, contact our consulting arm at Anywhere.

See you next week,

Peter & Torsten