It’s all getting very emotional

February 27 - March 3, 2025

Until recently, AI often felt distinctly robotic—smart enough, sure, but clearly missing the human touch. That’s changing fast. Now, AI tools aren’t just getting smarter and friendlier, warmer, and even surprisingly empathetic. It’s becoming easier than ever for marketing, sales, and customer success teams to use AI for genuine connections, even at massive scale.

Look at just the past week: we’ve heard AI voices so lifelike they’re giving listeners chills, and we’ve seen chatbots that respond with nuance and genuine empathy, sounding more like thoughtful colleagues than automated scripts. Even Alexa is stepping up its emotional game, shifting from a basic assistant to something closer to a personal concierge who remembers your preferences and anticipates your next move.

As AI keeps dialing up emotional intelligence, interactions that once felt forced or fake now feel authentic. For marketers and customer-focused teams, that means new possibilities to engage meaningfully, naturally, and comfortably with customers without the robotic awkwardness.

Let’s get into the details.

Peter & Torsten

AI Voice Tech Breakthroughs – Human-Like Voices at Scale

This week saw leaps in AI-generated speech, narrowing the gap between human and synthetic voices. We’ve seen (or, rather, heard) two new models forming a trend: they use a Large Language Model backbone to imbue voices with realistic emotion and context.

First came Hume AI’s Octave, an AI text-to-speech model that creates lifelike voices with rich emotions. Octave can generate custom voices and personalities from brief prompts, mimicking accents, tones, and styles (e.g. “gentle therapist” voice). A blind test found Octave’s audio more natural and high-quality than ElevenLabs’ output, with listeners preferring Octave’s voice in ~72% of trials.

For a few days, it felt Octave pulled ahead in the race… but then Sesame demonstrated a new Conversational Speech Model so lifelike that it “freaked out” tech journalists. Testers said chatting with Sesame’s AI felt like talking to a real person – one reporter was unnerved when the AI’s voice reminded him of an old friend.

Sesame (founded by Oculus’s co-founder) trained its model on 1 million hours of speech audio, focusing on emotional intelligence and consistent personality in responses. It even rolled out two demo personas, “Maya” and “Miles,” to show off natural back-and-forth conversation.

Why it matters:

We’re entering an era of AI voices that can carry a conversation.

For customer success and sales, this opens the door to truly conversational IVRs and virtual agents – imagine a support hotline where an AI voice calmly walks an upset customer through a solution, and the customer never realizes it’s not human. Early adopters in call centers are already eyeing such tech to enhance service (Sesame’s realism was praised as “insane” by the CEO of Shopify).

Marketers can also leverage emotive AI voices for interactive ads, voice-overs, or personalized audio messages at scale. The caveat is ensuring customers are comfortable – the “uncanny valley” of voices is real. Brands will need to deploy these voices thoughtfully, with transparency and an ear on user reactions, to tap their potential without creeping people out.

OpenAI Releases GPT-4.5 – More Creative, More Empathic, But Pricey

OpenAI rolled out GPT-4.5 on Feb 27 as a research preview of its latest AI model. Touted as the “largest and best” chat model yet, GPT-4.5 boasts a broader knowledge base, better intent-following, and higher “EQ” (emotional intelligence) for more natural interactions. Early tests show it hallucinates less and can generate more creative insights without needing step-by-step prompts.

However, this power comes at a cost: GPT-4.5 is significantly more expensive to use via API – about $75 per 1M input tokens (and $180 per 1M output) versus ~$2.50/$10 for the earlier GPT-4 model.

For now, it’s only available to ChatGPT Pro users ($200/month) and enterprise developers. A wider rollout is expected as OpenAI adds computing capacity.

EQ comes at a price:

On one hand, GPT-4.5’s enhanced creativity and human-like tone can help marketing and sales teams craft more engaging content and even get strategy advice that “feels like talking to a thoughtful person,” as CEO Sam Altman put it. Its better grasp of nuance could improve customer support chatbots, making them more empathetic and effective.

On the other hand, the steep API pricing is a barrier – many businesses may stick with cheaper alternatives for bulk tasks.

If you’re experimenting, use GPT-4.5 for high-value creative work (campaign brainstorming, sensitive customer communications) where its advanced capabilities shine, but watch budget and consider mixing in more cost-efficient models for routine tasks.

Alexa Gets Generative AI

Amazon unveiled Alexa+, the first major Alexa overhaul in a decade, now powered by generative AI. Alexa+ is more conversational and personalized: it can remember user preferences (e.g. dietary needs), make multi-step decisions like booking dinner reservations, summarize documents, and even control smart-home devices more intuitively. The service costs $19.99/month for non-Prime users but is free for Prime members, and it’s rolling out to U.S. users starting in March.

Why it matters:

For marketers, Alexa+ could open a new channel for voice commerce and advertising. Amazon hints at eventually weaving ads into Alexa’s more natural interactions, allowing brands to engage consumers via voice in the home. As Alexa+ potentially becomes a digital concierge that “knows almost every instrument in your life,” brands might leverage it for personalized promotions (e.g. recipe skill recommending a specific ingredient or product).

It’s a step toward more intimate, AI-driven customer experiences, but Amazon will need to balance monetization with user trust so Alexa doesn’t become a chatterbox of ads.

Chegg Sues Google over AI Search

Education company Chegg filed an antitrust lawsuit against Google, alleging that Google’s new AI-generated search summaries (“AI overviews”) are siphoning student traffic away from Chegg’s site.

Chegg claims these instant answers lift content from Chegg but keep users on Google, causing Chegg’s visits (and revenues) to plummet – its stock is down ~90% in a year. The suit argues Google is abusing its monopoly by using publishers’ content to satisfy queries without credit or clicks back.

SEO Alarm Bells:

Any business reliant on search visibility should take note. Google’s AI summaries represent a broader shift in SEO – users may get the info they need straight from the search page, bypassing your website. Chegg’s plight highlights the risk to content and SEO marketing models.

For marketers, this means doubling down on strategies for an AI-driven search era: integrating with answer engines, ensuring brand content is the go-to source that AI pulls from, and diversifying traffic sources. In customer support, it may spur more on-site AI chat solutions, so users don’t have to search the web (and potentially never reach you).

The bottom line: as AI transforms search, companies must adapt or risk “losing their homework” to the Googles of the world.

You.com’s ARI Research Assistant – AI-Powered Market Research.

You.com introduced ARI, an AI research agent that rapidly synthesizes information from 400+ sources, outputting reports complete with citations, charts, and key insights. ARI is designed to streamline competitive and market research, especially in data-intensive or regulated industries where accuracy is crucial.

Use cases:

• A marketing analyst could query ARI to get a distilled report on consumer trends or a competitor’s digital strategy, saving hours of manual research.

• Before client meetings, sales teams might use ARI to gather industry stats and talking points.

• By automatically compiling verified facts and visuals, tools like this can speed up and confidently enhance decision-making and content creation (e.g., whitepapers, strategy decks).

Alibaba’s Open-Source AI Video Generator – DIY Video Content via AI

Chinese tech giant Alibaba just made a splash by open-sourcing its Wan 2.1 AI model, which generates images and videos from text prompts. Four variants of the Wan 2.1 model (including text-to-video and image-to-video at 14-billion parameters) were released for free on Alibaba’s ModelScope and Hugging Face, available globally for researchers and commercial use.

Time to experiment:

For marketers, this is an opportunity to experiment with AI-generated video content at low cost. Until now, high-quality text-to-video has mostly been in closed labs or paid APIs. Now teams with technical skills can try Alibaba’s model to create short promos, concept videos, or interactive visuals without production crews or licensing fees.

For example, a social media manager could input “A serene night scene with our product on a lily pad” and get a short ambient video for an ad background. This also signals a competitive push: open-source models like this can spur faster innovation and cut down costs, ultimately giving businesses more creative tools.

Keep in mind that the outputs might not yet rival a professional studio—and you’ll need developers to run the model—but the barrier to entry for AI-driven video marketing just got lower. As open models improve, marketers can be less dependent on big AI vendors and more agile in trying bold visual content ideas.