
This is "The AI Economy," a weekly LinkedIn-first newsletter about AI's influence on business, work, society and tech and written by Ken Yeung. Sign up here.
This week, we’re exploring AI agents, an emerging topic that will dominate company discussions in the coming months as they reshape business interactions, both between organizations and consumers.
I just returned from New York City, where I attended Zendesk’s inaugural AI Summit as a company guest. It seems like AI summits are everywhere these days; Nvidia even hosted one in Washington, D.C., just before. For the customer experience company, their event showcased how they’re bringing their AI agents and copilot to call centers. This made me curious about how artificial intelligence will transform how we speak to a company representative.
On the sidelines of the event, I spoke with Kevin Boyer, Zendesk’s Senior Director of Product Marketing, to gain insight into how the introduction of AI might change the customer experience. Will consumers still feel frustrated when they have to pick up the phone, dial a 1-800 number, and repeatedly press zero just to reach an operator?
Plus, stick around for this week’s roundup of AI news headlines you may have missed!
Disclosure: I attended Zendesk's AI Summit as a guest of the company with my flights and hotel covered. Zendesk did not dictate the contents of this post. These are my words.
Zendesk’s Use of AI to Restore ‘Humanity’ to IVR
With Zendesk expanding its AI agents and copilots to email and voice, I asked myself—will this be the end of dreadful customer support calls? Jon Aniano, the company’s senior vice president of products, told me that voice is a customer service channel people both love and hate. They love it because of the one-on-one interaction, complete with vocal cues, human emotion, and empathy. But there are plenty of reasons to hate it. So, with Zendesk inserting AI agents and, eventually, its copilots, will the interactive voice response (IVR) system we’ve adapted to and equally loathe be a thing of the past?
“Your traditional IVR is designed with the productivity and efficiency of the company in mind,” Boyer explains to me. “Yes, it can often get a customer to the right answer faster if there’s a big queue of customers waiting ahead of you. But that’s not really the purpose.” He claims it’s more about deflection, to route customers to the appropriate department, but the onus is on the customer, not the company.
“You’ve taken a service that’s that would typically have been provided by a human face-to-face, and you’ve pushed it back on the consumer, and now the consumer is doing the work for you,” he remarks. “And in some cases, that makes sense. If you’ve got a limited IVR tree, if you have a limited number of inquiries, and you really like trying to move people from one category to another to get them into the right queue. But the more complex the conversation gets, the more difficult it is for that to really achieve any level of satisfaction.”
How AI Can Eliminate Fear of Calling the 800-Number
Zendesk’s AI infusion is designed to restore “humanity” to the process, and it believes its AI agents and copilots can accomplish that. Boyer states that everything makes it feel like you’re talking to a human, from the tone, accent and the conversation. “You don’t feel that same level of anxiety when you’re hearing a robot, and you’re trying to use the phrases that a robot would understand because generative AI can understand all of that already.”
He offers this take on the call journey: “You added humanity by taking the human opus of conversation, giving AI access to it, so it can then communicate to you as a human would because it knows the way humans would communicate. It doesn’t create a genuine human connection, but it adds a human-like experience to the front end of that. So now your deflection process is not just deflection, it’s actually resolving your issue, and it feels more satisfying to the person, and so you get a higher customer satisfaction rate while you’re getting the resolution of much more complex issues.”
Boyer pitches that technologies such as Zendesk’s AI agent copilot can provide the proper context. “It’s giving you language more about the situation itself, a tone that represents a calming influence,” he says. The bots can detect the customer’s emotions to make better judgment calls. “It’s making the humans more human…almost,” Boyer points out. “In a voice conversation, that’s going to be important because there’s no real time to go and check all the information — to have information in your ear or…on the screen in front of you…and at the end, if you’re evaluating empathy, tone or completeness of the solution or speed of resolution as part of your QA, you can then use that to coach your human agents or your AI agents to be better.”
“It adds humanity back into the full conversation at every step of the way,” he declares.
During our interview, Boyer previewed some of the other voice enhancements that will be coming to Zendesk. In two weeks, the company plans to launch Session Initiation Protocol, or SIP, which will offer data dripping for IVR. SIP enables enterprise customers to modify their communication processes using the Web programming language of their choice. In addition, the company will soon support “bring your own carrier” and launch new agent reporting metrics, specifically within voice. One could argue that Zendesk might be encroaching more on Twilio’s turf.
What does he think about the future of IVR? Boyer personally hopes these systems disappear, as he feels they don’t truly help customers. However, he acknowledges that IVR is here to stay. The future, he believes, lies in integrating AI into these outdated systems, allowing the technology to listen, interpret, and either resolve customer issues directly or transfer them to a human agent best equipped to handle the situation “without you having to press a bunch of keys.”
“AI creates the opportunity for us to really add back in the genuine human, whether real human or ‘human’ connection, because of the precision of AI. And IVR is a good example of something right for replacement—not just revolution—but a complete replacement, in my opinion.”
Will Customers Embrace AI Support?
While some companies may rush to adopt AI immediately at launch, there’s no guarantee that their customers, the buyers of their product, will embrace it. Boyer cautions that the public will only be content with having AI handle their problems if “they feel that the AI piece of this is better or as good as what they would get from a human.” Did the AI provide accurate information? The technology remains prone to hallucinations, regardless of where the data comes from. Was the issue resolved quickly, and was the resolution one that the customer expected—or better—than what a human agent could have provided?
Boyer argues that human behavior will only change once software providers prove their technology can rise to the challenge. “We have to lead that because we can’t expect humans to change until they have the experience that meets their expectations that they would get with another human being.”
Zendesk is seeing promising early results, with its AI helping automate, in some use cases, up to 50 percent of customer issue resolutions without a human getting involved. Much of that is thanks to its partnership with PolyAI, a startup dedicated to helping call centers deliver a “best-in-class” voice experience. However, when a human is needed, Zendesk’s AI-powered workforce management software kicks in and instructs agents to respond. Advanced cues and training will ensure that the right call is assigned to the right human agent based on their skills or language.
The Difference Rests in the Expertise and Technology
What advice would Boyer provide companies looking at AI solutions for their call centers or to handle customer service? After all, Zendesk isn’t the only software provider offering AI for this industry. Remember when Google demonstrated Duplex at its developer conference, which some considered a significant tipping point for AI? The market has become increasingly crowded, featuring players like Salesforce, Twilio, Google Cloud, Talkdesk, Genesys, Five9, and startups Observe.ai, Cresta, Replicant, and Sierra.
“What you have to do is, you have to have the technology, and then you have to have the expertise that has to come together to be able to put forth the solutions humans will use and thus, behavior will change,” Boyer responds. “That’s why I think you see a difference now between what you would see two or three years ago and what you see today.”
He says Zendesk is in a strong position to lead this field because the company possesses billions of conversations that it can evaluate to expand its AI. As that improves, so do the firm’s customers’ databases and conversations stored on the Zendesk platform.
“You can’t just throw something at people who don’t understand customer service and expect it to be useful for customer service,” Boyer argues. “It’s really the technology improvement, the [customer experience] expertise that Zendesk has, our commitment to being easy to use and easy to scale, which means we can bring it to life for customers without as much substantial investment on their part, and then specifically, you then bring in the voice piece of it, and I think that’s the difference.”
▶️ Read more about the announcements Zendesk made at its AI Summit (My Two Cents)
Today’s Visual Snapshot

This chart shows monthly web traffic in 2022 to the leading stock photography sites. It also displays traffic delivered to AI image generators, OpenAI and Midjourney. At one point, the AI sites are close to that of iStock, Getty Images and Shutterstock. It will be worth seeing how this compares to 2023 and 2024, when the generative AI space heats up, and more AI image generators have emerged in the market, including Google’s Imagen, Adobe Firefly, Stable Diffusion, and Runway. The trend also casts a spotlight on the ethical use of AI and the risks it poses to non-AI creatives and photographers who find themselves in a fight for survival against a blossoming technology.
Quote This
“One key difference between graduate students and AI is that graduate students learn. You tell an AI its approach doesn’t work, it apologizes, it will maybe temporarily correct its course, but sometimes it just snaps back to the thing it tried before. And if you start a new session with AI, you go back to square one. I’m much more patient with graduate students because I know that even if a graduate student completely fails to solve a task, they have potential to learn and self-correct.”
— UCLA mathematician Terrance Tao, the “Mozart of Math,” explains why he described OpenAI’s o1 model as the equivalent of a mediocre research assistant.
“I’m particularly proud of the fact that one of my students fired Sam Altman. And I think I better leave it there…”
— Professor Geoffrey Hinton, the recipient of the 2024 Nobel Prize in Physics, remarks on how he felt when former student and ex-OpenAI Chief Scientist Ilya Sutskever participated in the brief ousting of CEO Sam Altman in 2023. He would elaborate further: “Over time, it turned out that Sam Altman was much less concerned with safety than with profits, and I think that’s unfortunate.”
This Week’s AI News
🏭 Industry Insights
- “Godfather of AI” Geoffrey Hinton and computer scientist John Hopfield awarded Nobel Prize in physics for “fundamental discoveries” in machine learning (CNN)
- Google DeepMind’s Demis Hassabis and John Jumper join scientist David Baker in being awarded the Nobel Prize in chemistry for work in predicting and designing the structure of proteins (Associated Press)
- Agents are the future AI companies promise — and desperately need (The Verge)
- Google’s grip on search slips as TikTok and AI startup mount challenge (The Wall Street Journal)
🤖 General AI and Machine Learning
- AWS CEO Matt Garman on generative AI, open source, and closing services (TechCrunch)
- Hugging Face’s new tool lets developers build AI-powered web apps with OpenAI in just minutes (VentureBeat)
- DeepMind’s Michelangelo benchmark reveals limitations of long-context LLMs (VentureBeat)
- Uber to launch OpenAI-powered AI assistant to answer driver EV questions (TechCrunch)
- BIg AI thins out the competition as startups quit the race to build large language models (Fortune)
- Anthropic challenges OpenAI with affordable batch processing (VentureBeat)
- Website builder Squarespace says it’s training its AI tools with curation and taste (TechCrunch)
✏️ Generative AI
- Forget chat: AI can hear, see, and click is already here (MIT Technology Review)
- OpenAI’s GPT Store has left some developers in the lurch (Wired)
- OpenAI to integrate SearchGPT into ChatGPT before the end of 2024 (Search Engine Land)
- The incredible blandness of AI photography (The Verge)
- Goodnotes adds an AI that can read and explain even the worst handwriting (TechRadar)
- Zoom will let AI avatars talk to your team for you (The Verge)
- New high-quality AI video generator Pyramid Flow launches in open source (VentureBeat)
🛒 Retail and Commerce
- Amazon’s new AI-powered vision tech tells drivers which packages to deliver (TechCrunch)
- Amazon debuts AI-powered shopping guides (TechCrunch)
- Walmart bets on multiple models with new Wallaby LLM (VentureBeat)
☁️ Enterprise
- AI startup Writer launches new model to compete with OpenAI, currently fundraising at a $1.9 billion valuation (CNBC)
- Atlassian’s Rovo, an AI-driven search engine for enterprises, is now generally available (My Two Cents)
- Vectorize debuts agentic RAG platform for real-time enterprise data (VentureBeat)
- Driver launches an AI-powered platform for creating technical documentation (TechCrunch)
- Sierra AI gives its AI agents a voice, delivering lifelike customer service over the phone (My Two Cents)
- Google’s Gemini enterprise coding assistant shows enterprise-focused coding is growing (VentureBeat)
⚙️ Hardware and Robotics
- Tesla reveals 20 Cybercabs at its We, Robot event and says you will be able to buy one for less than $30,000 (TechCrunch)
- AMD launches Instinct MI325X AI chip to rival Nvidia’s Blackwell (CNBC)
- Intel debuts new flagship CPUs that run cooler and more efficiently for PC gaming (The Verge)
- Nvidia CEO Jensen Huang says demand for next-generation Blackwell AI chip is “insane” (CNBC)
- The seven major announcements Nvidia made at its AI Summit this week (VentureBeat)
- AI Platform Alliance brings system and chip companies together (VentureBeat)
- Rabbit CEO Jesse Lyu isn’t thinking too far ahead (The Verge)
- OpenAI leaders say Microsoft isn’t moving fast enough to supply servers (The Information)
- Shield AI’s founder explains the AI weapon “no one wants” (TechCrunch)
- Rise of the robots: AI to shape UK defense review (Politico)
🔬 Science and Breakthroughs
- How AI “twins” could help fill women’s health gap (Axios)
- CareYaya’s QuikTok is an AI phone companion for lonely aging adults (VentureBeat)
💼 Business and Marketing
- The OpenAI talent exodus gives rivals an opening (Wired)
- Amazon Web Services VP of AI, Matt Wood, is leaving the company (GeekWire)
- Amazon Web Services CEO Matt Garman explains company’s “any model you want” strategy (Axios)
- Meta debuts new generative AI tools for creating video-based ads (SiliconAngle)
- TikTok joins the AI-driven advertising pack to compete with Meta for ad dollars (Digiday)
- Adobe has a new tool to protect artists’ work from AI (The Verge)
📺 Media and Entertainment
💰 Funding
- Bret Taylor’s AI agent startup Sierra nears deal that could value it at over $4 billion (The Information)
- Unify, an OpenAI Converge accelerator grad, lands $12 million for “warm outbound” messages (TechCrunch)
- Generative AI fact-checking firm Infactory raises a $4 million seed (TechCrunch)
- AI legal tech startup EvenUp raises $135 million to hit unicorn status (Crunchbase News)
- VESSL AI secures $12 million for its MLOps platform that aims to cut GPU costs by up to 80% (TechCrunch)
- CoreWeave secures $650 million credit line from Wall Street banks to expand the cloud infrastructure company’s business and data center portfolio (CNBC)
- Healthcare startup Suki raises $70 million to build AI assistants for hospitals (Reuters)
- Basecamp Research draws $60 million to build a “GPT for biology” (TechCrunch)
💥 Disruption and Misinformation
- OpenAI: Bad actors are using its platform to disrupt elections, but with little “viral engagement” (CNBC)
- The rise of AI-powered job application bots (404 Media)
- Wimbledon tennis tournament replaces line judges with AI in break with tradition (Associated Press)
- The editors protecting Wikipedia from AI hoaxes (404 Media)
- AI is a multi-billion dollar industry. It’s underpinned by an invisible and exploited workforce (The Conversation)
- Study: AI-assisted tutoring boosts students’ math skills (The 74)
- Hurricane Helene and the “fuck it” era of AI-generated slop (404 Media)
🔎 Opinions, Analysis and Research
- Why we’re teaching LLMs to forget things (IBM)
- What feels real enough to share (Garbage Day)
- Senator Chris Coons: AI deepfakes threaten Americans and our election (U.S. News & World Report)
End Output
Thanks for reading. Be sure to subscribe so you don’t miss any future issues of this newsletter.
Did you miss any AI articles this week? Fret not; I’m curating the big stories in my Flipboard Magazine, “The AI Economy.”

Connect with me on LinkedIn and check out my blog to read more insights and thoughts on business and technology.
Do you have a story you think would be a great fit for “The AI Economy”? Awesome! Shoot me a message – I’m all ears!
Until next time, stay curious!
Subscribe to “The AI Economy”
New issues published on Fridays, exclusively on LinkedIn
Leave a Reply
You must be logged in to post a comment.