This week has seen several major AI firms secure enterprise agreements: Zendesk introduced new AI agents that claim to handle 80% of customer support cases, Anthropic and IBM revealed a strategic alliance, and Deloitte also entered into a partnership with Anthropic. Additionally, Google rolled out a fresh AI platform aimed at businesses.
However, large organizations adopting AI may still face significant challenges. The Deloitte announcement, for example, coincided with news that Australia’s Department of Employment and Workplace Relations required the consulting giant to issue a refund after submitting a report that reportedly contained numerous AI-generated inaccuracies.
On the latest Equity podcast episode, Kirsten Korosec, Sean O’Kane, and I explored the newest AI developments, comparing them to last week’s launch of the Sora app. While AI companies might eventually profit from consumer social apps, enterprise contracts are providing a faster route to substantial earnings.
Below is a shortened and edited excerpt from our discussion.
Anthony: This actually connects to our previous conversation about GenAI social platforms. We considered those as a possible future revenue stream for AI firms, which I still believe, but it’s a long journey. Enterprise deals may not seem as exciting as consumer products, but that’s where the real financial gains are right now.
Perhaps Sora will be OpenAI’s main revenue source in five years, but for now, these enterprise agreements are how these companies are generating income.
The Deloitte situation was particularly noteworthy. It can feel repetitive to highlight that these models aren’t always fully reliable, but I think it’s positive that the Australian government took a stand and said this isn’t acceptable.
It’s not that AI should never be used in creating such reports—though some might argue that—but if you do, you must take responsibility for the results. You need to verify that the cited information is accurate. You can’t just input data into a model and say, “My work is done, here’s the bill.” Anyone who does that should feel ashamed and face penalties.
Kirsten: Absolutely. Sean, Zendesk made an announcement this week as well, and they’re developing tools designed to automate almost all customer service tasks, essentially removing human involvement. In your daily life, or in the way car manufacturers handle service, for example, are you noticing this kind of automation becoming more common?
Sean: Yes, I’ve covered this topic several times. There are numerous startups building comprehensive customer service platforms, including voice agents and large language models for handling emails and texts from dealerships and service providers. I think it’s a valuable approach, not because there aren’t enough people for these jobs or that it will eliminate jobs, but because it’s often impossible to reach someone by phone or you get transferred repeatedly.
When seeking service, you’re often redirected to the service department, and everyone is busy. If automation can accurately capture requests and make it easier for customers to get answers, the real question is whether businesses will adopt and continue using these solutions. Over the years, dealerships have tried various technologies like web forms, but often abandon them, leaving customers thinking they work when in reality, they don’t, because the business prefers phone calls.
So I’m cautiously optimistic that these tools could become the first point of contact for customers. We’ll soon see how it plays out.
Equity is TechCrunch’s main podcast, produced by Theresa Loconsolo, with new episodes every Wednesday and Friday.
You can subscribe on Apple Podcasts , Overcast , Spotify , or any podcast platform. Follow Equity on X and Threads at @EquityPod.
