Big deals from Zendesk, Anthropic, Deloitte, and Google show enterprise AI adoption is accelerating — but the rollout is proving messy, high-stakes, and far from foolproof.
AI for Business: The Real Revenue Engine
This week brought a flurry of enterprise AI announcements, signaling that major corporations aren’t waiting around to test whether generative AI is “ready” — they’re deploying it now.
- Zendesk unveiled AI agents designed to autonomously handle 80% of customer service issues.
- Anthropic partnered with both IBM and Deloitte to expand Claude’s enterprise footprint.
- Google announced a new AI-for-business platform aimed at making its AI tools more accessible across large organizations.
These partnerships reflect a clear pattern: while consumer-facing AI apps (like OpenAI’s Sora) may be splashier, enterprise deals offer more immediate and predictable revenue.
“Maybe Sora is how OpenAI will make money five years from now,” said Equity podcast host Anthony Ha. “But this is how these companies are going to make money now.”
AI-Generated Reports Gone Wrong
Of course, enterprise AI adoption isn’t without growing pains — or accountability problems.
On the same day Deloitte announced its partnership with Anthropic, the firm was ordered by the Australian government to refund a contract after delivering a report riddled with what appeared to be AI-generated hallucinations. The report included citations that couldn’t be verified, and officials determined it failed to meet professional standards.
“If you’re going to use AI, you still have to take responsibility for the outputs,” Anthony emphasized. “You can’t just feed it into a model and say, ‘My job is done.’”
The incident underscores a key concern: AI tools can’t replace professional accountability. If enterprises move too quickly without quality control, trust and credibility could suffer — along with real-world consequences.
Zendesk’s AI Agents: Customer Service Gets a Makeover
At the other end of the spectrum is Zendesk’s AI-first approach. The company claims its new virtual agents can handle the majority of support tickets without human intervention, drastically reducing wait times and improving resolution rates.
“It’s not about taking jobs — it’s about actually getting a response,” said reporter Sean O’Kane. “You call a dealership or a service center and get bounced around. These tools could finally solve that bottleneck.”
The key question? Will businesses stick with it — or will AI fall into the same trap as past tools, like forgotten web forms that never worked properly?
A Tipping Point in Enterprise AI
- IBM + Anthropic: A partnership with implications for secure, regulated industries like banking and healthcare.
- Deloitte + Anthropic: Consulting giants are embedding AI into strategy, compliance, and operations — but must avoid cutting corners.
- Google’s new AI platform: Targeting a broader base of businesses looking for customizable, secure LLM integrations.
Enterprise AI is rapidly shifting from experimentation to execution. These aren’t just pilot programs anymore — they’re being marketed as core business solutions.
Responsible AI Still Lags Behind
Despite the optimistic headlines, there’s a growing disconnect between AI’s promise and performance. When enterprise vendors push AI into production without safeguards, the risk isn’t just faulty output — it’s regulatory backlash, reputational damage, and loss of public trust.
The Deloitte-Australia incident is a case in point. The AI-generated content wasn’t the issue itself — it was the lack of oversight, the absence of fact-checking, and the assumption that automation equals deliverable.
“Anyone who delivers that kind of report and bills for it should be embarrassed — and fined,” Anthony said bluntly.
Where This Is All Heading
As Zendesk’s rollout shows, there’s a clear appetite for AI systems that can streamline repetitive work and improve user experiences. But if Deloitte’s mistake is any indication, hype without responsibility will only go so far.
Companies that get this right will likely become market leaders in AI deployment. Those that don’t could end up spending more on damage control than innovation.
