Insights
AI confidence doesn’t buy protection in the tech industry
The industry that builds AI must be years ahead at running it, right? Not quite. Tech is the most confident industry when it comes to AI readiness, but its confidence is hiding a deeper issue that only surfaces after agents are already live.
Sinch’s AI Production Paradox report surveyed 420 global tech leaders about the role of AI in their customer communications strategy. The findings reveal a gap between how confident the industry feels and how ready it actually is.
The tech industry is deploying AI with confidence
According to Sinch’s research, 59% of tech leaders are very confident in their AI deployment readiness. Tech’s confidence level is the highest of any industry and nine points above the global average.
How confident are you that your organization is ready to deploy AI agents that autonomously communicate with customers at scale?
This confidence tracks with the industry’s higher-than-average AI production rate. 69% have broken out of pilot projects and into production with AI communications agents, compared to 62% overall.
of technology organizations have AI communications agents in production.
of technology organizations are very confident in their AI readiness
When asked which channels they’re integrating with AI agents, tech leads every industry on website and in-app chatbots (75%) and email (70%). Its top priority for deploying agents overall is customer service(55%), eight points above the overall average and the highest of any industry. This is followed by fraud prevention and security (41%), and identity verification and authentication (38%).
Which areas of customer communication is your organization prioritizing for AI agent deployment? (Select up to three.)
If you’re looking at just this part of the story, before an AI agent gets shipped, the tech industry has every reason to feel confident. But this is only the beginning.
Confidence doesn’t mean immunity from failure
Reaching production is the first test, but keeping an agent live is the harder one. Across the global study, 75% of the leaders who call themselves confident have already rolled back a live agent.
Among all technology organizations running AI agents in production, 66% have had to roll one back after a governance failure.
That is the lowest rollback rate in the entire study, sitting eight points below the industry-wide average. The question is, does a lower rollback rate mean the industry experiences fewer failures or is it just worse at catching them?
Sinch’s global research found that higher rollback rates are actually linked to better, not worse, governance. Programs with fully mature guardrails roll back at a higher rate than average (81% vs 74%), likely because closer monitoring surfaces failures that would otherwise get missed.
Whether tech is failing less or just seeing less, when two thirds of AI programs need to be shut down after going live, the real question you should be asking yourself is why.
Tech’s weak spot is where most failures come from
The number one reason AI agents have to be shut down or rolled back in the tech industry is PII and data leakage (28%).
Has a deployed AI agent ever been rolled back or shut down due to a governance failure?
Private user data ending up where it shouldn’t is exactly the kind of thing the industry fears most. When asked about their biggest technical challenge, the most-cited challenge was data privacy at 49% (vs 44% overall), the highest of any industry.
Tech organizations have reason to be concerned. Their agents typically handle much more customer data than in other industries, from product usage to account and billing records, support history, and their own customers’ stored data. Each of these is another place data can leak from. A support agent that reads account details to answer a billing question is also capable of pasting one customer’s invoice into another customer’s chat.
The consequences of failure hit the brand hardest
When AI agents fail due to a technical issue, most industries feel these failures most in their support queue. Users want to know why the agent gave them the wrong information or why their personal data ended up in a public channel. The consequences are serious, but usually resolved internally without long-lasting damage.
Tech is one of only two industries we surveyed that said the most significant business impact was actually reputational damage (37%).
When an AI agent-driven customer interaction fails due to a technical issue (slow response, lost context, incorrect information), what is the most significant business impact?
When an AI agent fails in a tech organization, the public often knows about it. Picture an AI support agent quoting a rate limit that does not exist. In most industries, that’s a support ticket. In tech, it’s a screenshot in a developer forum, picked apart by the very people the company most wants as its customers.
But you can only protect the brand from a failure you can see coming, and within tech organizations, not everyone is looking at the same picture.
The confidence is highest furthest from the build
Our findings revealed a clear visibility gap within organizations. For every vertical in the study, the further from the consequences, the more confident the leader. And tech is no exception.
69% of senior technology leaders are very confident in their AI programs compared to 53% of Directors and Managers who run the programs. The closer tech leaders are to the work, the less confident they are.
of senior technology leaders (VP and above) are very confident in their AI programs
of technology Directors and Managers running the programs are very confident
When the person approving the AI program budget is more confident than the person responsible for delivering it, problems go unfixed.
“As a C-level leader, you’re often not involved in every detail of making technology, but that’s precisely the point. Executives need to operate at a different altitude, with a longer horizon. You see how the technology behaves, and more importantly, you can anticipate where it’s heading and what it will unlock next. One thing every leader needs to internalize is that the AI we have today is the worst AI we will ever have. From this point forward, it only gets better, faster, and more capable. That’s the lens C-level leaders bring — the ability to see past the near-term friction and recognize where it’s going. And that trajectory is only accelerating.”
So what is that confidence hiding?
Beneath the confidence, the tech industry is grappling with the same issues as every other industry in our study. Just as in retail, financial services, and healthcare, tech is prioritizing investment in trust.
81% of tech leaders say trust and security is their number one investment focus for AI communications. That’s the highest of any industry, and ten points ahead of AI development itself (71%).
Yet, only 30% say they have fully mature guardrails. The trust investment is well intentioned but it’s being sent to the wrong layer, where safety gets bolted on rather than built in.
The real barrier is hiding a layer below. Tech’s number one reported barrier to getting business impact from AI is infrastructure readiness, at 35%, which is higher than anywhere else. Communications infrastructure is its lowest investment category, and the lowest of any industry (53%).
This is concerning given that our study found infrastructure satisfaction to be the single strongest predictor of deployment success.
The AI Production Paradox: See how tech compares
The lead tech has stops at deployment. Its confidence is not buying protection once AI agents go live. Protection from failure is earned with a reliable, AI-ready communications infrastructure.
The AI Production Paradox covers insights from 2,527 enterprise leaders across six industries, looking at what happens once AI agents get to production, the cost of failure, and the importance of a strong foundation. See how tech compares, and what makes an AI program successful.
Author:
