95% of AI Agent Projects Fail to Reach Production. Here's Why | Manoj Saxena, TrustWise

Topics

Read next

• Helft bedrijven stuurt agenten naar productie die evaluaties haalden maar faalden

  Onderzoek van 157 bedrijven toont aan dat 50% een agent heeft uitgerold die interne evaluaties passeerde maar dan bij klanten faalde, terwijl slechts 5% volledig automatische evaluatie vertrouwt.

• 54% van bedrijven ondervindt beveiligingsincidenten met AI-agenten

  Onderzoek van 107 bedrijven toont aan dat meer dan de helft al AI-agent beveiligingsincidenten of bijna-incidenten heeft meegemaakt, en dat meeste agenten nog steeds credentials delen zonder op-scope identiteiten.

Description from the channel

It takes a few hours to build an AI agent. It takes six to seven months to get it into production. Manoj Saxena - the man who commercialized IBM Watson - has built a company around solving exactly that problem. In this episode, Craig Smith sits down with Manoj Saxena, CEO and founder of TrustWise, to examine what he calls the most underaddressed crisis in enterprise AI: the gap between deploying intelligent agents and actually controlling them once they're running. Saxena's central argument is captured in a single line ("intelligence without control is not deployable") and the statistics back it up: 95% of agent projects are failing to move from pilot to production, not because the models aren't capable enough, but because enterprises have no infrastructure to govern agent behavior at runtime, at scale, and across multiple vendors simultaneously. TrustWise's answer is what Saxena calls the AI Control Tower, a vendor-agnostic layer that sits above agent orchestration frameworks and evaluates every tool call, action, and output in 10 to 300 milliseconds against up to six layers of alignment, from global human rights frameworks down to customer SLA requirements. The conversation also covers TrustWise's demonstrated results, 83% cost reduction, 40% safety improvement, 60% latency reduction, and a milestone that Saxena describes as the equivalent of data traffic surpassing voice on AT&T's network: last month, for the first time ever, agent traffic on the internet exceeded human traffic. The episode closes with a preview of TrustWise's next product: Genesis agents, designed not just to prevent bad outcomes but to surface beneficial hypotheses, looking 95 moves deep into domains like revenue leakage and fraud, in the way that Deep Blue looked 95 moves deep in chess. Key Topics Covered: ● Why 95% of enterprise AI agent projects stall between pilot and production, and why the bottleneck is governance, not model capability ● The token consumption paradox: why one input into an agentic system can trigger 20 to 50 actions and consume 40x more tokens than a generative AI interaction two years ago, even as token costs fall ● How TrustWise's AI Control Tower operates in four modes, live, sidecar, batch, and simulation, and evaluates agent behavior against six layers of alignment in 10 to 300 milliseconds ● Why agent traffic on the internet exceeded human traffic for the first time last month, and what that milestone means for enterprise AI infrastructure ● The new Harmony AI product launch: Guardian agents (preventing bad outcomes) and Genesis agents (surfacing unknown unknowns by looking 95 moves deep in domains like fraud and revenue leakage) ● Why OpenAI and Anthropic can't solve the enterprise AI control problem, and why the real data to make AI truly transformative belongs to enterprises, not model providers As agentic AI moves from demos to production systems running autonomously for hours and days inside critical business processes, the absence of runtime governance isn't a technical detail,it's the primary reason AI ROI is failing to materialize at enterprise scale, and this conversation offers the clearest framework currently available for understanding what solving that problem actually requires. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI. Craig Smith on X: https://x.com/craigss EYE On A.I. on X: https://x.com/EyeOn_AI Connect with Manoj Saxena LinkedIn: https://www.linkedin.com/in/manojsaxena Timestamps 00:00 Why AI trust has become a critical issue 01:14 From IBM Watson to founding TrustWise 04:31 Why agentic AI needs runtime governance 07:05 Using AI to control AI 11:49 Guardian agents and AI shields 14:14 How humans manage an AI workforce 19:43 Why AI governance is now a C-suite issue 24:33 Protecting enterprises from powerful AI agents 27:40 How the AI Control Tower works 31:21 The hidden cost of agentic AI 38:10 The emerging cyber trust market 41:59 Navigating AI regulation and compliance 44:24 Why AI agents require a new enterprise stack 47:28 The rise of multi-agent workflows 49:39 How AI alignment actually works 54:34 Can AI learn human values? 58:31 Guardian agents, Genesis agents, and the path to AGI