MarketWatch host Christine G interviews IBM’s David Cox about why AI may matter more as a reliable, boring enterprise tool than as a flashy AGI story. Cox argues most enterprise pilots fail because they’re expensive, hard to secure, and too broad; the real value comes from narrow, efficient models embedded into business workflows.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
Christine G opens by framing the current AI debate as hype versus bubble, then brings on David Cox from IBM, who argues the market is overrating “human-like” AI and underrating practical, enterprise-grade automation. His core thesis is that AI’s real value is likely to come from narrow, reliable, boring systems that solve specific business problems at the right cost, rather than from pursuing AGI-style general intelligence for its own sake. Cox says the current phase of AI is a “growing pains” period: the technology is improving quickly, but it is still hard to use reliably, securely, and consistently. …
Near term, the risky part is the gap between AI spend and provable ROI: expensive deployments, legal liability, and security failures can still hit sentiment fast. Tactical enthusiasm is better reserved for companies showing concrete enterprise monetization rather than generic AI exposure.
Over the next few months, AI likely keeps spreading through enterprise workflows, but the winners should be the firms that prove narrow, reliable automation at sane cost. If open models continue closing the capability gap, pricing power for closed frontier labs could come under pressure.
Structurally, AI looks less like a temporary hype cycle and more like a utility layer that will be absorbed into software and operations. The long-run question is not whether AI matters, but whether the benefits are broadly distributed through open ecosystems or concentrated in a few dominant providers.
AI will be valuable, but the current phase is a difficult growing-pains period for reliable enterprise use.
He says the tech keeps getting better and will change everything, but is still hard to use reliably, consistently, and securely.
Many gen-AI pilots fail because companies build easy MVPs, overpay for deployment, and run into security and legal problems.
He attributes the 95% failure narrative to economics and non-functional requirements, not just model capability.
Consumer chatbot behavior does not translate cleanly into enterprise monetization or acceptable enterprise risk.
He contrasts laughing off consumer mistakes with business contexts where errors can create lawsuits or reputational harm.
Why is the disconnect happening between companies trying to adopt AI for productivity gains and failing?
David Cox says AI is in a growing pain phase where it's hard to use reliably, consistently, and securely. He describes it as a 'cocky adolescence' that needs to mature before it can truly change everything.
What went wrong with the 95% of gen AI pilots that MIT found were not adding value?
David Cox identifies several issues: companies build easy MVPs that don't add real business value; AI applications are very expensive to deploy and costs can overwhelm them; and there are non-functional problems like security and legal risk. He believes the spirit of the MIT report is true.
Will AI value come from enterprise or consumer applications?
David Cox says it will be both, but they look completely different. Enterprise has enormous money in back-end systems and workplace productivity. Consumer AI quirks are forgivable, but in business those same issues could cause lawsuits or reputation damage. He also notes consumer AI like ChatGPT has huge user bases but most don't pay, making monetization difficult.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
