Current and former insiders shed light on IBM's missteps with Watson, as IBM takes a less ambitious approach to commercializing Watson in the corporate market
IBM's artificial intelligence was supposed to transform industries and generate riches for the company. Neither has panned out. Tweets: @counternotions , @leapingrobot , @scottnations , @patentology , @nytimes , @sangernyt , @slavov_n , @storyneedle , and @claudiawilliams Tweets: Kontra / @counternotions : 2011: Steve Jobs on IBM ↓ 2021: Reality on IBM https://www.nytimes.com/... https://twitter.com/... W. Patrick McCray / @leapingrobot : An overhyped AI has failed to live up to expectations https://www.nytimes.com/... Scott Nations / @scottnations : The most interesting line? IBM's stock price is down 10% since Watson's victory on “Jeopardy” in 2011. https://www.nytimes.com/... Mark Summerfield / @patentology : Like many other past contestants on Jeopardy, IBM's Watson has gone on to do very little else of any great note https://twitter.com/... @nytimes : IBM's Watson was supposed to transform industries and generate riches for the company. Neither has panned out. Today, it stands out as a sobering example of the pitfalls of hype and hubris around AI. https://www.nytimes.com/... David Sanger / @sangernyt : Fascinating read from the great @SteveLohr on IBM's travails with Watson, once considered in the vanguard of AI. A take of over-ambition and over-expectations. https://www.nytimes.com/... Nikolai Slavov / @slavov_n : It's remarkable how industry can waste billions of dollar based on premisses that a technically informed observer could have easily seen as unsubstantiated https://www.nytimes.com/... Michael Andrews / @storyneedle : IBM Watson is a content marketing anti-pattern. The narrative sounded great but it over-promised and under-delivered. Some candid remarks here from IBM insiders. https://www.nytimes.com/... @claudiawilliams : Striking difference between todays @awscloud announcement “we will help you organize messy data” https://press.aboutamazon.com/ ... and past failure of IBM Watson “we will solve health's hardest problems with AI” https://www.nytimes.com/... First is far less sexy but way more effective
Context & Ripple Effects
The arc here is a decade-long retreat. After the 2011 Jeopardy win, IBM pitched Watson as a universal engine — deployments across 75 industries within four years — then spent heavily on staff and healthcare bets, insisting the investment was yielding profitable opportunities even as a 2017 investigation found Watson for Oncology fell far short of what IBM had promised clinicians.
Today's reporting closes that loop: insiders say the technology never transformed an industry or produced riches, IBM shares are down roughly 10% since the Jeopardy victory, and the company is now pursuing humbler corporate-market uses — following February's move to explore a sale of the unprofitable, ~$1B-revenue Watson Health unit.
First-order effects
- IBM abandons the 'transform every industry' framing for Watson and repositions it as a narrower enterprise tool, ceding the grand-ambition AI narrative it built after Jeopardy.
- Current and former insiders publicly attribute the failure to IBM's own handling of Watson — a reputational cost on top of the financial one for a company still selling AI services.
Second-order effects
- Enterprise buyers who took Watson's healthcare and industry pitches at face value now have documented grounds for skepticism about big-vendor AI claims, raising the bar for IBM's remaining AI sales.
- Rivals in enterprise AI can position against IBM's decade of overpromising, making demonstrated vertical outcomes — not demos — the new proof point for winning corporate accounts.
Third-order effects
- The Watson arc becomes the canonical case study of demo-to-enterprise commercialization risk: a famous research win does not guarantee a product-market fit, and boards may discount flagship-AI narratives accordingly.
- If divestitures like the Watson Health sale continue, IBM's structure drifts toward shedding ambitious but unprofitable AI units rather than subsidizing them inside the core business.
The trend: Flagship AI programs are being judged by commercial results rather than demonstration wins, forcing vendors like IBM to shrink their ambitions or shed the businesses their hype created.