Aug 12, 2016

    'Elementary, my dear Watson, we don't have cash to pay IBM'


    INTERNATIONAL Business Machines (IBM) is in an unusual fix in telling big United States banks that they can use its Watson software of Jeopardy-winning fame as a cost-saving solution.

    Bankers said they like it but cannot afford it.

    Banks are in the fifth year of belt-tightening campaigns that began in 2011, chasing billions of dollars' worth of savings.

    And vendors that offer everything from technology to janitorial services are getting squeezed.

    With persistently low interest rates hurting revenue and businesses like bond trading hemmed in by new regulations, few on Wall Street expect the austerity to end any time soon.

    For IBM, the irony lies in the fact that senior bank executives believe its artificial intelligence software could help them achieve cost-cutting goals in coming years, but are not ready to pay for Watson today.

    Several technology executives from large banks told Reuters that while the software may have enormous potential, they would struggle to convince top managers, laser-focused on quarterly results, to sign off on investments that do not offer an immediate payoff.

    People familiar with the matter said IBM had offered presentations of its software to banks including Bank of America, Barclays and Morgan Stanley.

    The software, known as Watson, got attention in 2011 for winning the game show Jeopardy.

    IBM said it can learn and process human language, and analyse large amounts of unstructured data, like social media posts and digital photos.

    IBM's pitch is that Watson can do everything from answering customers' questions in retail branches to detecting credit-card fraud to helping wealth managers make better recommendations to clients.

    Bank technology executives said the minimum cost of using software like Watson, including due diligence and training, can reach a few million dollars.

    However, a spokesman for IBM noted that companies can develop their own applications using Watson's underlying code - if they do not want to pay for a full-scale implementation.