Algorithms everywhere: Can IBM automate business decisions?

Algorithms everywhere: Can IBM automate business decisions?

Summary: IBM outlined a vision---and of course a new services unit to go with it---that takes a little time to grok. Big Blue talked about the "information journey," fact-based enterprises and nudging out gut calls in everyday management for decisions based on hard, cold facts.

TOPICS: Security, Banking, CXO, IBM

IBM outlined a vision---and of course a new services unit to go with it---that takes a little time to grok. Big Blue talked about the "information journey," fact-based enterprises and nudging out gut calls in everyday management for decisions based on hard, cold facts. But when you boil it all down, Big Blue is talking about providing a bag of algorithms that will automate many of your business decisions.

Sitting through IBM's series of presentations Tuesday on how we'll all work for fact-based enterprises in the future left me with a few nagging questions (with a dearth of concrete answers): What's the role of intuition and gut calls in management? Are we all becoming quant black boxes of management? Can we see around corners for real? What's the prototype of a fact based enterprise? How long will this journey take and what does the end state look like?

The answers after a few rounds with IBM execs: Intuition will still matter, but for the bigger decisions. Some decisions will be automated (think call center manager who will look for a simple yes, no answer when deciding on whether to extend a warranty). Predictive modeling will be everything. And this journey will take a while and the blueprint will be industry specific (rest assured IBM will consult with you every step of the way).

"We're developing a bag of algorithms to be plugged together," said William Pulleyblank, chief of IBM's center for business optimization. Pulleyblank should know: He led the Blue Gene supercomputing effort. Remember Blue Gene? It might be your boss someday (only half kidding).

There's little question that IBM could hit a sweet spot for its consulting business. Simply put, risk management is on every manager's mind. Anecdotally, execs talk about risk management more. And I've seen it up close. Our risk management report on Goldman Sachs and webcast had a lot of interest, but it wasn't just the financial wonks on board---it was IT folks. Turns out everyone wants to see around the corner for the latest bogeyman. Risk management and mitigation was something that used to be tucked away in the corner. Today it's everyone's business.

In some respects, IBM's big pitch boils down to better data analytics. How can you take all that stuff you've been collecting and find some real intelligence? How can you account for systematic problems? Real-time reaction to customer needs? Altering pricing on the fly?

It would be easy to dismiss IBM's effort as another wrapper to sell software, hardware and services, but the vision is big and makes a lot of sense. However, I couldn't help but feel a little uneasy. Algorithms are partly responsible for this financial mess. Sometimes the black box fails miserably. And sometimes a little human intervention is needed. And the biggest worry: If managers wind up just looking to a screen to make a call---yes or no---doesn't that make us a business equivalent of a GPS slave where no one will be able to read a map in a decade. Thankfully, IBM's vision doesn't include some artificial intelligence theme (maybe next decade).

Meanwhile, IBM's Fred Balboni, head of the business analytics and optimizations services unit, says Big Blue will go easy on pitching the "intergalactic projects" that may spook clients. The game plan is to serve up valid business cases for being more analytic and deliver returns within a year. Over time, companies can become the so-called fact-based business.

Here's the big picture:

But are "applied semantics" better than "human insight?"

In Corporate America you can easily (predictively) model some cultural issues. CEOs will love this "fact-based" management, but the front line folks can resist. IBM notes the hurdles, but expects rapid adoption---at least something faster than the ERP revolution of the 1990s. Why? Younger folks already look at their PCs---and Google---as an answer machine, said Brenda Dietrich, vice president of business analytics & mathematical sciences at IBM Research.

This blur of data, computation and decision making looks great on paper. The reality is that data architecture is messy already and could use the clean up. Ask the Bill Eimicke, FDNY Deputy Commissioner. Eimicke got on this fact-based enterprise bandwagon in 2007 following the deaths of two firefighters as the Deutsche Bank building was being dismantled (it was heavily damaged on Sept. 11, 2001).

There were "flammable things" amid the demolition work. The problem: The FDNY didn't know there were flammable things because it didn't have access to the data from the plethora of city agencies that inspect buildings.

Eimicke, a Columbia University academic on loan to the FDNY, said that the data was there, but not in a form that was usable. His project: Aggregate all of the data on buildings in New York so it can prioritize inspections. Maybe a few extra facts will save another Deutsche Bank building from happening. "That crisis triggered our project," said Eimicke.

Indeed, firefighting is an obvious fact based business. Financial services, health care and retail are other obvious verticals ripe for some algorithm love. Financial services firms are the farthest along on this algorithm utopia.

What can go wrong?

There are multiple industries that could benefit from a little fact-based decision making, but there are landmines ahead. When I asked Pulleyblank he rattled off a few items. Given more time, a top 10 list would have emerged.

Here are Pulleyblank’s landmines ahead:

  • Data quality: Any automated process is only as good as the data being used. If the data has errors in it IBM's algorithms won't work as well. The solution will be better filtering to diminish "noisy data," says Pulleyblank. It's not like a company is going to go back and clean 30 years of data.
  • Risk management techniques: This entire concept of managing systemic risk---and determining everything that could go wrong---is young.
  • The need for real-time reactions: Are companies ready to respond and make decisions in real-time?
  • An unexpected shift in the world. Pulleyblank acknowledges that predictive modeling only goes so far. What are the Black Swans ahead?
  • A spectacular failure. "We can't afford a spectacular failure," said Pulleyblank. What would be a spectacular failure? Try a major Northeast blackout where the algorithms in the smart grid are at fault.

It's a bit early for those concerns, but it is something to think about. If IBM's vision plays out algorithm management will be a key component of businesses everywhere.

Topics: Security, Banking, CXO, IBM

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • IBM made the same claim 25 years ago with MRP

    MRP (Material Resource Planning) and MRP II were going to solve all the problems with inventory in manufacturing. Not only would it track and order materials for a manufacture, it would do so for all the upstream and downstream companies. (Your vendors and your customers)

    There are plenty of horror stories of companies that almost went bankrupt trying to make it work.
    • Isn't MRP what all retailers and manufacturers do today? nt

      Isn't MRP what all retailers and manufacturers do today? nt
      • In very limited ways.

        Yes, its still around and is used but... The smart companies use it as a "indicator" of what to order when but the decision is made by a human being. That's a far cry from what IBM tried to say it would and could do.

        I'll try to give an example. Lets say your company makes and sells bread. In a perfect world MRP would look at demand in the stores and seasonal trends. (Gathered from their MRP system.) This would generate a forcast for you and schedule production as well as ordering your raw materials and packaging. That in turn would tell your vendors (who also use MRP) what they must produce and order their materials automatically. Lets say the vendor makes the plastic bags for bread. They in turn would automatically order plastic resin to make the bags. That in turn would generate an order to the resin folks, on and on and on...

        The problem of course was (is) that its far to complex a system and a snag anywhere in the loop brought every thing to either a stop, or worse, way over order materials.

        So yes, MRP is used as a forcasting tool but I know of no one that actually allows it to place purchace orders automatically and instead leave those decisions to human beings.
      • By the way...

        IBM tried for three years to make MRP work for their own manufacturing and finally gave up.
      • How many examples would you like?

        How many examples would you like? Aerospace industries would collapse wihout MRP / ERP. WallMart without Supply Chain Management???
        David Gale
        • MRP and Aerospace

          I disagree that MRP is useful to Aerospace(at least with the aerospace company I work for). The majority of our orders are custom or do not repeat in a predictable fashion, and the volumes are relatively low.

          Our past orders are generally a poor predictor of our future orders. One year we may have an order for 1000 units of a specific variety, and the next year it will be a completely different unit, with almost no orders for the first one. The result of this, of course, is massive lead times. I don't really see how you could solve this with algorithms, as most of the variation is a result of unpredictable one time decisions(E.G. the government approving a new satellite program), rather then predictable events occurring at a predictable frequency(E.G. the christmas retail sales boom).
    • Saying "Never"

      Yeah, I agree that business decisions can't be automated. But I guess I would like to hedge any bets. Recalling my M.S. C.S. days (mine is thesis #14 at U of Maryland), computers could "never" play chess or translate between languages. If I recall correctly, there was even a theoretical basis, as well as the experiential basis, for those conclusions.

      I think it was Niels Bohr who said predictions are hard, particularly when they involve the future.
    • And Some Companies Are Still Weighed Down By It

      Like so many powerful tools, MRP was often selected by people who did not and do not understand the business in which they are employed. It is great for producing many millions of slightly different widgets. It is deterimental to businesses making truly custom, one-off products.
      It was one of the causes of a debilitating inefficiency in a company where I used to work. Frankly speaking, the less intellect people had, the more whole-heartedly they embraced the system. It gave them power over the more clever, knowledgeable and experienced people who rightly resisted.
      I envision the same sort of disasters occuring, but on a much grander scale, once algorithms replace true inductive and intuitive decision-making.
      The only potential benefit would be that CEOs and their breathren can only justify measely salaries for following what their purchased model tells thenm to do.
  • Eating their own dog food

    IBM has to be its own first customer and have one or more impartial observers evaluate the software and its decisions. If IBM won't use it, should anybody else?
  • RE: Algorithms everywhere: Can IBM automate business decisions?

    Ibm is a decent company but it does it by outsourcing... a very very large part of IBM is outsourced labor even its creativity is often subcontracted.

    Its a shame that IBM isnt the IBM of the past, when they were american people inventing the technologies and methods.

    There best stuff nowadays is from India & Argentina and China.
  • Why this is better than intuition.

    Today, huge companies are basing decisions on surprisingly little real data.

    IT collects vast amounts of data and unable to do anything with it.

    Call it, LINO, Lots In, Nothing Out. For any number of reasons data is collected, and nothing is done with it. While there is an art to running a business, most of business is science, not art. IBM's approach could bring a higher degree of consistency and repeatability to business processes.
  • IBM need to re-think their software stack

    It is all possible but needs a greatly simplified and unified approach to software that reflects how people work where source information is created. IBM's historical strategy of buying old disconnected software components will make this a huge challenge - in fact that challenge starts by the very talk of this vision - hold off buying their old technology?
    We are close to the intelligent process that recognises past actions and dynamically changes to reflect new circumstances. It is already possible to have basic automated decision making by applying integrated process, rules, state and calculations - so nothing radically new - as ever IBM follows the innovators?

    David Chassels
  • RE: Algorithms everywhere: Can IBM automate business decisions?

    If you program decisions, those smart enough to take advantage of the data constraints that you place on your decision making will reap the oportunity.Intuition (human computation) is not bound by quantification, and contains emotions like greed (like game playing to win at all costs)that seeks advantages outside the rules.

    Sounds like a socialist plot to eliminate greed in the market place and free thought in the masses.
    Talk'n trash
  • What would a major failure look like?

    How about a major financial meltdown where the algorithms used by lenders pretty much said that any mortgage was a good deal because property values would go up faster than people could possibly default?
  • RE: Algorithms everywhere: Can IBM automate business decisions?

    Hi -

    There is some merit here, but engineering people out of the equation always delivers enormous problems.

    There is a fundemental social reorientation of the enterprise underway. "Algorithm management" is a contribution insofar as it augements collective intelligence and enterprise information markets. See:


  • RE: Algorithms everywhere: Can IBM automate business decisions?

    The one-word answer to the title question posed: "NO."

    But not because the technology does not exist or cannot be brought to bear on the business issues and considerations surrounding decisions. It cannot be done because it will never be accepted by the individuals and organizations which might otherwise have the most to gain.

    Regardless of whether it makes for better decisions, the lack of human touch inherent in "automated" processes makes people uncomfortable at best, irrelevant at worst.

    We humans tend to fear what we do not understand. Reducing decision making to its component process steps; using words like "algorithm" in this context; suggesting that a computer program can improve human decision making... these are inherently misunderstood - and thus feared - by the majority of business managers.
  • I can't believe anyone actually hires ibm...

    I realize their products are nigh on impossible to implement without help, but everything from their operating system to their website reeks of half-assedness. One day, they might discover sort functionality, but I don't think that will happen until after they dump RPG, rational, the cobol core, and make websphere a decent product out of the box.

    Just to clarify any confusion, none of that will ever happen.
  • RE: Algorithms everywhere: Can IBM automate business decisions?

    Can business decisions be automated? Mostly. Can IBM automate business decisions? Probably not. Hopefully, however, IBM will provide the methoda and tools by which business people can create autonomy of much higher order than the mess of IF:THEN with no ELSE's that current formulas for business process managment has created.
    The key step will be to get beyond the decision vs. intuition view and learn how to combine and harmonize both.
  • Algorithms everywhere

    This will will be useful in giving suggested answers based on preset parameters, but humans will have to make the actual decision. Business decisions often have to take into account one time data points which can have a dramatic effect on the results; at the CEO level and at the call center level.
    GM Fedorchuk
  • RE: Algorithms everywhere: Can IBM automate business decisions?

    see "Smart (Enough) Systems" by James Taylor.

    This is not about automating decisions made by people, it is about automating decisions that don't need people. Dig deeper, and this is about Business Rules. In situations where all responses to a question can be defined depending on data about the situation, having people pick an answer is a waste. Biggest example? Credit Checks and your Credit Score, and 'deciding' to approve a request for credit. This is being done thousands/millions/whatever times per hour/day, and there not enough people available to do this.

    And this is never going to be about Artificial Intelligence. Smarter systems are made of the same stuff as 'dumb' systems, they just know more, enough to make a decision based on facts. If the facts aren't available, the system is not going to 'think' or 'reason' its way to a decision, its just going to ask for the facts, please.