The challenge has been choosing between the quick fix of 'plug and play' rather than a systemic approach which may be more disruptive and expensive to the organization and its processes, but that provides a more profitable and productive solution.
Artificial intelligence is no different, in this regard, than its technological predecessors. JL
Sam Ransbotham reports in MIT Sloan Management Review:
Most organizations are still in the early stages of AI implementation. An imminent reality is that AI is agnostic and can benefit both good and bad processes. An insidious risk is that AI gives new life to poorly conceived processes. The danger is when AI gets just good enough to let a less-than-ideal system limp on. No amount of gee-whiz AI used to improve a process beats not having to rely on that process at all.
Scenarios describing the potential for artificial intelligence (AI) seem to gravitate toward hyperbole. In wonderful scenarios, AI enables nirvanas of instant optimal processes and prescient humanoids. In doomsday scenarios, algorithms go rogue and humans are superfluous, at best, and, at worst, subservient to the new silicon masters.
However, both of these scenarios require a sophistication that, at least right now, seems far away. Our recent research indicates that most organizations are still in the early stages of AI implementation and nowhere near either of these outcomes.
A more imminent reality is that AI is agnostic and can benefit both good and bad processes. As such, a less dramatic but perhaps more insidious risk than the doomsday scenario is that AI gives new life to clunky or otherwise poorly conceived processes.
Consider faxes in the health care industry. Despite being obsolete in most places, “like the floppy disk or the CD player,” faxes are still a fundamental part of the medical infrastructure. Because of the long history and strong network effects, the medical industry still sends a staggering number of faxes every day.
Invented in the 1840s, well before the telephone, faxes illustrate how difficult it is to change an entrenched process.
Despite widespread use, sending faxes is, for the most part, a horrific way to transfer information. The process is typically (1) extracting and printing information from a computer system, (2) scanning it into an image, and (3) transmitting it somewhere else via fax. The burden rests squarely on the recipient to interpret a pixelated approximation of the original information. Structured digital information has become unstructured. With every fax, a data scientist gets their wings ripped off.
The conversion from structured information to unstructured and back is a waste. No one wins — a patient may be waiting for approval, medical staff may lack information, errors can creep in, etc. At a minimum, time and effort are needlessly spent.
Advances in AI and image processing are making significant progress in reducing this problem. Organizations can use AI to recognize images, automate the interpretation, and restructure the information. Certainly, this is a welcome improvement. No one gets competitive advantage or business value from wasted time spent restructuring information.
However, this improved image processing is a pyrrhic victory. While it may be more efficient, perhaps even significantly, the expended resources are lost forever. In the absolute best case, the original structured information is recovered, but it will be expensive and difficult to get close to that best case. The realistic case is far less promising.
The danger is when AI gets just good enough to let a less-than-ideal system like this limp on. Improved image processing can allow organizations to cut costs and automate much of the processing (this applies beyond faxes and beyond health care), but AI may also mask the symptoms of bad processes. It provides a bandage, not a cure.
If AI were to fail completely at interpreting the faxes, we might be better off: The system would be untenable. Costs would rise. People would notice. Change would be inevitable.
Yes, it is better to reduce this annoying work than to continue doing it. This is a justifiable reason to apply AI in organizations and a great example of reducing the scut work. However, gains in AI applied to bad processes may staunch wounds, not heal the organization.Successful organizations will continue to improve underlying processes. But, in many industries, like health care, with legacy systems and embedded processes that involve many people and many organizations, that will be difficult.
One risk we see is that upstart organizations, unencumbered by legacy processes, will be able to start fresh. They won’t have to make significant investments of time and resources to apply AI to processes that perhaps should not exist in the first place. These new entrants, possibly from outside your industry, will apply other tools and technology to completely bypass the process your organization is struggling to bandage with AI.
Are there processes in your organization that AI can improve? Great. But, be sure to ask yourself: Should these processes exist in the first place? No amount of gee-whiz AI used to improve a process beats not having to rely on that process at all.
0 comments:
Post a Comment