The last holiday season was replete with the sound of ads for Amazon Echo, Apple Home, Google Home, and Microsoft Cortana. All these products promise their owners the ability to speak aloud and have the receiving device trigger specific actions such as turning on the heat or reading the results of a web search. These devices are still buggy but improvements in natural language recognition performance and artificial intelligence-driven mechanisms portend a near-term future where voice commands can trigger all manner of multi-stage activities.
What will enable the next generation of voice-driven personal assistant “bots” is the proliferation of APIs that allow secure access to transactional database services. The AI converts a voice request into API calls that perform very complex queries quickly and include purchase functionality. The simplest example often cited is the one-button “buy again” mechanism for things like printer cartridges and laundry detergent often used as an example of how the Internet of Things can make a consumer’s life easier.
In this simple case, the purchasing mechanism is configured on the Internet or via a downloaded app’s preferences and then triggered by the device. The next stage of these bots’ evolution will be to include more purchase decision-making steps in cases where the goods and services offered are more variable.
For instance, you will soon be able to send a command to your personal assistant to “Book me on a flight to Bangalore on April 23rd” that communicates with your pre-configured travel bot, which will contain frequent flyer preferences and general flight preferences such as window seat or no more than one layover. The bot will make a tentative reservation via an air carrier’s booking API, send an alert to get a “yes” or “no” to the chosen flight, and then either buy it or try again.
As convenient as these new personal services would be, there are much larger opportunities for B2B supply chains and enterprise resource planning (ERP) services that incorporate similar automated alert and purchase mechanisms. Algorithms or routines that can automatically assemble an order for industrial supplies based on historical seasonal production data and contractual parameters could be built fairly easily, allowing managers to place an order in advance to take advantage of volume discounts or early-bird pricing and ensure just-in-time arrival of raw materials.
Advanced custom mechanisms for these sorts of things have been built for years but the commoditization of AI, the proliferation of APIs for shipping, travel, and other services along with the increasing sophistication of calendar and communication apps will make the next steps toward automation far faster, cheaper, and easier to implement. When we turn this corner, vast swathes of support positions at organizations of all kinds will be permanently displaced. Meanwhile, the job of those who build and maintain information services will be a critical one to the inevitable automation of complex B2B workflows.