Our research has long shown that demand for translation continues to grow at a fast clip. On the other hand, stagnant translator productivity endangers the supply of high-quality translation services. To bridge this demand-supply mismatch, we recommended that buyers and suppliers of translation invest in the human workforce, improve processes, and use automation more intelligently, especially machine translation (MT).
 
What form does “improving processes” actually take in real life? We recently spoke with Jonathan Kirk, chairman of Elanex, a U.S.-based language services provider (LSP), who told us about his solution to the problem – automate the management of translation workflow to speed up execution by an order of magnitude. Beyond accelerating the process, he wants to provide an Amazon-like experience for translation buyers, what we call “transactional translation.” That means that the terms and conditions are clearly laid out, the flow of control is intuitive, and you don’t haggle about the price – but you may choose various delivery options. In most cases, no human project managers are involved. However, flesh-and-blood translators still produce the translation.
 
Of course, buying translation is different from picking up the latest bestseller online. It’s much more like configuring a PC at Dell.com. In fact, Kirk says that even a simple translation job can involve up to 130 steps, many of which typically require costly human touches such as e-mailing the client, contacting potential translators, and preparing the files. For these tasks, Kirk has applied artificial intelligence (AI) algorithms to codify human knowledge.
 
The process knowledge that Elanex coded into expressIt, the company’s recently debuted translation portal, comes from observing the project managers (PM) who shepherd jobs through a workflow. From reviewing nearly a decade of projects, Kirk uncovered substantial blocks of downtime as PMs orchestrate a series of synchronous interactions with and patiently wait for responses from the client, then from a staff member who categorizes the job by subject matter. Next the PM contacts potential translators and editors who might be qualified to do the job, then the translators who actually do the work, an editor, the client again, and sometimes the translators and editor again. Throughout, the PM identifies the availability of human and technical assets to aid in the translation.
 
PMs no longer initiate and shepherd a job to its delivery, but are called in only when expressIt’s expert system logic senses an exception. We asked Kirk how long it took to reach the point where problems requiring human intervention no longer regularly stopped the system. Starting from the purpose-built software that powers the company’s translation business, Kirk and his team spent a good part of the last two years reviewing nearly a decade of human-managed projects and “boiling out” the human PM interactions. The promise is shortening the turnaround time by 90%, all from automating steps formerly performed by the PM.

Kirk admitted that Elanex is cautiously introducing expressIt to the market. While the portal’s workflow engineered ultimately to deliver translations in as little as an hour, the company will initially commit to turnaround times of a few hours for anything up to a few hundred words. Today’s version limits document formats to Office, plain text and HTML; project size to 1,000 words; and whichever languages can be supported by the translators who happen to be online and are qualified to work with the document’s subject matter. Of course, Elanex will also accept bigger jobs with other document types for any of its language pairs through its normal workflow process.

The bottom line is that expressIt promises to address translation volume and velocity issues while delivering the levels of quality and domain specificity that buyers demand. We are interested in how other language service providers and in-house translation departments are dealing with the translation demand-supply mismatch. E-mail us with your innovative or pragmatic solution. We also ask freelance and in-house translators to let us know what it’s like to work with these systems.