X

Our Analysts' Insights

01Aug

Planning for the Onslaught of Articifical Intelligence

In Ernest Hemingway's novel The Sun Also Rises, Bill asks, "How did you go bankrupt?" Mike answers, "Gradually and then suddenly." Both buyers and suppliers of language services and technology have had a similar experience with the evolution of artificial intelligence. After decades of science-fiction depictions of AI and fits and stops in the actual science, the last few years have seen a rapid move toward incorporating artificial intelligence in a broad range of processes and products.

As a result, artificial intelligence has become a regular topic in many of our discussions with language service buyers and suppliers. Both groups invariably ask about the existential threat that AI poses to LSPs. After that, conversations quickly devolve into details about particular technologies, where AI might best be used, and who's using it best. Here are a few issues that we regularly consider in our advisory sessions with language service providers. Many of the same issues come up when we talk to enterprise planners who buy from these LSPs.

  • Just what is AI? LSPs often intermix terms in a way that recalls a truism that we've heard in academia and in Silicon Valley: If you're trying to raise money, it's AI. When you're hiring developers, it becomes machine learning. But when you're developing a product or an internal agent, it's all about the data. In short, AI is the buzzword. Machine learning is the technology. Predictive analytics is the art or science. 

    What it means: Be clear about what you're trying to accomplish. Machine learning experts don't come cheap, so begin training engineers you have on hand. A decade of mainstream investment in big-data technology like analytics and visualization combines with massive volumes of linguistic, process, and supplier performance data collected by LSPs in the course of their everyday work. That endows LSPs with statistics to feed into machine learning frameworksand thus creates a starting point that might allow them to skip some expensive hires. 
     
  • Will AI cannibalize revenue in the industry? Yes, some projects will go to neural MT (NMT) and buyers will reduce total spend – mostly at the top of the buying pyramid. Big-spend buyers look to do more with less, but that is not how or where the conversation ends. Rather, lower cost opens up new content to translate through improved ROI. In the past, growing companies have not used lower costs to reduce total spend, but rather to justify a bigger spend. Neural MT and "lights-out" project management powered by machine learning will be no different. 

    What it means: That shift of spending is a new opportunity for LSPs to grow. Large players will see some revenue loss at the program level. Medium players may see some account loss when customers shift vendors. Small companies serving big LSPs should expect some loss of revenue unless they have a unique offering.
     
  • Will there be winners and losers from AI? In the short term, large players with R&D budgets bigger than some competitors' entire revenue streams can be fast movers, gaining market share through efficiencies and lower prices. So far, most vendors use advanced rule-based systems to capture more margin rather than drop prices. This allows them to pour more money into R&D – or to take more money out of the business. We already see AI – broadly defined – appearing in TMS solutions, such as Memsource and SmartCAT, to handle discrete tasks in the process. 

    What it means: As both NMT and lights-out project management get mainstreamed and become more competitive, the pricing war could become hazardous to unprepared LSPs. However, the window of disruption keeps getting shorter as the pace of innovation increases. Companies that invest too much, too fast, will be at risk as state-of-the-art commercial tools become available at a fraction of the cost of custom development.
     
  • How much can I gain from investing in automation now? Democratization of AI will arrive through commercial off-the-shelf (COTS) solutions, bringing NMT and AI-based project management to medium-sized and then small LSPs that don't have the cash flow to hire specialists to build it themselves. Small innovative LSPs will use advanced forms of automation to rapidly gain market share, thus becoming M&A targets for bigger companies. 

    What it means: If your company has a legitimate opportunity to innovate, don't wait to do it. Decide whether your strategy is to: 1) lower prices to gain market share; 2) increase margin to fund greater investment or return; or 3) position your company for M&A. In 2018, AI is still an innovator's game but by 2019 or 2020 it will already be a fast-follower pursuit. If your company does not have a credible chance of making fast gains in the next 12 to 24 months, rather than spending on development instead prepare for the arrival of AI in COTS solutions by formalizing how you record project data.
     
  • Will my current language software vendors add AI components? Yes, they will. TMS suppliers such as XTRF are already adding core capabilities to create rules, based on patterns of behavior the system observes. More pre-configured options that require less know-how will come later, but the move is already underway. You won't have to be an AI expert – machine learning will come to your software stack from multiple products and integration-ready cloud services. In short, AI will show up in the market as wizards that configure machine learning. 

    What it means: If your current technology supplier isn't investing in AI that will make your workflows and linguistic processing simpler and less human-intensive, ask them what they're waiting for. AI may seem that it's coming slowly to a sector or a software category, but once it does arrive, it will seem very sudden. Meanwhile, get your data ready.

Whatever course of action you take, keep your eye on this space to avoid being surprised by new developments – CSA Research analysts are closely following advances in the field. AI will provide real efficiency gains and opportunities for savvy LSPs to differentiate themselves and deliver superior services to clients.

About the Author

Donald A. DePalma

Donald A. DePalma

Chief Research Officer

Focuses on market trends, business models, and business strategy

Related

Building a Comprehensive View of Machine Translation’s Potential

Building a Comprehensive View of Machine Translation’s Potential

It is no secret that machine translation (MT) has gone from a relatively niche solution to seeming u...

Read More >
Augmenting Human Translator Performance

Augmenting Human Translator Performance

In the first episode of an iconic sci-fi television series, a NASA test pilot was seriously injured ...

Read More >
Making the Best of a Bad Year: Five Lessons for 2021

Making the Best of a Bad Year: Five Lessons for 2021

As we look back at the annus horribilis that was 2020, what are some things we can learn and take fo...

Read More >
Microsoft Custom Translator: A Big Step Forward

Microsoft Custom Translator: A Big Step Forward

In late 2015 most developers still treated Neural Machine Translation (NMT) as a future technology t...

Read More >
Global BPO to Buy Lionbridge Data Annotation Business

Global BPO to Buy Lionbridge Data Annotation Business

TELUS International announced that it would buy Lionbridge’s artificial intelligence business unit ...

Read More >
TMS Is Dead. Long Live TMS.

TMS Is Dead. Long Live TMS.

A recent examination of how computing power has changed over time calculated how much it would cost ...

Read More >

Subscribe

Name

Categories

Follow Us on Twitter