That ultimately depends on the agreement between you and your LSP client. However, it is my unabashed opinion that you should reject hourly rates on any task that can be properly measured by output.
When clients pay by the hour, they invariably set a minimum output you must achieve—but they refuse to reward you for working more efficiently. This creates a fundamental trust gap. Without controls, they must trust you to report your hours honestly, even while financially disincentivizing you to do so.
The industry’s solution to this gap is toxic: big brother surveillance tactics built directly into their platforms. As translators inevitably look for ways to survive and game the system, the tools find new ways to close loopholes and tighten the leash (even building in ways to keep you from using your own tools that would help you deliver better, as if you’d want to when your hourly rate is capped). This exhausting arms race takes the focus off the actual translation, insults the dignity of the translator, and fosters a hostile working environment. As a result, output quality is measured in minimum standards, not excellence. Further, it gives the LSP a second lever to arbitrarily announce that their automation efficiencies have moved the goalposts, demanding even more output for the exact same hourly rate.
And for all these reasons, the amount clients are willing to pay hourly is always less than what an efficiency- and quality-driven translator can earn on a per-text unit basis. The hourly model actively drives top-tier translators out of the market, leaving only mediocre time-card-punchers and system gamers enthusiastic about staying. This entirely saps the talent pool for conventional MTPE, making automation the sole driving factor, which further accelerates these negative outcomes in a vicious cycle.
No wonder translators hate conventional MTPE so much! Such an approach will never lead to the excellence and client-vendor trust the GAIT-Augmented MTPE workflow is designed to achieve. At the end of the day, who cares how long the job took? Time is an input; the end client is buying an output. They are paying for a high-quality product measured by the amount delivered.
That is why the GAIT-Augmented MTPE workflow is designed to operate on a per-text unit basis, such as per word or per character. You, the post-editor, are responsible for delivering good work, but the boutique LSP gives you the discretion to choose your tools and work however you wish to achieve it. The only question that matters at delivery is whether the work meets the specifications. Nobody asks questions about time—an irrelevant metric that cannot be tracked without surveillance. It makes sense that our industry has mainly worked on a per-text unit basis for decades, and there is no reason to change now.
In terms of the actual rate, the rough target is 1/3 of your standard full-human rate, which has traditionally been the proportion allocated for human revision tasks. In this model, the backend technical processes (including MT processing) account for a portion, leaving the end client with an attractive savings, likely somewhere around 50% off full quality.
Before you panic at the idea of working for a third of your normal rate on a post-editing project, consider two things. First, the baseline is better. In the past, LSPs would often send jobs to relatively cheap human translators (yesterday’s “machine translation”) and then go to high-quality revisers to fix that for around a 1/3 rate. The GAIT-Augmented MT you receive from me should be better than what you would have gotten from a mediocre human translator. Second, the goal is pragmatic. Remember the mindset shift: post-editing is not intended to deliver your version of stylistic perfection. Your mandate is to deliver a clear, fluent, and correct translation. You are not expected to fine-tune every single sentence.
Because you are starting from a contextualized baseline and editing for pragmatism, you should be able to move through most GAIT-Augmented MTPE tasks at approximately 3X the speed you used to achieve in full-human translation pre-AI. You hit your necessary hourly earnings; you just produce more volume to get there. And if we work together, we can push those MTPE volumes up and help everyone.
