PMs don’t get paid to transcribe requests. They get paid to translate meaning.
That distinction clicked for me while talking with Andrew Saxe (VP of Product at Smartling, the enterprise translation platform) during his interview on The Way of Product Podcast.
Andrew mentioned they see thousands of feature requests. He also shared that on a few occasions they built very specific requests for big customers—only to watch those features go unused.
That’s what happens when we transcribe. We copy a sentence from the customer’s language into our roadmap’s language. It looks faithful. It often misses the point.
Translation is different. In language, you don’t carry over every word—you carry over the meaning. Think about the Spanish phrase “échale ganas.” Transcribed, it reads “throw desire.” Translated, it means “try your best!” The words change; the meaning lands.
In product, you don’t ship the exact sentence a stakeholder spoke—you ship the solution that relieves the underlying pain.
So you slow down and look at the moment. What were they trying to get done? Where did the workflow, policy, or constraint bite them? Then you step back and look for echoes across accounts—different words, same friction. When the words change and the pain repeats, you’ve found the thing to build.
Andrew’s team aims to “manage by exception, automate the rest, and guard against obvious failure modes."
That spirit applies here too. Translate the signal, not the syntax. Ship the exception-handling that solves the real problem, even when it looks nothing like the original request.
Bad PMs transcribe meetings into tickets. Great PMs translate meetings into specs that work.
If your last spec reads like a quote, you probably transcribed. If it reads like an answer, you probably translated.
Listen to The Way of Product: Apple Podcasts or Spotify
Declining MAU can be a product win,
depending on your product strategy...
Andrew Saxe, VP of Product at Smartling, joined me to unpack why.
When a system translates intent well and runs in the background, people don’t have to log in—and the work still gets done.
We discussed designing for “manage by exception,” placing guardrails where failure would be costly, and allowing automation to carry the load.
We looked at the feature request funnel too: why thousands of asks don’t equal thousands of tickets, and how building a few verbatim specs for large customers can still lead to unused features, tech debt, and wasted opportunities.
Listen to The Way of Product: Apple Podcasts or Spotify
Timestamps:
02:48 The niche translation industry and Smartling’s impact
11:28 The role of AI in translation
13:55 AI’s impact on Smartling’s product roadmap
16:38 Managing translations with AI and human oversight
22:00 Optimizing translation services for market expansion
22:47 Challenges and nuances of multilingual customer research
23:41 The art and science of requirements to spec translation
29:18 Feature requests and customer feedback
31:26 The importance of understanding user needs
34:27 The future of translation and AI
38:37 Concluding thoughts and contact information
24:56 AI and Translation: Balancing Automation and Human Touch
28:19 Feature Requests and Customer Feedback
30:27 The Importance of Understanding User Needs
35:42 Managing Translation Budgets and Quality
37:39 Conclusion and Final Thoughts
