I left the recent Boston Product Management Association (BPMA) and Lokalise event thinking about Data from Star Trek.
An extraordinarily capable android, curious about people, still working to understand the subtleties of human life. Humor. Social cues. The distance between what someone says and what they mean.
That feels like a useful way into the conversation about AI agents, and the human work that makes them useful.
At the event, Sahil Gambhir told a story about interpreting nods in Bulgaria, where the gesture could mean the opposite of what he expected. His humility made the story stick. Living across countries had given him experience and plenty of opportunities to find out what he didn’t know.
If people can misunderstand something as apparently simple as a nod, think about what we leave unstated when we ask a machine to communicate for us.
Giving the agent somewhere to go
There is a technical development behind this conversation: MCP, the Model Context Protocol. It gives AI applications a standard way to connect to tools and information outside the model. An agent can use those connections to get context and carry out actions.
In the Data analogy, the agent is our capable crew member; MCP provides a connection to the ship’s systems. Those are different parts of the picture.
Lokalise’s MCP server makes the idea concrete. An assistant can manage localization tasks, work with glossary terms, and attach screenshots that show where words appear. A person can describe an outcome in ordinary language and have the assistant work through connected tools.
That is a meaningful change in what the technology can do. It also makes the quality of our direction more consequential.
The human knowledge behind the prompt
Lokalise calls its broader approach Language-Market Fit. The white paper describes content that fits the audience, market, channel, context, and business purpose. It proposes turning human knowledge into reusable assets and using review and performance feedback to improve them.
That is a useful challenge for writers and marketers. We have to make what we know explicit enough for someone, or something, else to use.
“Make it friendly” leaves a lot to chance. Is the person excited, skeptical, or already frustrated? What do they need to understand? Which references will feel familiar? Is a joke welcome here?
Good writing brings those questions into focus. It identifies the audience, chooses the point, and explains what matters. Creativity helps us see a possibility that wasn’t in the original request. Cultural knowledge helps us recognize when an apparently reasonable choice could land badly.
AI can generate humor and adapt tone. The work for us is to establish what is appropriate in this relationship and assess whether the result connects. Access to more tools does not settle that question.
The value of paying attention
Ennio Carboni’s discussion of executive engagement and business value stayed with me too. He was sharing his personal perspective, rather than speaking officially for AWS. It made me think about how much useful knowledge begins in a conversation and never reaches the brief.
Someone uses an unexpected phrase. A customer hesitates. An executive explains a priority in a way the organization’s own messaging never has. A person laughs, and the reason matters more than the words of the joke.
Ennio also made a practical point in the recorded conversation: pivots can be expensive. His preference for doing the groundwork before making a decision is a useful reminder for anyone tempted to equate speed with progress.
These are observations we can bring into the work. We can record them, test our interpretation, and use them to improve the direction we give our systems.
Lokalise’s approach also makes an important point about scale: human expertise can be built into guidance so people do less repetitive correction. The opportunity is to put our attention where it changes the result.
Help the capable crew member
For me, the Data analogy is about collaboration and curiosity. We can connect an agent to more systems and ask it to do increasingly useful things. We also have to help it understand what matters to the people those actions will affect.
That takes listening, clear writing, and the willingness to question our assumptions. Sahil’s nodding story is a good place to start: what seems obvious to us may be exactly what needs explaining.
The more capable our tools become, the more valuable it is to know what we mean, and to be able to say it well.