German philosopher Ludwig Wittgenstein argued that the meaning of language isn’t fixed by abstract definitions but by how words are actually used in everyday life. In his early work, he saw language as a logical system that mirrors reality. Later he shifted, claiming that meaning comes from “language games” - shared social practices where words gain meaning through use and context.

In the AI era, these ideas have become more relevant than ever. For organizations who collaborate to create value, the meaning of words depends on their context. The word "application" for instance has completely differentt meaning in HR vs IT. Organizations that achieve the most internal clarity on their internal "language games" stand to benefit from AI productivity gains the most, since effective use of LLMs is entirely dependent on truthful and reliable interpretation of words.

Plot starts from a couple of simple, familiar ideas:

- get your act together - it takes all sorts to make a world - genius doesn't scale - a picture is worth a thousand words - what you see is what you get - think before you act - choose your words carefully - keep your story straight - those who cannot remember the past are condemned to repeat it

Unfortunately, today's AI practice generally does not abide by these rules. It is largely individual: one person prompting in isolation, acting as if they have complete context, even though real work generally depends on multiple perspectives. It’s text-only: everything gets flattened into prompts and responses, losing the structure, shapes, colors, and relationships that help people actually understand and design complex systems. It’s scattered: knowledge lives across individual chat histories, personal tools, and loose sessions with no coherent intent tying it together. And it’s ephemeral: insights appear in a conversation and then vanish, with little memory, reuse, or accumulation—exactly the opposite of “remembering the past.” The result is dissolving workflows that rely on the clever ideas of individuals and on coincidence rather than strategically aligned teams and orchestrated agenda's. Consistency, predictability, shared understanding, and durable, compounding knowledge are the exception rather than the norm. In the end, the only parties that truly benefit from this arrangement are the AI vendors, because the company's own knowledge never hardens into reusable structure. This hurts strategy, teamwork, planning, repeatability, and is ultimately very costly and ineffective. Tokens are not getting any cheaper anytime soon, but regardless, spending them effectively is rapidly becoming an ever more significant competitive advantage.

Large language models can absolutely achieve massive productivity gains, but they need to be treated like the programmable computers that they are, rather than the friendly, supportive companions they have been programmed to impersonate. But an AI chatbot's primary objective is not to be helpful, but to keep talking no matter what. They will confidently say things that are outright nonsense, because that is inherent to how they work. They have no understanding, no morality or judgement, no memory, all they do is continually calculate the most likely next letter to respond with based on the patterns they have seen before. This is why individual chat sessions are a hugely ineffective way to program an LLM. But chatting is in fact programming, and when organizations decide to hand out Pro AI subscriptions to large parts of their staff, they quietly assume that suddenly everyone has become a programmer, which is obviously not the case. Prompting is a skill in itself, and collective, systematic prompting like Plot enables yields dramatically better results.

To be used effectively, LLMs need clear boundaries, curated context, and deliberate, complete instructions that leave no room for doubt or interpretation. And they should execute against a system that people understand, shape, and govern together, not force the people into scattered, individual expression formats that are entirely unsuitable for collaboration. Plot was built around these beliefs. It enables the real, human 'mixture of experts' that organizations are to discuss and capture ideas collectively, visually, and in real time. From that shared understanding, Plot can generate systematic LLM instructions that encode company policy, strategy, processes, data models, technology decisions, and broader context in a controlled way. Plot's LLM instructions sharply reduce the room for assumption, fabrication and hallucination because the model is no longer improvising against fragments, but sees the whole intent and context at once in a form it can effortlessly ingest.

Plot introduces real-time collaboration, visual communication, knowledge accumulation and determinism into AI workflows. It turns organizational know-how into tangible results that teams can depend on rather than fleeting artifacts they merely hope will work again next time. And with Plot, organizations are inherently perpared for whatever the AI world will think of next, because your organization's intent will always be the thing that will be driving the machine, and that is what Plot helps you capture. With Plot, you jump ahead of the competition by spending far fewer AI tokens on much higher quality AI output, by focussing on coherent intent rather than scattered execution.