DoView Boards
The new, faster, way of communicating to AI what you want to do
Quick Overview
Instead of having to type instructions to AI, you should be able to just visually show AI what you want it to do by communicating with it directly onto a visual plan showing the action you want taken. This is what builders do when they point at a specific room on a blueprint plan, or generals do when they point at a town on a map. You can now use a DoView Board in a similar way as a clear visual plan of what needs to happen and point at the steps you want the AI to do next.
Use the prompt below to get Claude to draw an on-the-fly interactive visual plan of what you want it to do. Then you can have a focused discussion with Claude by just pointing to the particular box you want to talk about. No more meandering textual chats with AI that get filled with too much unstructured information.
Get the Claude prompt to get Claude to automatically build a DoView Board. If you like this way of working, give feedback to the AI companies that you want a button on their interfaces that says ‘DoView it’ so you can communicate with their AI systems using this visual approach.
Please note that the Claude prompt used here is just a prototype for the use of DoView Boards. It is not a fully featured product because the Claude interactive diagram feature (wonderful though it is) has not been optimized for working with DoView Boards.
Using DoView Boards for communicating with AI is just being developed and at the bottom of the page, there is information for developers wanting to push the DoView Board approach further.
DoView Boards
A DoView Board is simply a diagram with a set of boxes on the left that need to be done first, to achieve high-level outcome boxes on the right. They can be used to plan anything from a two-day holiday to an entire country’s economic strategy. This works because the common starting point in any planning is identifying the ‘This-Then’ steps that need to be done. Often, when we are interacting with an AI system (and always when we are working with AI agents), we are planning things we want done.
A simple DoView Board of a cycling trip
Below is a static image of a DoView Board created on-the-fly by Claude using the DoView Board Prompt. Claude was given the prompt and then told to ‘create a DoView Board for a cycling trip to the Wairarapa for five adults’. It came up with all the detail in this DoView Board itself. As shown in the subsequent diagram you can use an interactive DoView Board rather than having to have a long-winded text-based discussion with Claude about the trip
An interactive DoView Board of a cycling trip
When using a DoView Board in within Claude you interact with it. Try clicking on a box on the left in the board below. You can enter into it information you want Claude to know about that box. When Claude gives you responses (e.g. where to stay, you put can put that under that box). You can also ask Claude questions about the box (this option does not work in the DoView Board below because you are not using it in Claude). And you can put in your own notes for yourself about the particular box.
Notice how when, for instance, in the Info for AI option, you say ‘completed’, the traffic light on the box will turn green, and if you do this for all of the boxes with orange traffic lights, Claude will then turn orange more of the boxes for you to have input into. So Claude understands the dependencies in the DoView Board.
At any time, you can ask Claude to output the contents of the DoView Board in any format you like (e.g. as an itinerary, a list of gear, etc.) as you normally would when having a long text-based chat with Claude about planning a trip.
A more complex DoView Board
DoView Boards are scalable so that you can use them to plan for anything. Here is a DoView Board for everything you need to do to build a house.
For developers wanting to push the DoView Board concept further
As noted above, the use of DoView Boards as a GUI for interacting with AI systems is a recent development. However, the concept is not new; it has a fully established theoretical foundation in outcomes theory. The approach comes out of strategy psychology. It conceptualilzes any action in the world as always being undertaken by an outcomes system. Such systems can be human, human/AI or A| systems. Outcomes systems specify outcomes, prioritize them, align actions to them, delegate actions, measure their success and improve. This is in addition to dealing with issues such as agents accountability and management.
as a way of reducing the cognitive load associated with planning and implementation. It does this by getting the mental ‘This-Then’ logic model out of people's heads and into an external shared thinking tool that can be used for group decision-making.
DoView diagrams have been used in thousands of instances to build human organizational ‘This-Then’ models. And the now legacy DoView app for drawing DoView diagrams won Gartner Cool Vendor recognition.
There are multiple ways in which the concept needs to be developed further if you are interested:
Optimizing the DoView Board user experience
On-the-fly building of DoView Boards has only become widely doable because of Claude’s interactive diagram feature. However, this feature in Claude (wonderful though it is) is not fully optimized for using a DoView Board as the central artifact. This is because when you ‘Ask AI’ under a box within the DoView Board Claude returns the answer down the page and scrolls away from the DoView Board so the user has to scroll back to it. Also Claude does not fully update the DoView Board until you type DoView and it takes some time to redraw it. At the moment in Claude if you use the Artifact window the DoView Board will not interact as it does when it is an inline interactive diagram in the Claude chat. You can mock up how an interface would work if you put the HTML for the DoView Board into Open AI’s Atlas browser and then open up the chat window on the right hand side of the DoView Board in the browser. You need to instruct ChatGPT to look for changes in the Your Notes field which are questions and then ChatGPT will return a response in the right hand window which means you can still see the DoView Board. This does not have the full functionality of using the interactive diagram in Claude, it merely demos a better user interface for using a DoView Board as a GUI to an AI system.
Using a DoView Board to structure code development
In theory a DoView Board could be used to control code development with different modules being located under different boxes in the DoView Board.
Using a DoView Board to control AI agents
DoView diagrams were developed in order manage and control human agent(s). There is no reason in theory why they could not be wired up to control agents. Their drill-down structure makes them scalable as you can just develp subpages with more and more detail to inform and control agent behavior.
Could DoView Boards be used by AI agent swarms to plan, coordinate and implement
At the moment DoView Boards have been optimized for humans. For instance, they are always broken up into subpages which a human can easily parse. DoView Boards could, in theory, have an accompanying version which is optimized for AI. If you look at the way in which Claude is using the existing DoView Board Prompt, it is building into the board some logic about dependencies. How much this needs to be formalized in the desing of t
however a theory based approach that has been used extensively in to plan human action in the world and the way of working is directly transferable to humans working with AI systems.
For developers wanting to push the DoView Board concept further
As noted above, the use of DoView Boards as a GUI for interacting with AI systems is a recent development. However, the concept is not new; it has a fully established theoretical foundation in outcomes theory. The approach comes out of strategy psychology. It conceptualilzes any action in the world as always being undertaken by an outcomes system. Such systems can be human, human/AI or A| systems. Outcomes systems specify outcomes, prioritize them, align actions to them, delegate actions, measure their success and improve. This is in addition to dealing with issues such as agents accountability and management.
as a way of reducing the cognitive load associated with planning and implementation. It does this by getting the mental ‘This-Then’ logic model out of people's heads and into an external shared thinking tool that can be used for group decision-making.
DoView diagrams have been used in thousands of instances to build human organizational ‘This-Then’ models. And the now legacy DoView app for drawing DoView diagrams won Gartner Cool Vendor recognition.
There are multiple ways in which the concept needs to be developed further if you are interested:
Optimizing the DoView Board user experience
On-the-fly building of DoView Boards has only become widely doable because of Claude’s interactive diagram feature. However, this feature in Claude (wonderful though it is) is not fully optimized for using a DoView Board as the central artifact. This is because when you ‘Ask AI’ under a box within the DoView Board Claude returns the answer down the page and scrolls away from the DoView Board so the user has to scroll back to it. Also Claude does not fully update the DoView Board until you type DoView and it takes some time to redraw it. At the moment in Claude if you use the Artifact window the DoView Board will not interact as it does when it is an inline interactive diagram in the Claude chat. You can mock up how an interface would work if you put the HTML for the DoView Board into Open AI’s Atlas browser and then open up the chat window on the right hand side of the DoView Board in the browser. You need to instruct ChatGPT to look for changes in the Your Notes field which are questions and then ChatGPT will return a response in the right hand window which means you can still see the DoView Board. This does not have the full functionality of using the interactive diagram in Claude, it merely demos a better user interface for using a DoView Board as a GUI to an AI system.
Using a DoView Board to structure code development
In theory a DoView Board could be used to control code development with different modules being located under different boxes in the DoView Board.
Using a DoView Board to control AI agents
DoView diagrams were developed in order manage and control human agent(s). There is no reason in theory why they could not be wired up to control agents. Their drill-down structure makes them scalable as you can just develp subpages with more and more detail to inform and control agent behavior.
Could DoView Boards be used by AI agent swarms to plan, coordinate and implement
At the moment DoView Boards have been optimized for humans. For instance, they are always broken up into subpages which a human can easily parse. DoView Boards could, in theory, have an accompanying version which is optimized for AI. If you look at the way in which Claude is using the existing DoView Board Prompt, it is building into the board some logic about dependencies. How much this needs to be formalized in the desing of t
however a theory based approach that has been used extensively in to plan human action in the world and the way of working is directly transferable to humans working with AI systems.
Design considerations when constructing systems using DoView Boards
DoView Boards are not just some random type of box diagram thrown together. The are drawn according to a very specific set of rules. These rules can be seen as ensuring that any DoView Board is drawn in a standard way. Building DoView Boards in a standard way means that they are then fit for purposes for use throughout the planning, prioritization, alignment, delegation, implementation, performance measurement, evaluation and improvement cycle.