Most nonprofits have a theory of change somewhere, either implicit in how the team talks about the work or drawn up once for a funder and rarely revisited. The boxes in a logic model are usually easy to fill in. What's harder, and where programs tend to fall apart, are the arrows between those boxes, because each arrow rests on assumptions that rarely get written down. Claude can draw the whole chain inline as you describe your program, making every arrow clickable so you see what each link is actually assuming, and the act of clicking through them is often the first time those assumptions get stated plainly.
Here a program officer runs a youth mentoring program and wants to see where the logic might be thin before the next evaluation conversation. Claude draws the chain from inputs to graduation rates, and clicking any arrow surfaces what has to be true for that step to hold.
I run a youth mentoring program. We pair high schoolers with adult mentors for weekly meetings over a year, and the goal is higher graduation rates. Show me our theory of change, and I want to be able to click the arrows between boxes to see what assumption each one is riding on. Where are the weak links?
No files are needed for this. A sentence or two about who you serve, what you do, and what outcome you're aiming for is enough for Claude to fill in the causal logic between, and asking "where are the weak links" in the prompt tells Claude to focus on the arrows rather than just drawing boxes.
Nothing to upload.
If you have a formal logic model or a paragraph from a grant narrative handy, paste it so the chain gets built from your framing rather than a generic one. That said, starting vague often surfaces assumptions you didn't realize you were making.
Claude draws your program's causal chain as five connected boxes running left to right, with every arrow between them clickable. The boxes will look familiar since they're the standard logic model categories, but the arrows are where the real work happens. Click one and a short note surfaces naming the assumption that link depends on, which is often the first time you've seen it written down.
Click any arrow and Claude expands on what's underneath it, naming the assumption, the things that typically break it, and what early signal would tell you whether it's holding.
Focus on the link between contact hours and stronger school connection. What would have to be true for that to hold, and what would the early warning signs be if it isn't?
Describe a change to the program and Claude redraws the chain with the new component slotted in, surfacing the assumptions you've just introduced alongside the ones you already had.
Redraw this with a family engagement component added alongside the mentoring. What new assumptions does that introduce, and does it change any of the existing arrows?
Claude takes the assumptions it flagged as shakiest and proposes one concrete thing to track for each, turning the theory of change into a data plan you can bring to an evaluation conversation.
For each of the arrows you flagged as weak, give me one thing I could measure this year to know whether the assumption is holding.
Asking "where are the weak links" is what gets you clickable arrows with assumptions under them rather than just a static diagram. A plain "draw my theory of change" tends to produce the boxes without the stress-test. The same phrasing works on any chain of cause and effect, whether that's a policy argument, a product strategy, or anything where the steps sound right but haven't been pushed on.
The assumptions Claude surfaces are a reading of your program, not a verdict. If Claude marks an arrow weak that you've got solid data on, that's useful confirmation. If Claude marks strong an arrow you've been privately worried about, it's worth telling Claude what you've seen so the note under that arrow can be rewritten around your evidence.
Hover over the chain for options to copy it as an image for a slide or save it as an Artifact if you'd reopen it later to click through again. You can also ask Claude to write the narrative version, which turns the boxes and arrows into paragraphs and gives you a starting draft for the logic model section of your next proposal.
