The prompts
Every Supercharged Prompt carries all seven moves. Paste one in and AI runs it with you.
See Supercharged PromptsThe If Possible Method
You wouldn't let a new hire mail the appeal on day one
AI shows up knowing everything on the internet and nothing about your world. You've onboarded smart people who didn't know you yet. This is the same job.
A real year-end campaign
What can one person do in eight hours?
In September, Colleen Cook sat down to find out. Working alone, with AI, she built a full year-end campaign for an If Possible customer.
It wasn't a better prompt. It was better context and better decisions.
Read how Colleen built it91pieces, including
- A campaign plan, ready for leadership
- An appeal letter for each audience, with reply cards and envelopes
- GivingTuesday emails, ads and social
- A giving page with a case video
- Weeks of reels, carousels and animated posts
- A thank-you card made for handwritten notes
- One campaign hub for staff and board
The method
What is the If Possible Method?
The If Possible Method is how Colleen Cook and Scott Williams teach nonprofit teams to work with AI. You treat AI like a smart new hire who doesn't know you yet.
You give it your context before you ask for content. You talk your thinking out. You let it ask questions, one at a time. You stay the one who decides. You react to a plan before anything gets made. You check its work. You save the steps.
None of that is new. You've done all of it with people.
You're just training a new employee. This is what it is. This is a graphic designer who says, I am so good at graphic design, but I know nothing about your brand.
Here are the seven moves, and how each one worked on Colleen's campaign.
Move 1
Give it your context
What should AI know before it starts?
Who you are, who you serve, how you sound, how you run.
Put it in one place AI reads before every task. We call that a Mission Brain. A one-page brief is a fine start.
Skip this and AI won't stop to ask. It will guess, and its guess is the average nonprofit.
- Who are we?
- What do we do?
- How do we sound?
- How do we run?
Mission Brain
The memory AI reads before every task.
On the campaign
The context was already there. The brand and the case for support took about a year of work, and none of that was AI generated. The Mission Brain held it all before the clock started. It's why eight hours was enough.
Move 2
Say it out loud
Why talk to AI instead of typing?
Record yourself talking through the job, or record the meeting where you already did. Hand AI the transcript.
A typed request gives it one tidy sentence. Talking gives it the reasons, the people and the history.
On the campaign
Colleen didn't start by typing. She talked through what she wanted the campaign to be: the focus, the ask, the goals, where she saw opportunity in the donor base.
all of this came out of a conversation that I had with my phone
Move 3
Let it ask
How do you stop AI from guessing?
AI won't ask questions unless you tell it to.
So tell it. Have it read what it has, then ask you one question at a time, the most important first. Have it say what it plans to do before it does it.
You'll find decisions nobody on your team ever made.
On the campaign
Before anything got made, AI found what Colleen hadn't said, and asked.
- What are the goals?
- How did last year's campaign perform?
- Is there a matching gift or challenge grant?
- What story should lead the campaign?
- Do we have consent to use these stories?
Five guesses it didn't make.
Just said a lot of things in a kind of incoherent way. Ask me any questions before we send this to clarify your understanding. Tell me what you're gonna do, and then let's do it.
Move 4
Make the decisions
Who makes the decisions when you work with AI?
You do. The strategy and the final yes stay with a person.
AI drafts. You decide who it's for, what it says and whether it's good enough to go out with your name on it.
When AI makes those calls, you get slop: work that sounds fine and says nothing.
The meat proxy
Pastes what AI wrote into an email and hits send. Adds nothing. Six words of thought go in, and 600 land on someone else's desk.
The director
Names who it's for. Writes the one point first. Asks for versions, not one answer. Says yes or no.
On the campaign
Colleen had AI build a small app, her command center, to hold the plan. Every open question sat there until she answered it. Each answer became the creative. That's where half her time went.
over the course of those eight hours, I would guess I probably spent four of the hours in the command center thinking through decisions, building strategy, and working with the plan.
You don't have to be an expert in this, you just need to have the expert mindset of I'm the director, I'm the leader of this. It's my critical thinking. I know my organization best.
Move 5
See a plan first
Why ask AI for a plan before the finished work?
Reacting is easier than starting.
Ask for a plan, an outline or a few versions. Say what's wrong in one line. Throw out what doesn't work and ask again.
On the campaign
Colleen's first ask wasn't a letter. It was a proof sheet: one deck showing every piece she could make.
- First sheetThe copy was missing. She went back and asked for everything.
- Second sheetThe copy was there. She moved it forward with notes.
- Third sheetPhotos, matched to what each one could be used for.
- Approved19 slides. Copy, photos and layouts. Only then the build.
I have reviewed and approved copy and photos and layouts for everything in the 19 slide proof sheet. It's time to go build.
Move 6
Check its work
How do you check AI's work?
Read AI's work line by line. It writes smoothly, and it can be wrong with total confidence.
Assume something is off and go find it. Check every fact, number, name and quote against your own records. Ask what it assumed. Ask whose story it's telling.
Then say exactly what's wrong, so it doesn't happen twice. Nothing goes out until a person says yes.
AI is the smartest, dumbest thing you'll meet, so check its work
On the campaign
Colleen read the work and said what was wrong, in plain words. When 91 pieces got to be too much to track, she had AI build a review board with every piece in one place. Some of her notes:
please fix the overset text here. This needs to print as a one page letter.
I don't like the logo treatment.
now that I'm seeing this one live, I'm realizing the copy doesn't quite work.
I still see way too much white space there.
Eight hours got the campaign to third base. The customer's team still refines it and chooses what to use.
Move 7
Save the steps
How do you make good AI work repeatable?
When something works, save it. The brief, the prompt, the rule you had to say twice. Give it a name and keep it where your whole team can find it.
Then the next job doesn't start from zero.
When something comes out great, you save it and then you give it a name
On the campaign
Mid-build, Colleen told AI to always use footnotes for credits, then told it to save that rule, so she wouldn't have to say it twice.
The command center she built to run the campaign didn't stay on her desk either. Nonprofit teams in our Intensive built their own, and one of them adapted the prompt for a strategic plan dashboard.
You don't have to memorize this
The articles
The Art of the Possible goes deeper on every move, with real stories from nonprofit teams.
Read the articlesThe quiz
Not sure where your team stands? Thirteen questions, three minutes, three next steps.
Take the AI Readiness Quiz
Who's behind it
Who teaches the If Possible Method?
Colleen Cook and Scott Williams, the co-founders of If Possible, an AI consultancy for ambitious nonprofit leaders.
They also run Vinyl, the agency where they've helped 100+ nonprofits raise more than $200M. They work this way themselves, and they teach it live to nonprofit teams.
Questions
What leaders ask first
What is the If Possible Method?+
A way of working with AI, taught by Colleen Cook and Scott Williams at If Possible. You treat AI like a smart new hire: context before content, talk your thinking out, let it ask questions one at a time, stay the one who decides, react to a plan, check its work and save the steps.
What is a Mission Brain?+
It's what we call one place that holds what your nonprofit knows about itself, written so AI can read it before every task. It answers four questions: who are we, what do we do, how do we sound and how do we run. You can start with a one-page brief.
What is a meat proxy?+
Someone who pastes AI output into an email, chat or document and sends it on without adding anything of their own. The term isn't ours. We just work hard not to be one. Read more: What is a meat proxy, and how do you avoid becoming one?.
Do I need to be technical?+
No. If you can explain your work to a colleague, you can do this. You'll talk, read and decide. Colleen isn't a graphic designer, and she built those 91 pieces.
Is this prompt engineering?+
No. A clever prompt helps once. Great context keeps paying off on every task after. Most of this method happens around the prompt, not inside it.
Does it work in ChatGPT, Claude and Gemini?+
Yes. It's a way of working, not a product. We use Claude. If your team uses something else, the same moves apply. Here's how the tools compare.
Is it safe to put our information into AI?+
It can be, with the right setup. Keep donor data out of free tools. Use a paid plan with model training switched off, and follow your own AI policy. Read what to check first.
Where should we start?+
With your context. The Mission Brain Brief is one free prompt that interviews you and hands back a one-page brief. Then take the next real thing on your list and run the seven moves on it.
Start with your context
One free prompt interviews you and hands back a one-page brief. Every other prompt starts from it.
Start with your contextKeep reading
From The Art of the Possible
Implementing AI
Treat AI like a new hire, not a new tool
The nonprofits pulling ahead onboard AI the way they would onboard a person, giving it context, a role, and time to learn the work.
Implementing AI
Your first 30 days with AI: a nonprofit rollout plan
A nonprofit AI rollout plan for month one: build the Mission Brain, ship one workflow, add the checks, then expand. One reliable thing at a time.
Implementing AI
Why nonprofit AI pilots stall, and how to fix it
Most nonprofit AI pilots stall for one reason: no owner, no goal, no feedback loop. Set up those three and the work finally sticks.