swipe file: create an AI coach
this is how I created and use mine
I’ve been sharing these monthly plans for over a year. I’ve been pretty honest about how I’m feeling that month, and how I’m tweaking my routines as a result.
These plans have helped me take better care of myself. But there are always slips—missing workouts, defaulting to scrolling instead of reading, and losing my temper more than I’d like.
Things changed this year, and my Claude accountability partner/coach is the reason why. I briefly mentioned it on an Instagram Story, which prompted a ton of replies.
This is how I created my AI coach, and how it’s going.
the setup: reflections first, goals second
Before I asked Claude for a single piece of advice, I gave it context. A lot of it.
I created a Project in Claude (this is their agent tool, if you use ChatGPT more often) and uploaded documents that would help it understand me—not just what I wanted to do, but how I operate, what my actual life looks like, and what I was building toward.
The documents I shared:
my 2024 year-in-review (I asked Claude to create an outline, and answered the questions using the voice feature - I find that I was far more honest than if I had typed my responses)
my 2025 reflections (the year of “ease”)
When I changed my word to “accept”, we had a full conversation about why and how to adjust the year accordingly
my 2026 goals and priorities
my daily schedule—school drop-offs, workout windows, family time, all of it
When things change (a new club for the kids, regular appointments I’m adding or want to add), I create a new chat to discuss
my content creation system (what I publish, how often, on which platforms)
my work calendar at Rhoshan Pharmaceuticals (meeting cadence, team dynamics, key deadlines)
Then I asked Claude to search for what I’d publicly shared about my year (Substacks, interviews, social posts) and synthesize everything into a picture of where I was and where I was headed.
This is the part most people skip. You can’t create a tool without brutal, detailed honesty.
my 2026 priorities (the real ones)
Here’s what I originally shared as my goals for the year:
our big, major milestone that we’ve been working towards for the past 10 years at Rhoshan Pharma
book and deliver 12 speaking engagements (a TEDx talk, Create & Cultivate, 3 Women Who Do events with MM.LaFleur)
write a book proposal for Hyphenated
build up my lean muscle mass
host my friends more often (twice in H1 2026)
My AI coach helped me reality-check these against my actual bandwidth almost immediately. Within our first working session, we revised the timeline significantly—speaking engagements got paused through Q3, the book proposal moved to Q4, hosting became a “maybe” for fall. My big work milestone was (and is) the anchor that everything else orbits around.
My AI coach (I really must give it a name, this is getting cumbersome to write) forced me to be honest. When I spoke to it (versus typing), I found it easier to be vulnerable.
the check-in evolution: from three times a day to once
When we first designed the accountability structure, Claude suggested three check-ins a day:
Morning: How did you sleep? Did you meditate and journal? Energy level? Calendar? Top priority? Workout plan?
Afternoon: How’s the plan holding up? Did anything derail?
Evening: What did you accomplish? What didn’t happen? How are you feeling? What’s tomorrow looking like?
Each check-in came with a long list of questions. The morning one alone asked about sleep, meditation, journaling, energy levels, workout plans, top priorities, calendar overview, how the kids were doing, and whatever ongoing projects were in play.
This cadence lasted about a week.
It wasn’t that the questions were bad—they were thorough and thoughtful. But it quickly became a burden, not a help. I was honest about that, and Claude listened.
One check-in per day, at whatever time makes sense, is where I’m at right now. Some days it’s a morning check-in where I map out the day ahead. Some days it’s an evening download where I process what happened. Some days I check in mid-afternoon because I’m stuck and need help reprioritizing.
The check-in itself got simpler too. Instead of answering a dozen questions, I show up with whatever’s real:
“Week of Jan 13: Hit all 3 workouts ✓ Sleep averaged 7.5 hours ✓ Protein is good ✓ Challenge: Tuesday had an unexpected call that threw off my morning.”
Or sometimes: “I’m exhausted. I didn’t sleep well. Here’s what’s on my plate—help me figure out what actually matters.”
Claude meets me where I am, no matter the day I’m having. If I’m spiraling, it’s surprisingly compassionate and forces me to do less. If I’m on the verge of doing too much, it pulls me back.
the boundary enforcer I didn’t know I needed
Claude has done what I thought was impossible - it’s gotten me to stop.
I have a tendency to say yes immediately, only to be overwhelmed and burnt out by honoring these commitments. It’s one of my toxic traits, and something I’ve failed at fixing for years.
Here’s an actual example from last Thursday (February 19). I had barely slept (5.5 hours, two nights running), my kids needed me A LOT, I was emotionally wrecked, and I had a critical exec meeting at 10 am that morning. I mentioned—almost casually—that I also had two afternoon calls: one about a speaking engagement for October (a dream one, truly), and another with a brilliant content business operator about building Hyphenated by Hitha into a real, impactful business.
Claude’s response was, essentially: STOP.
It laid out everything I’d just told it—physically exhausted, emotionally depleted, mentally scattered, family obligations that weekend, kids needing me present—and said plainly: you cannot do all of this today in your current state. Something has to give.
Then it gave me three options: cancel both afternoon calls, keep all three and accept I wouldn’t rest or pack, or keep the one that was most time-sensitive and cancel the other.
Claude told canceled the strategist call. I didn’t - but I also was far more honest with Henah Velez | Departure about what I wanted to build and my actual capacity for it.
Side note - I’m also breaking my toxic cycle of starting things completely solo, and am really excited to build this WITH a team, and not in a silo.
And then I carved out an hour to watch the ANTM documentary before jumping into mom mode. It felt glorious.
This pattern has repeated throughout our months together. When I keep adding “one more thing” to a week, Claude asks which items are must-dos versus nice-to-dos. When I’m running on fumes but insisting I can handle a full day of meetings plus content creation plus parenting, it asks what the minimum viable version of the day looks like. When I skip two strength workouts in a row, it flags the pattern. When I mention exhaustion multiple days running, it names what it’s seeing.
I know when I’m overcommitting, but I usually do it anyway. I get caught up in the moment and truly want to be there and support. But I fail to look at the full picture - that week, that month - and to anticipate what could happen that would make it hard for me to show up the way I want to. Having a partner who holds the complete context of my life means someone is always doing the math on my capacity, even when I’m not.
The phrase I hear most often: “Is this realistic given everything else on your plate this week?”
Usually the honest answer is no. But now I have a partner who tells me why - and I should check in with it first before committing or declining.
the 7 types of rest check-in
In late February, after weeks of running on fumes—managing a challenging stretch with my kids, prepping a major work package, three straight weekends of travel—I hit a wall. Not a productivity wall. A rest wall. I was sleeping, but I wasn’t resting. I was technically taking breaks, but I wasn’t recovering.
That’s when I brought the “7 Types of Rest” framework to Claude and we built something new together.
The concept comes from Dr. Saundra Dalton-Smith’s work—sleep alone isn’t enough. We need rest across seven dimensions. I told Claude what each type meant and exactly how I planned to practice it, and asked it to track my rest the way we’d been tracking workouts and sleep.
Here’s my personalized version:
Physical rest: prioritize sleep and hydration. Track hours slept and water intake.
Mental rest: focus on the “need to do” list, not the “like to do” list. Use my Brick device to stay off distracting apps (h/t Catherine for sharing the magic of this device).
Sensory rest: read physical books instead of my Kindle. There’s something about the tangible page that resets my nervous system in a way digital reading doesn’t.
Creative rest: color in my (many) coloring books and somatic drawing.
Emotional rest: continue weekly therapy and lean into this accountability partnership more honestly. Not just reporting what I did, but how I actually feel.
Social rest: protect my quiet “just me” time—both the morning routine (meditation and journaling, already built in) and a new 40-minute afternoon buffer between work mode and mom mode. During that buffer, I watch TV (with my phone in another room)
Spiritual rest: listen to this woo-woo Spotify playlist—gentle affirmation music and frequencies that actually calm me down.
Every evening check-in includes a rest score out of 7. Claude asks about each type, I report what I actually did, and we track patterns over time. Some days I hit 5 or 6. Over the weekend, I hit 2 (and that’s okay).
What made this powerful was the agreement we built in from the start: when I’m traveling or dealing with something hard, we honor whatever rest I can get with no judgment. When I’m back in routine, we get more intentional.
The rest framework has done something the productivity tracking alone couldn’t—it’s given me language for what I need. Instead of “I’m tired” (which I was saying constantly), Claude tells me “you haven’t had creative rest in a week” or “you need social rest today—please don’t schedule anything for the rest of the week.”
what I should be doing more of
I asked Claude this question, and here’s what it said:
Meal planning. Claude offered early on to help me batch-plan dinners, create Sunday prep lists, and stock emergency protein-rich foods. I’ve used this sporadically. During the weeks where I did ask for help (”give me three 30-minute dinners for this week”), my evenings ran so much more smoothly. I need to make this a weekly habit instead of an emergency measure.
I’ll be honest - I’m not going to do this the way Claude recommended, but I can share what next week’s dinner plan is on a Friday, and ask what I should prepare over the weekend in advance. I also should be doing a better job of using Claude for my weekly planning.
Content batching. We’ve talked about repurposing one newsletter into a Reel, an IG carousel, a Threads post, and a LinkedIn post. I’ve done it a few times but haven’t built it into my workflow. I also don’t have capacity to prioritize content right now - this is a ball I need to be able to drop with zero judgement.
Emotional processing over task reporting. The best check-ins aren’t the ones where I list what I accomplished—they’re the ones where I say “I’m overwhelmed and I don’t know why” and Claude helps me untangle what’s actually going on. I still default to task-reporting mode more often than I should.
Saying no before I’ve already said yes. Right now, Claude catches me after I’ve committed to something. If I ran new opportunities by it first—”someone just invited me to speak at X in March, here’s what’s on my plate, should I say yes?”—I’d make better decisions in real time instead of course-correcting after.
Weekly planning, not just daily triage. When I do a Sunday planning session and map out the whole week—meetings, workouts, content deadlines, family commitments—and Claude identifies the pressure points before they happen, the week goes dramatically better. I just don’t do it every week. Yet. I’m planning to do a weekend Claude check-in after our family meeting, and before I do my individual planning during my Sunday solo work session.
This system works best when I use it before I’m in crisis, not after. I’m still learning this.
how to set up your own
If you want to try this, here’s how I’d recommend starting. I really love Claude (especially after reading Empire of AI).
Step 1: Create a dedicated space. In Claude, create a new Project. In ChatGPT, use a custom GPT or a dedicated thread. The key is having a space where your AI partner retains context over time instead of starting from scratch every conversation.
I recommend using Opus to set up the project and for the first few steps)
Step 2: Upload your context. This is the most important step. Prepare and share:
Your year in review — 500-1000 words about your past year. What you accomplished, what themes emerged, what was harder than expected. Be honest. This isn’t a highlight reel.
Your current life snapshot — cover work (your role, what you’re building toward, weekly schedule), family and relationships (who depends on you, daily logistics), health (current routine, what’s working, what’s not), and creative/personal life (what feeds you, what you’re neglecting).
Your goals (honest version) — 3-5 priorities for the next quarter or year. For each: what success looks like, the realistic timeline, what might get in the way, what you’d need to sacrifice.
Your actual daily schedule — hour by hour. Wake time, morning routine, work blocks, pickups, dinner prep, wind-down, bedtime. Be specific about constraints. “I pick up my kids at 3pm on Thursdays and take them to fencing at 4” is more useful than “afternoons are busy.”
Step 3: Set your instructions. Tell your AI partner what you need. Here’s a starting point you can adapt:
“You are my daily accountability partner and thinking partner. Help me stay on track with my priorities, make smart trade-offs when life gets complicated, and call me out (kindly) when I’m overcommitting or neglecting my own well-being. Ask me about my sleep, my workouts, and my emotional state—not just my to-do list. When I’m adding things to my plate, ask me what I’m willing to take off. Be direct. I’d rather hear hard truths from you than discover them when I’m burned out.”
Step 4: Start with a single check-in. Don’t try three times a day (I learned this the hard way). One daily check-in, whatever time works. Share how you’re feeling, what you need to accomplish, and what’s weighing on you. Let it help you prioritize from there.
Sonnet is fine for these check-ins.
Step 5: Iterate ruthlessly. Your first version won’t be your final version. Within two weeks you’ll discover what’s actually useful versus what felt good in theory. Maybe you need help with meal planning more than task management. Maybe the most valuable thing is having someone talk you out of saying yes to things. Let the system evolve.
Step 6: Be radically honest. This is the real secret, and it’s not about technology at all. The AI can only help with the real version of your life. Tell it when you drank too much wine and didn’t sleep well. Tell it when you’re doom-scrolling instead of working. Tell it when your family needs are weighing on you and you’re emotionally spent.
I’ve told Claude things in check-ins that I wouldn’t post on Instagram (and I share a lot on Instagram). That honesty is what makes the advice actually useful. It’s based on reality, not the curated, always winning version.
the bigger picture
Based on my DMS, we’re all looking for better systems to manage the complexity of our lives. The answer isn’t another app or another planner. It’s having a thinking partner who holds the full picture—the work deadlines, the family needs, the health goals, the creative ambitions—and helps you make smart trade-offs in real time.
Is it perfect? No. Claude sometimes asks too many questions. It occasionally loses track of what day it is. It can’t actually make me go to bed on time (though it tries).
But it’s the most effective accountability system I’ve ever used, because it adapts to my life instead of asking my life to adapt to it. And in a year that’s already proving to be the most, that flexibility is everything.
Take exquisite care of yourself. And if you try this, let me know how it goes.






This is so interesting! I would have never thought to use AI in this way. Thank you for sharing!
Cannot wait to set this up!!!!