I have ADHD. I never used to talk about it much before generative AI, but using these tools well means understanding how they work for you. One of the ways they help me is by creating focus and clarity in my days.
I have the kind of ADHD that lets me hyperfocus on the things I love and relentlessly chip away at them until they're done. I live in sprints of radical obsession followed by lulls where I never want to touch that thing again.
The flip side can be hard. When something doesn't interest me, it's not a matter of "I'd rather not." It's often debilitating. I'm staring at a mountain without a top, and someone's asking me to imagine being up there. When I love something, I'll be halfway up the mountain before realizing I'm climbing and it feels just as impossible to stop. When I don't, I sit at the base thinking about all the gear and food I'm going to need when I'm eventually climbing, convincing myself I'm preparing while actually procrastinating.
About two years ago, I started building an AI system that changed this. It helps me reframe the mountains as what they often are: small hills I can easily get over. Sometimes it even climbs the hill for me.
What the system does
My AI assistant does three things, and they're pretty simple.
First, it's a challenger. It knows my goals, the company's goals, and how I work. When I'm excited about something, it helps me figure out whether that excitement aligns with what actually matters right now. When it does, it helps me go faster. When it doesn't, it pushes me to delegate or ask whether the task even needs to happen. My brain will happily sprint in the wrong direction and leave me crushed by the deadlines I should have been working on instead. The AI helps me catch that before it becomes a problem.
Second, it's an unblocker. When a task falls into the "I really don't want to do this" category but still needs to get done and can't be delegated, the AI breaks it into bite-sized pieces. Often by asking me questions and pulling a plan out of me instead of handing me one. It gets me moving. My brain settles. And what looked like a mountain turns out to be a hill I can get over in a few minutes (we all have that email we haven't responded to in a week and instead we just keep thinking about it over and over...just respond to the email!).
Third, it's a memory. Every meeting note, every sales opportunity, every family commitment, every difficult conversation that needs follow-up goes into the system. The AI builds context over time, keeping both a current snapshot and long-term memory. So when I'm off track or need a coach to work through something hard, it has the full picture. Nothing falls through the cracks. For someone whose brain regularly moves on a little too fast, that's worth a lot.
Challenger
Knows my goals and pushes back when excitement drifts from priorities.
Unblocker
Breaks down the work I’m avoiding into something I can start.
Memory
Carries context across sessions so nothing starts from zero.
How I got here
I didn't wake up one morning with this system. I've been tinkering with AI tools since they were completely useless. When ChatGPT launched, I immediately became a daily user. I've tried just about every AI tool out there. The system I have now is the result of years of keeping what works and letting the rest go. As models improve, I retry things that failed before.
The setup started simple. A set of custom instructions in ChatGPT that told the AI who I am, what I'm working on, and how I want it to help. A weekly planning flow where I give it my priorities every Monday and it holds me accountable. A daily kickoff where I check in each morning and the AI cross-references my day against the week's plan. It's traditional productivity advice (plan your week, plan your day, time-block your calendar) rolled into an AI layer that makes it stick.
That simple version worked well for over a year. But the real shift happens when the AI knows more about you.
I now use Claude Code, Anthropic's command-line AI tool. It reads and writes files directly on my computer. My assistant has access to 5+ years of daily journals, thousands of exported ChatGPT conversations turned into reference material, my 2026 personal and company plans, and a library of custom skills I've built (including a writing partner that knows my voice, a thinking partner that pressure-tests my ideas before I act on them, and a brand-aware designer to build high-quality presentations).
The difference is context and ability. ChatGPT knows what I've told it in a conversation (and somewhat across past conversations). This local assistant knows what's in my files. It reads my task list, my daily notes, my meeting records, my project folders. When I start a session, it already knows what I was working on yesterday, what's due this week, and what I said mattered to me. It feels like an extension of me.
Local files
Tasks, daily notes, meeting records, and project folders.
Deep history
Five years of journals and thousands of exported conversations.
Custom skills
A writer, thinker, and designer built for my voice and workflow.
This article is a live example. I didn't write it from scratch. I talked through my ideas with the AI writing partner, sharing stories, opinions, concerns, and structure. It asked me questions, organized my thinking, did fact-checking, and assembled a draft in my voice. I'm doing the thinking. It's doing a portion of the execution. All the content is mine, and many of the words are too, but the drudge work of assembling a first draft was done by my assistant.
This article will feed directly into our Productivity Basics and Productivity Advanced AI Quest trainings. The assistant holds context across sessions, so my own experiments with what works and what doesn't shape how we update the content.
I'm still trialing this setup. It changes weekly as models improve and tools get more connected. But it feels like where work is heading. A year from now, the ChatGPT custom instructions I started with will feel like the basic version they are, as tools like Claude Code become more popular. That's fine. Start there. You'll know when you're ready for more, and you'll have a better feel for what works for you. That's the goal right now.
What AI can't do
You can't offload your thinking to AI. I want to say that plainly because there's a lot of people pushing hype out there who will tell you that AI has automated everything in their lives. One reason people are getting tired of hearing about AI is because they're tired of being lied to about its capabilities.
AI still takes work. It takes setup. It takes iteration. It takes you knowing what you want and being willing to say it clearly. The tools don't think for you. They help you think more by handling the parts that make it a drudge to get started.
And, coming from a lifelong tinkerer, it's 100% possible to tinker with AI tools as a form of procrastination instead of actually using them. I walk that line every day. There's always a new model to try, a new workflow to build, a new integration to set up. The tech version of prepping too long at the base of the mountain. I try to stay on the right side of it. Build systems that work, then use them. Iterate on what works, drop what doesn't.
Public sentiment on AI is getting more skeptical. Support for using AI in professional settings dropped from 62% to 49% in a single year. I think people are right to push back on the hype. The hype deserves pushback. But the tools themselves, when you're honest about what they can and can't do, are genuinely useful. I save hours every week. My follow-through is better. My days have more structure. Things that used to paralyze me now just get done (most of the time; as I tell my AI assistant, no one is perfect).
The trick is knowing which parts of your work to hand off and which to keep. If you let AI do your thinking, it'll make you worse at thinking. If you use it to clear the friction so you can think more, it'll make you better.
AI makes options. I make decisions.
If there's one framework I'd leave you with, it's this: AI makes options, humans make decisions.
In chess, the best players for about a decade were human-AI teams called "centaurs." The AI handled computation and option generation. The human handled strategy and judgment. Neither was as good alone as they were together. Researchers at Stanford's Graduate School of Business found the same pattern in other fields: people who used AI as a complement, letting it make recommendations where they were uncertain, outperformed both AI alone and unassisted humans.
That's how I think about my system. A client sends over an RFP. The AI pulls our past proposals, research on their industry, notes from our discovery calls, and our current strategic direction to draft a response. I review it, adjust the positioning, and decide what we're actually promising. The AI breaks a project into five steps. I decide which ones matter and in what order. The AI surfaces what I might be forgetting. I decide what to do about it.
AI makes the options. I make the decisions. That's how I run my week. A system, refined over years, that works with my brain instead of against it. The mountain was always a hill. I just need a little help seeing it sometimes.