I’ve been using AI to work through ideas. Over the past two weeks, I’ve been testing how it can help with the work that follows: preparing an email, checking a schedule or getting a presentation ready.
The early groundwork happened in my existing AI conversations. By September 30, I was coordinating more work through Jarvis, including voice.
By “operating system,” I mean a way to connect priorities, information, preparation, action and follow-through. I’m still building it. These two weeks have produced useful work, exposed gaps and clarified where human judgment belongs.
Key takeaways
- Give AI a specific job, the right source material and a clear definition of done.
- Separate preparation, execution and approval.
- Check the result, including what happens after the visible button.
- Keep your experience involved throughout.
Where did the shift begin?
On September 25, I was working through “Experience + AI = Growth.” By September 28, AI had helped turn a service idea into a written proposal. On September 29, the operating-system concept had its own page on my website.
The next step was connecting that thinking to ongoing responsibilities. What needs attention? What can Jarvis prepare? What still needs my decision?
That’s the foundation of Bouf Ops, my developing personal and business operating system.
What work are we actually doing together?
Bringing the day into focus
Briefings and dashboard updates brought calendar commitments, active work and pending decisions into one view.
On October 3, we corrected a dashboard showing yesterday’s meetings. It was a snapshot requiring updates.
I also asked Jarvis to work through one item at a time and pause for my answer. A useful assistant has to fit how I make decisions.
Preparing communication and protecting time
Jarvis reviewed email threads, found supporting files and prepared replies with the relevant links and attachments. Some stayed as drafts. Others were sent after I authorized them.
Calendar work included private holds and conflict checks. On September 30, adding a speaking commitment exposed an overlap with another meeting. On October 1, a proposed meeting time remained a hold while the group’s confirmation was pending.
Those distinctions keep a proposed plan from becoming a false certainty.
Working from existing material
For an upcoming talk, Jarvis adapted an existing presentation to a new venue and date, then prepared a smaller delivery copy, preserving its slides, notes and links.
Town-guide revisions involved comparing versions, returning to an earlier design direction and preparing previews for review. My feedback determined the direction.
Video work taught an equally important lesson: an approved script needs to stay intact. We tightened checks for exact wording, speaker assignments and the correct version before moving ahead.
The common pattern was source material, specific instructions, a reviewable result and human feedback.
Checking the work behind the website
For the AI Growth Plan website, we worked through page revisions, assessment behavior, forms, confirmation emails and a booking option. Testing found a repeat-request setting that prevented an email from going out even though the form had saved.
By October 7, a booking button was live with business-calendar conflict checks. My separate personal calendar wasn’t included, and a complete test appointment still needed verification.
A page can look ready while important steps remain unfinished.
Reconciling records and planning the next build
League scheduling meant comparing emails, the official schedule and a referee spreadsheet. We needed to distinguish confirmed changes from proposals, and a blank referee field from proof that nobody had arranged coverage.
Client operating-system planning used the same discipline: identify recurring work, define responsibilities and decide where review belongs before promising automation.
The connections between these ordinary jobs are where the value starts to grow.
What does proactive help look like?
Reactive help begins with my request: review this email, adapt this deck, check this schedule.
Proactive help brings a relevant issue to me before I ask. That included flagging an upcoming game with no referee listed and catching a phone assistant’s meeting promise that had no corresponding calendar booking.
Jarvis could surface the gap and prepare the next step. I still had to supply missing context, make decisions and authorize the appropriate action.
What needed fixing?
On October 6, we confirmed that the content dashboard’s tracking controls weren’t connected to a working production process. A completed video also didn’t prove that an ongoing production queue worked.
We separated what the screen displayed, what had happened and what still needed testing.
The lesson: ask for evidence of completion. Open the file. Check the sent message. Test the form. Review the exact video.
My 90/10 vision, with AI handling more routine operations and people focused on strategy and judgment, remains an aspiration. I haven’t measured a percentage achieved, hours saved or financial return.
How can you try this?
Choose one recurring job you already understand.
Give your assistant the relevant, appropriate source material. Describe the output you need. Specify what it may do and where it must stop for approval.
Then review the result against the real task. Keep what works, correct what doesn’t and repeat the process before expanding it.
Frequently asked questions
What do you mean by an AI operating system?
I mean a connected way of organizing work around an AI assistant: priorities, useful context, preparation, actions and follow-through. Bouf Ops is my version of that approach. Its value depends on which connections actually work and whether the results can be checked.
Where should a beginner start?
Start with a repeated task you know well, such as preparing a meeting brief from approved notes. Your experience helps you recognize a useful result and catch mistakes. Give the assistant a defined output, review it and improve the instructions before adding more responsibility.
Can Jarvis work proactively?
In my experience, it can surface relevant gaps from information it can access, such as a scheduling conflict or an unconfirmed assignment. Proactive help still needs context and boundaries. A flag starts a decision; it doesn’t prove the situation or authorize every possible response.
What should stay with the human?
For my work, that includes priorities, relationship judgment, factual review and the approvals appropriate to each task. I also need to decide whether an output represents what I intended. A polished draft or finished render still needs someone accountable for using it.
Have you reached 90 percent automation?
No. That’s a direction I’m working toward, not a measured result. These two weeks show concrete task types, useful outputs and problems that needed correction. The honest next step is to keep testing the work and measure results before attaching percentages or savings to it.
Your experience is part of the system
What interests me most is the developing collaboration: I bring context and judgment; Jarvis helps gather, prepare, connect and follow through within the agreed scope.
Start with work you understand. Build a process you can trust. Grow from there.
Experience + AI. That’s growth.