People ask me all the time how Claude Code fits into my actual day-to-day work as an AI consultant. Not the marketing version — the real one. What does a Tuesday morning look like? How much time do I spend in Claude Code versus meetings versus manual work? When do I reach for it, and when do I just write the code myself?
Here's an honest breakdown of my Claude Code daily workflow. I've been doing this full-time for over a year now, mostly with Vancouver-based clients but some remote work too. This is what works for me — your mileage will vary depending on whether you're agency-side, in-house, or solo like me.
Morning: Client Work and Rapid Prototyping
I block 8am to noon for deep work. No meetings, no Slack. This is when I do all the actual building — prototypes, automations, client deliverables that require focused attention.
Most mornings start with a quick review of what's in my queue. I keep a simple Notion board with three columns: This Week, In Progress, and Waiting on Client. If something moved into "This Week" overnight, I'll prioritize it first.
A typical morning task might be: build a lead capture form for a Vancouver real estate agent that feeds into their CRM, sends a confirmation email, and logs everything to a Google Sheet for backup. A year ago that would take me 4–5 hours of manual coding. Now it's a 45-minute Claude Code session.
Here's how I structure that session:
- Initial prompt — I describe the full system in plain language: "Build me a lead form with name, email, phone, message. POST to this webhook URL. Send confirmation email via SendGrid. Log every submission to this Google Sheet via API."
- Scaffold review — Claude Code generates the HTML form, the server-side logic (usually Node or Python), and the integration code. I review the structure before it writes any files.
- Iteration — I test locally, catch issues (usually API auth or CORS stuff), and ask Claude Code to fix them. Usually takes 2–3 rounds.
- Polish — I adjust styling, add client branding, make sure mobile works. This part I often do manually because it's faster than explaining design nuance.
By 10am I have a working prototype deployed to a staging URL. I send it to the client with a Loom walkthrough. They usually respond with change requests by end of day, which I batch for the next morning session.
Midday: Async Communication and Content
Noon to 2pm is when I catch up on everything else. Client emails, Slack check-ins, posting to LinkedIn, writing blog content like this. I also use this window for any meetings that couldn't be avoided — usually discovery calls with new leads or check-ins with active clients.
For content work, I use Claude Code differently than I do for client projects. I'm not asking it to write full blog posts (I tried that early on — it doesn't work). Instead I use it for:
- Outline generation — I'll give it a keyword and ask for a logical H2 structure based on search intent
- SEO audits — I have a workflow I built for automating technical SEO that I run on my own site and client sites
- Code snippets — If I'm writing a tutorial-style post, I'll ask Claude Code to generate example code blocks so I don't have to context-switch into a code editor
Lunch is usually at my desk. I know that's bad. I'm working on it.
Afternoon: Iteration, Testing, and New Builds
2pm to 5pm is my second building block, but this one is more exploratory. If I have client revisions from the morning, I knock those out first. Then I'll usually spend an hour or two on internal projects — experiments, tools I'm building for my own workflow, or side projects like the directory sites I've shipped.
This is also when I test new Claude Code features or workflows. Anthropic ships updates pretty regularly, and I like staying on top of what's changed. Sometimes that means rebuilding an old script to see if the new version handles it better. Other times it's just reading release notes and trying the examples.
One pattern I've developed: I keep a "playbook" repo where I store reusable prompts and code templates that work well across multiple clients. When I solve a problem once — like automating client reporting — I document the exact prompt sequence and file structure so I can replicate it faster next time.
The biggest productivity gain from Claude Code isn't speed on a single task — it's being able to reuse solutions across clients without copying and pasting a bunch of brittle code. The prompts become the reusable asset.
When I Don't Use Claude Code
It's worth being clear about when I bypass Claude Code entirely and just write code the old-fashioned way:
- Quick CSS tweaks — If I just need to adjust padding or change a color, opening an editor is faster
- Debugging complex logic — Claude Code is great at generating working code, less great at tracing subtle bugs in existing code. For that I use Chrome DevTools and manual inspection.
- Performance optimization — If a client site is loading slowly and I need to profile it, that's still manual developer work
- Infrastructure work — Setting up DNS, configuring a server, managing environment variables — I do that myself or hand it off to a DevOps person
Claude Code is a tool, not a religion. Use it where it makes you faster. Skip it where it doesn't.
Evening: Planning and Admin
I try to wrap up client work by 5pm. After that it's admin: invoicing, updating my Notion board, reviewing analytics for client projects, scheduling next week's calls.
I also spend 20–30 minutes most evenings reading — usually AI/ML papers, blog posts from other consultants, or case studies from companies shipping Claude-powered products. Staying current is non-negotiable in this field. Things move too fast to take a week off from learning.
Once or twice a week I'll do a longer evening session if I'm working on something I'm excited about. That's when I built most of the projects I've written about here — nights and weekends when I had the mental space to experiment without client pressure.
What This Workflow Optimizes For
If you're thinking about structuring your own Claude Code workflow, here's what mine is designed around:
- Deep work in the morning — The best thinking happens before lunch. I protect that time aggressively.
- Async-first communication — I don't do daily standups or status check-ins. Clients get updates via Loom or email. This keeps meetings to 2–3 per week max.
- Batch similar tasks — If I have three clients who all need forms built, I do them back-to-back in one session. Context switching kills productivity.
- Reusability over speed — I'd rather spend an extra 15 minutes documenting a solution properly than save 10 minutes now and have to reinvent it next month.
This setup works because I'm solo. If you're on a team, you'd need more coordination time. If you're agency-side, you probably can't block off mornings the way I do. Adapt the principles, not the schedule.
Tools I Use Alongside Claude Code
Claude Code is my primary building tool, but it's not the only thing in my stack. Here's what else I use daily:
- Cursor — Code editor with AI assistance, great for when I need more control than Claude Code's interface offers
- Notion — Project management, client docs, personal knowledge base
- Loom — Screen recordings for client updates and demos
- Google Sheets + Airtable — Quick databases for prototypes before building something proper
- Figma — Wireframing and design handoffs, though I'm not a designer and my Figma files look like it
- GitHub — Version control for everything, even solo projects
Most of these integrate with Claude Code in some way. For example, I'll often ask Claude Code to generate an Airtable schema or write a GitHub Actions workflow. It's good at connecting tools together.
Key Takeaways for Your Own Workflow
If you're building your own Claude Code daily workflow, here's what I'd focus on:
- Protect deep work time — Claude Code is fastest when you can give it 60–90 minutes of uninterrupted focus
- Document your successful prompts — Build a reusable library so you're not starting from scratch every time
- Know when to code manually — Don't force Claude Code into tasks where a human is genuinely faster
- Batch similar work — Context switching is expensive, even with AI assistance
This workflow isn't static. I change it every few months as I learn what works and what doesn't. A year from now it'll probably look different again. That's fine. The goal isn't perfection — it's a system that lets you ship fast without burning out.
If you want to talk through how this might apply to your client work or internal projects, book a call. And if you're just getting started with Claude Code and want the basics first, start with my getting started guide.