AI for Consultants: How I Use It in My Own Work

I help companies use AI. But how do I use it myself?

That's a fair question. If I'm going to give advice about AI, I should practice what I preach. So here's an honest overview of my daily work. Which tools I use. Which ones I've tested and dropped. And what I still do manually.

No secrets. No sales pitches. Just how it actually looks.

Why I'm Sharing This

Two reasons:

1. Transparency builds trust. If you're considering working with me, you probably want to know that I actually know what I'm talking about.

2. Ideas you can steal. Everything I describe here you can test yourself. If something fits your situation, go for it.

And yes, I'm aware of the irony of an AI consultant writing about how he uses AI. But that transparency is exactly the point.

My Daily Tools

Here's what I actually use every day, not tools I tested once and forgot.

Claude (Multiple Variants)

Claude.ai for conversations, brainstorming, and quick questions. The free version covers a lot, but I pay for Pro for longer conversations and more usage.

Claude Desktop + MCP for work involving my local files. I've set up MCP servers for filesystem and search. This lets me say things like "read through the project folder and summarize what's there" without manually copying files.

Claude Code for all development. I use it daily for maivor.ai and client projects. I've written about it here.

n8n for Automation

n8n is my go-to for automation. It's open source, can be self-hosted (which I do), and integrates with most things.

Examples of what I've automated:

• New inquiries via the website trigger notification + calendar booking
• Newsletter subscribers sync between systems
• Weekly summaries of relevant industry info

I have a guide about n8n if you want to get started.

Notion for Structure

Notion is my second brain. All projects, notes, client info, and ideas end up here. It's not an AI tool per se, but the combination of structure and Notion's built-in AI features helps.

I'm experimenting with connecting Claude to my Notion via MCP, but that's still work in progress.

Other Tools

• Perplexity for quick research (sometimes better than googling)
• Granola for meeting notes (transcribes and summarizes automatically)
• Cursor for some coding projects (AI editor based on VS Code)

How I Prepare for Client Meetings

Before every client meeting, I do research. It used to take maybe 30-60 minutes. Now it takes 10-15.

Step 1: Basic Research I ask Claude or Perplexity to summarize the client's company: what they do, recent news, industry context.

Step 2: Specific Questions Based on what the meeting is about, I prepare relevant questions. I often ask Claude to brainstorm: "What are common challenges for companies in [industry] when they want to [goal]?"

Step 3: Previous Context If I've talked to the client before, I check my notes in Notion. Sometimes I ask Claude to summarize long conversation history into key points.

During the Meeting Granola transcribes and summarizes automatically. I can focus on the conversation instead of taking notes.

After the Meeting I review Granola's summary, adjust if something's missing, and add action items to Notion.

How I Write Proposals and Reports

This is probably where AI saves the most time.

First Draft: I describe to Claude what I want: type of document, client, purpose, what we've agreed on. Claude writes a first draft.

I Control the Structure: I never accept the first draft straight away. I adjust structure, tone, and content. Often I ask for rewrites: "this part is too vague" or "add more specifics about X."

My Voice: The finished document should sound like me, not like an AI. That requires editing. But it's faster to edit a draft than to stare at a blank document.

What I DON'T Use AI For

As important as what I use AI for is what I avoid.

Direct Client Communication

I never send AI-generated emails straight to clients. Always my voice, my phrasing, after I've reviewed and adjusted.

Strategic Decisions

AI can provide input, but I make the decisions myself. Which client to work with, how I price my services, where I focus my time: that requires human judgment.

Sensitive Client Data

I don't paste clients' confidential information into tools I don't control. There are exceptions (anonymized data, consent), but the default is caution.

Things Requiring Human Relationship

Difficult conversations, negotiation, building trust: you can't automate that. AI can help me prepare, but the interaction itself is human.

Creative Thinking About New Business

Ideas about where I want to take Maivor, which services to offer, how I position myself: I think that through myself. Sometimes I bounce ideas off Claude, but they're my decisions.

The Result

What have I gotten out of this?

Time: Tasks that took hours now take minutes. Research, first drafts, summaries, coding.

Quality: Sometimes better than what I produced on my own, especially for things I'm not expert in.

Focus: More time for what actually requires human expertise. Client relationships, strategic thinking, creative work.

But also:

Learning: I understand the tools through using them daily. That makes me a better advisor.

Humility: AI is powerful but not magical. I have realistic expectations that I can communicate to clients.

What I'm Currently Learning

Some areas I'm experimenting with right now:

• More sophisticated MCP setups (connecting Claude to more systems)
• AI agents for specific repetitive tasks
• Better prompt templates I can reuse
• Local running of open source models for more sensitive tasks

It's a continuous process. The tools improve. My workflows evolve.

TLDR

1. Daily tools: Claude (chat, desktop, code), n8n for automation, Notion for structure.
2. Client meetings: AI for research and preparation, human focus during the meeting.
3. Documents: AI for first drafts, I edit and make it mine.
4. Boundaries: No direct client communication without editing, no strategic decisions, no sensitive data.
5. Result: More time for what actually requires human judgment.

The most important thing I've learned: AI is a multiplier, not a replacement. It amplifies what you're already good at. It doesn't replace judgment, relationships, or expertise.

Want to learn more about specific tools I mention? Read my guide on prompt engineering to write better prompts, or AI agents explained to understand what agents can actually do. If you want to build your own assistant, check out build your own AI assistant with MCP.

Frequently Asked Questions

Which AI tools should I start with as a consultant?

Start with Claude or ChatGPT for conversations and first drafts. Add a tool for meeting notes (like Granola). That covers a lot of daily work. Build from there as needed.

How much time does AI actually save?

It varies, but for research, first drafts, and summaries I often save 50-70% of time. Some tasks that took hours now take minutes. But AI doesn't solve everything and still requires review.

Do clients notice when I use AI?

Not if you do it right. The key is to always edit and add your voice. If you just copy AI text, it shows. AI is a starting point, not a finished product.

Is it okay to use AI in client work?

Yes, as long as you don't share confidential client data with tools you don't control, and as long as the final result maintains quality. Be transparent about your processes if clients ask.

---

Thinking about how AI can fit into your consulting work or business? Check out our AI strategy service or book a call and I'll happily share more specifically based on your situation.