Using AI Honestly: How Imagewize Integrates Claude and Mistral Into Daily Development
Most web development agencies are using AI tools like ChatGPT, Claude, or Mistral in their daily work. Ask them privately and they’ll agree—these tools are useful for coding, writing, debugging, and more. Ask them publicly? That’s where it gets quiet.
There’s an odd reluctance to admit it. The fear seems to be that clients will devalue the work, or that it implies the team is less skilled. We think that’s backwards.
Our Stance: Use the Tools, Own It
Imagewize has no issue admitting we use Claude and Mistral Le Chat in our daily workflow. We think agencies should move with the times.
The key is knowing how to use these tools well—not just knowing they exist. That means understanding when AI is the right tool, when it isn’t, when it’s heading in the wrong direction, and how to correct it. None of those steps are possible without real technical knowledge. AI amplifies expertise. It doesn’t replace it. And for clients, that distinction matters.
What This Means for Your Project
When you hire Imagewize, you’re not getting AI output with a logo slapped on it. You’re getting a team that knows its stack inside and out, uses modern tools to work faster and smarter, and applies professional judgment at every step.
AI helps us deliver better work in less time. The benefit passes directly to you—tighter timelines, fewer revision cycles, and lower overall cost.
A Real Example: PHP 8.4 Upgrade
Recently we upgraded our main server to PHP 8.4. The driver was our Sage 11-based Nynaeve theme—its latest Composer packages require PHP 8.4. This involved adjusting Trellis PHP settings, re-provisioning the local VM, testing, and then provisioning production.
Working with Claude on this, two things happened that illustrate the point well.
Where the AI Went Off Track
First: Claude checked the PHP version on the host machine rather than the Trellis VM. Our Mac already runs PHP 8.4 via Homebrew—but the actual development environment lives in the VM, and that’s where the change needed to happen.
Second: Claude initially focused only on updating the Composer requirements, without thinking through the full environment stack that needed to align.
Neither issue was catastrophic. But if you don’t understand the Roots Trellis/Bedrock/Sage architecture, you’d have followed AI instructions that led you in the wrong direction. AI takes the path of least resistance. A senior developer redirects it. That’s the expertise clients pay for—and it’s the part AI cannot replicate.
Know When Not to Use AI
Not every task benefits from AI involvement. Some things are better handled by purpose-built shell scripts.
Routine jobs—creating pull requests from commits, checking for banned IPs, enforcing pattern standards, updating theme versions and CHANGELOG files—are well-defined, repeatable, and fast to automate. Using AI for them is slower, costs more, and introduces unnecessary variance. Build a script once, run it indefinitely.
The same applies to straightforward backend changes. Sometimes a config tweak or quick database update is faster done by hand than explaining the full context to an AI. You develop a feel for this. The question to ask: would setting up the AI context take longer than just doing it?
Claude vs. Mistral Le Chat: Different Tools, Different Strengths
We use both Claude and Mistral Le Chat, and they genuinely feel different to work with.
Claude handles complex, context-heavy coding tasks well—strong on multi-step reasoning and architectural decisions. Mistral Le Chat has a different reasoning style and personality, which means you interact with it differently to get good results.
We’re using Le Chat more and more, especially for writing and lighter coding tasks. It’s improving fast. Having committed to a year, we’re giving it a proper evaluation. Whether we converge on one tool or run both in tandem remains to be seen. Two AIs with complementary strengths isn’t a bad position for an agency to be in.
The Bottom Line: Senior Developers Are Still Essential
AI doesn’t make developers redundant. It makes experienced developers faster.
The agencies that get the most value from these tools are the ones with enough expertise to guide them, catch mistakes, and know when to set them aside. That’s where the real skill gap shows—not in who uses AI, but in how well they use it.
We’re transparent about this because we think honesty builds better client relationships. The work is real. The tools are real. And the expertise required to combine them effectively is very real too.
If you’re looking for an agency that uses every professional tool available—and has the experience to use them correctly—get in touch. We’re ready to take on your project.