has turn out to be the one only approach for me to resolve problems. Most problems I encounter at work might be solved effectively by utilizing agents. That is in contrast to manually solving tasks or coding up an answer yourself.
In this text, I’ll give a high-level overview of how I approach problems and solve them using Claude Code. As an engineer, you’re essentially an issue solver. Your job ought to be something like:
- Discover and discover a very powerful problems to resolve
- Give you an answer to the issue
- Execute
And this doesn’t just apply to programming tasks. It also applies to tasks in marketing, sales, and customer management. I do know this because that is what I do each day at work as a Data Science Lead of a series-A funded startup.
Why it’s best to solve problems with Claude Code
I all the time think it’s essential to know why you do something. When you take any engineering major at university, this can be a mindset they’ll instill in you.
All the time understand the of what you’re doing
You need to solve problems with Claude Code just because it’s often probably the most effective method to solve an issue. Now, after all, you should use Claude Code alternatives, comparable to Cursor, though I’ll check with Claude Code throughout this text, since it’s the tool I exploit.
Nonetheless, Claude Code isn’t only effective in step 3 of the problem-solving process (executing on the answer). It’s also effective in discovering and identifying problems, together with how essential they’re. It’s also super effective at coming up with solutions to the issues you’ve discovered.
You may, for instance, discover essential problems by:
- Having Claude analyze production logs and inform you of any issues
- Give Claude access to your CRM system to scrub up and notify you of a very powerful areas to concentrate on
- Give Claude access to all of your social media posts and their performance, to research what works and what doesn’t
After discovering issues just like the ones listed below, it’s essential to prioritize them. One among the best tools for that is the worth effort graph.
You just list all problems to resolve on a 2D graph, with the axes being the worth you get from solving the issue and the hassle required to resolve the issue. After doing this, you just pick the issues which have the very best value, relative to the hassle required to resolve them.
Apply Claude Code to find and solve problems
On this section, I’ll undergo the three steps of the problem-solving process, which I highlighted earlier: discover and prioritize problems, give you an answer to the issue, and execute on the answer.
To maintain the article organized, I’ll have one subsection per step of the problem-solving process.
Step 1: Discover and prioritize problems
Earlier within the article, I highlighted some specific ways to find essential problems with Claude Code.
I feel a very powerful consider finding essential problems to resolve is:
Give Claude Code access to all the knowledge you’ve got access to
When you don’t give your coding agents this access, you just can’t expect them to perform well. You may imagine for those who had to resolve a production bug without being allowed to take a look at the logs. After all, you wouldn’t find a way to resolve it, because you can not understand what the issue is, without the logs.
Thus, ensure your coding agent has access to all relevant resources:
- Project management tools like Linear
- Notes from Notion
- Log groups in AWS
- GitHub to take a look at commits
- Browser access to breed issues
And doubtless many more platforms. The purpose is: for those who use the platform to resolve an issue, it’s best to give your coding agents access as well.
Step 2: Coming up with solutions to problems
Okay, at this point, you’ve done the toughest part: You’ve identified a very important problem it is advisable to solve. In software engineering, we regularly say that:
Finding the bug, why it happens, and reproducing it’s the toughest part. Solving it from there is simple
So for those who’ve gotten here, you ought to be excited that you just only have the easy part yet. After all, the way you give you an answer depends lots on the issue you’re trying to resolve.
If it’s a production bug, you’ll be able to often just prompt Claude Code with:
Problem X is occurring in location Y. I consider it’s due to Z. Take a look at the CloudWatch log groups †o understand why the issue is occurring, and give you a plan to resolve it
That is enough for over 50% of the issues I work with, and it really works well to resolve production bugs and implement easy features. If the issue is a little more complicated, you frequently have to iterate several times on the plan Claude Code creates for you, read thoroughly through it, and ensure every part looks correct. This all ties back to general techniques and approaches I’ve previously discussed to get probably the most out of Claude Code.
For other problems, you would possibly have to prompt Claude Code otherwise, or ensure it has access to the proper tools. When you’re, for instance, analyzing your CRM tool, you’ll need to offer API access to the tool and supply Claude Code with all of the access it needs.
Normally, I urge you to be liberal with the access you provide your coding agent, because without proper access, it simply cannot solve problems for you. This is strictly the identical as for those who were to ask humans to resolve an issue. In the event that they don’t have access to the proper platforms and tools, you can not expect them to be good problem solvers.
Still, nonetheless, it’s best to after all ensure the agent cannot perform any destructive actions. When you’re giving it AWS access, for instance, it’s best to probably log in with Viewer access only, as a substitute of an admin user. Normally, you’ll be able to simply follow good security practices.
Step 3: Executing on solutions
The last a part of the problem-solving process is to execute the answer. When you’ve give you a very good plan with Claude Code, you’ll be able to simply tell it to execute on its plan.
Again, this works for a majority of the issues I apply Claude Code to. It’s gotten so good that it one-shots the answer.
Nonetheless, I still encounter more complex problems where this doesn’t work. Sometimes it is advisable to iterate several times with Claude Code. For instance, once I create marketing material comparable to:
- Scripts for webinars
- PDF carousels for LinkedIn
- Posters
I often have the agent make an initial design and begin reviewing and iterating on it. Normally, that is an awesome technique when working on complex problems. Give you a good initial solution, test it, review what works and what doesn’t, and iterate. By simply repeating this just a few times, you’ll often find yourself with an awesome solution.
Conclusion
In this text, I’ve discussed how I apply Claude Code to resolve problems. Normally, my mindset is that at any time when I encounter a brand new problem, I feel: “How can I apply Claude Code to resolve this problem for me. Loads of times, you’ll realize that you may solve an issue with a straightforward prompt. Other times, it requires more iterations, either to find the proper problems, prioritize the issues, create a plan to resolve the issue, or to execute on the issue. Normally, I strongly urge you to have the mindset of iteration in place, where you begin off with a good solution, and iterate on it until you’re completely happy with it. Most problems don’t require an ideal solution.
👉 My free eBook and Webinar:
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free Vision Language Models ebook
💻 My webinar on Vision Language Models
👉 Find me on socials:
💌 Substack
