How to find and start your first AI coding project

The AI Daily Brief explains that AI-assisted coding is no longer just for software engineers; it's a key productivity skill for all knowledge workers. The host outlines a framework for identifying 'software-shaped problems' within your daily work. Business operators can use this to decide whether to automate an existing task, upgrade a current tool, or invent a new solution, and then select a high-value first project.

In a recent episode, the AI Daily Brief presented a practical guide for non-developers to start using AI for coding and automation. The core idea is that any knowledge worker can learn to identify and solve 'software-shaped problems'—repetitive, rule-based tasks that can be addressed with simple scripts or applications built with AI assistance.

Identifying software-shaped problems

The first step is to develop an eye for processes that are ripe for automation. These often involve:

• Repetitive data entry or transfer between applications.
• Manual report generation from structured data (e.g., compiling weekly sales figures from a CSV file).
• Sorting, filtering, or triaging information (like customer support emails or internal requests).
• Any task where you find yourself repeatedly following the same set of digital steps.

A framework for action: automate, upgrade, or invent

Once a problem is identified, the host suggests a three-part framework to decide on the best course of action. This helps to scope the project and clarify the goal.

For beginners, the recommendation is to start with automating a small, annoying, and frequent task. This provides a clear goal and a high chance of a quick, tangible win.

The concepts presented in the podcast signal a fundamental shift in how businesses can and should operate. This isn't about turning your entire team into software engineers; it's about empowering them to solve their own problems, leading to significant gains in productivity and operational agility.

The democratisation of development

Modern AI models like Claude 3.5 Sonnet and ChatGPT-4o are powerful coding assistants. They can take a plain-English description of a task and generate functional code in seconds. This dramatically lowers the barrier to entry, enabling 'citizen developers' within your organisation—employees who are closest to a business problem and can now build the solution for it.

Compounding productivity gains

An automation that saves an employee 15 minutes per day may seem minor. But across a year, that single automation reclaims over 60 hours of productive time. When multiple team members start building these small solutions, the cumulative impact on your business's efficiency can be enormous. It shifts the focus from manual execution to system design.

From passive software user to active workflow builder

Historically, if a niche operational problem existed, a business had to wait for a vendor to create a commercial solution or hire expensive developers. Now, your team can build bespoke, 'good enough' internal tools to solve these specific problems. This allows your organisation to become more adaptable and efficient, tailoring its own tools to its unique workflows rather than changing workflows to fit generic software.

Building capability with low risk

Starting with small, internal automation projects is the ideal way to build AI literacy and capability within your team. These projects are low-risk—if they fail, the impact is minimal. But when they succeed, they not only deliver productivity gains but also equip your staff with the skills and confidence to tackle more complex agentic AI challenges in the future.

Business owners and operators should treat this not as a technical exercise, but as a strategic initiative in operational excellence. Here are five practical steps to get started:

1. Conduct a 'process audit'. For one week, ask your team to keep a simple log of their most repetitive, time-consuming digital tasks. Look for patterns involving spreadsheets, copy-pasting between tabs, manual report creation, and data entry.

2. Host a 'problem-finding' workshop. Gather your team to discuss the findings from the audit. Use the Automate, Upgrade, Invent framework to categorise the identified tasks. Focus the group on identifying the single most annoying and frequent task that could be automated.

3. Select a pilot project and a champion. Choose one simple 'Automate' task as your first project. Its goal should be clear and measurable (e.g., "auto-generate the weekly marketing report from our sales CSV"). Assign it to an enthusiastic team member who is keen to learn—their curiosity is more important than any prior coding experience.

4. Equip your champion for success. Provide them with a subscription to a capable AI assistant like ChatGPT Plus or Claude Pro. Encourage them to start by describing the task in plain English, iterating on the prompts, and testing the code generated by the AI. Frame it as an experiment, not a pass/fail test.

5. Document and share the win. Once the first automation is working, have your champion document the process, the final script or tool, and—most importantly—the time saved. Share this success story internally to demystify the process and inspire other team members to identify and build their own automations.