From OpenRouter to Github Copilot, these were all just buzzwords until they came up in that one All Hands. Then, all of a sudden, an email arrives. It mentions something about AI credits and “tokens” and watching your “usage”, and provides some vague links to an API key. Then nothing. No instructions, no practice sandboxes, just a new currency that you have no idea what to do with.
Let me try and clear the air for you. It’s easy to let credits sit unused, or conversely, to burn them on novelty prompts like asking an agent to evaluate your entire codebase just to make a one-line change. If you’re new to AI credits, here’s a breakdown of how to get real value from them.
Start with your most tedious work
The best first use of AI credits can be the work that you find most tedious, not the work that sounds impressive. Think about the tasks that can easily eat an hour of your time but don’t really utilise your full brain power. Those could be:
- Turning messy meeting notes into clean minutes and action items
- Drafting the first version of a specification, report, or email you’ve written several times before.
- Scaffolding a test suite that follows a similar pattern to the ones that came before it.
AI can be thought of as a partner in crime code, and a tool. You don’t have to take its advice. The value is in having something to build upon.
Learn something faster
Credits gave me a good excuse to pick up a skill I had been putting off, and for the record, learning to integrate LLMs into your workflow is a skill as well! Ask for an explanation of a concept at your level, then ask follow-up questions until it clicks, then ask it to quiz you afterwards. At its basic level, it can help you upgrade your writing and communication by offering rewrites and alternative versions of your pasted texts in different tones and levels of comprehension. It’s up to you to steal the best bits.
Automate the repeatable
If you notice you’re giving AI the same instructions repeatedly, save them as a template (typically an agent.md file). A good prompt for PR reviews, feature implementation, or even simple scripts, is a small asset with great returns. Over time you build a personal toolkit, and credits spent on refining it pay off again and again.
Get good at the craft of asking
Also known as prompt engineering, it is a given that better inputs correspond to better outputs. Consequently, it would be worthwhile if some credits are spent on the art of communication itself. A few habits help:
- Give context. Provide context on your codebase/document, explain who the audience is, what the goal /objective is, and what good looks like.
- Show examples. One sample of the coding convention, paradigm, or style you want can often be more than enough context to get the ball rolling.
- Iterate. Treat the first answer as a draft and refine it. I insist on using plan mode in Opencode before doing any implementation.
- Request questions. I like telling AI agents to ask me for clarification if anything I have given is unclear. A simple “What else do you need to know before you start?” often improves results dramatically.
Know the guardrails
Before you paste anything in, check your company’s privacy policy. Usually, client data, unreleased documents, personal information, and anything confidential may be off-limits for sharing with AI tools, or may need specific approved permission. Also, remember that AI can be confidently wrong. Verify facts, figures, and citations, especially for anything that leaves your desk, and always check its output before using it in production or sharing it with others.
Don’t hoard them, and don’t waste them
Credits that expire unused are a wasted opportunity, but using them uneconomically is just as bad. If you are just starting out, a sensible approach would be to pick two or three real tasks this week, try using AI on each, and note what worked. Don’t forget to share the wins with your team, because your discoveries will save someone else’s credits and time.
In short…
AI credits are a low-risk invitation to experiment. Start with something boring, build up to something ambitious, and pay attention to where it genuinely saves you time or improves your thinking. Use them both personally and professionally (if you have permission). By the time the balance runs low, you’ll know exactly where AI belongs in your work, and you’ll have a solid case for asking for more and this time – the evidence to show for it.
Happy coding! Until next time,
Nkeiruka.
#AI #Productivity #SoftwareEngineering #DeveloperTools #GithubCopilot #LifeAtCanonical
Leave a Reply