AnthropicUnderstand it. Question it. Apply it carefully.
We keep learning so we can choose useful jobs for AI, check the answers it gives and know when your team should step in.
The eleven credentials below were earned by our founder, Okasha Abdirahim, through OpenAI Academy and Anthropic’s Claude Academy. They cover building agents, checking their work, connecting software and understanding where AI can go wrong. Each entry links to the issuer’s public record.
The business examples in this note explain how we would apply that learning. They are illustrations of our approach, rather than claims about a particular customer project or a promised result. A course record provides useful evidence of completed study; the system itself still needs to prove that it works for the business using it.
Choose a useful job. Check the result.
Evaluate AI Applications
Learning how to test AI answers and spot mistakes.
Know what good looks like.
A quick demonstration is not enough. We test common questions, unclear messages and requests that need a person. This helps us check that your agent is useful in the situations your team actually faces.
Can we show that this works on the situations your team actually faces?
Scope AI Solutions
Choosing a useful job for AI and deciding what the work should include.
Choose a problem worth solving.
We ask where your team loses time and what would make the project worth paying for. Sometimes clearer information or a better process should come first. This helps avoid spending money on the wrong thing.
Which specific improvement would make this project worthwhile?
Build with Retrieval-Augmented Generation
Helping AI find answers in information your business has approved.
Give answers a useful source.
Your agent should check your guides and records before answering. We consider who can see each document, how it stays up to date and whether the answer shows its source. If the information is missing, the agent should ask for help.
Where did this answer come from, and is that information current?
Agents and Workflows
Planning the steps an agent can take and when it needs a person.
Delegate with clear boundaries.
An agent might collect an enquiry, prepare a summary and suggest a reply. We agree which steps it can do alone and which need checking. Your team should always know what has been sent and what is only a draft.
What can the agent do alone, and when must it ask your team?
Optimize AI Application Performance
Balancing useful answers, waiting time and running costs.
Make the system practical to run.
An answer that arrives too late can still slow your team down. We check speed, quality and running costs together, and plan what should happen if a tool is unavailable. The aim is help that is practical to use every day.
Does the experience stay useful at the speed and volume your business needs?
Help with the work. Keep people in control.
Claude Academy: Building effective human-agent teams (beta)
Making it clear how people and agents share the work.
Help people and agents work together.
Your team needs to know which jobs the agent handles and who checks them. We start with clear tasks and visible progress so colleagues can understand what it does before asking it to do more.
Does everyone know what the agent owns and who reviews its work?
Claude Academy: AI Fluency for creative work
Using AI to help with writing and ideas while keeping a human final check.
Keep the purpose and the voice.
Customer messages should still sound like your business. We start with who is reading and what they need to know. Your team checks facts, tone and the final message before it is used where approval is needed.
Does this still sound like your business and serve the people reading it?
Claude Academy: Model Context Protocol: Advanced topics
Understanding how an AI agent connects to other software.
Understand the connections underneath.
This course covers a technical way for AI to use tools, called Model Context Protocol. For your business, the useful part is a clearer connection: what information the agent can see, what it can change and how your team is told when something fails.
Can the team understand what a connected tool is doing and recover when it fails?
Claude Academy: AI Fluency for builders
Giving AI clear instructions and checking the work it helps produce.
Own the work from problem to release.
We turn the business problem into things the finished agent must do. We review the work and test it before release. That helps keep the focus on whether your team can use it to do the job more easily.
Have we checked that the thing we built actually solves the original problem?
Claude Academy: AI capabilities and limitations
Understanding why AI can give a confident answer that is still wrong.
Design around what AI can miss.
Your agent should not invent a policy when it cannot find the right document. We plan for missing information and unclear questions. It may need to ask another question or pass the job to someone who can decide.
What happens when the system does not have enough information to answer?
Claude Academy: Claude Code in action
Using AI to help build software, with people checking the changes.
Turn coding assistance into a controlled process.
AI can help with parts of the build. We give it clear tasks, review the changes and test what your team will use. Decisions about the finished product and its release stay with a person.
Can we explain, review and test the change before it reaches your users?
How the credentials were earned.
These are individual course completion credentials earned through the academies’ learning and assessment processes. OpenAI’s five public records name Okasha Abdirahim and show issue dates of 17–18 September 2026. The six Claude Academy verification pages name the same recipient and show 17 September 2026.
OpenAI Academy describes its badges as recognition for completing a course and passing its assessment. Claude Academy awards its completion badges after the required course quizzes are passed. The linked issuer records are the place to check the recipient, course and issue date; we do not publish or infer individual assessment scores.
Both platforms distinguish these learning credentials from professional certification programmes. They are evidence of completed learning, not a business licence, an issuer partnership or an endorsement of Cerebro. Continuing to study, test and review new capabilities remains part of the work.
For the academies’ current completion and assessment policies, see OpenAI Academy courses (opens in a new tab) and the Claude Academy FAQ (opens in a new tab). The OpenAI wallet and all six Anthropic verification pages were checked on 20 September 2026.
A stronger starting point for the work.
For a business owner, the useful question is what this preparation changes. It should lead to a clearer conversation about the problem, a realistic view of what AI can handle and a plan for checking the result. It should also make limitations easier to discuss before they become operational problems.
Imagine enquiries arriving by phone, email and your website. We need to decide what details to collect, where the agent finds its answers and what it can do next. We then test ordinary and awkward requests, and make sure someone can take over when needed.
The aim is less work for your team, without losing track of what is happening. Answers need to be useful and accurate, and someone needs to be responsible when a job needs help.
The starting point is still your business: its customers, its people and the work that needs to improve. Explore keeping teams in step, work between apps or answers for your team to see where that learning can be applied.