Every new technology promises to save people time. Artificial intelligence is beginning to promise something more ambitious: memory.
Instead of asking chatbots to answer questions or draft emails from scratch, a growing number of professionals are spending hours – sometimes days – teaching AI who they are. They are feeding it years of curriculum vitae, portfolios, writing samples, work histories, career decisions, and personal preferences until the software can represent them almost as consistently as they can.
The goal is not to write better prompts, but to build an AI that understands the person behind the prompt.
For Olanrewaju Habeeb, a Lagos-based communications professional, that shift began with frustration.
Finding remote work had become almost a full-time job. After finishing his day job in public relations, he spent his evenings searching job boards across the United States, Canada, and Europe, researching companies, tracking down hiring managers’ email addresses, and rewriting his CV and cover letter for every application. Most applications disappeared without a response.
“You cannot have just one source of income because of how the economy is,” he says. “Not every company will pay you well. You have to find another way to make money.”
Like millions of professionals, Habeeb initially turned to ChatGPT to speed up the process. Then he experimented with Claude. The chatbots helped him write faster, understand unfamiliar tasks, and tailor his applications to different employers.
Copy-paste
But over time, he noticed something else. As AI-generated applications became commonplace, they also began sounding remarkably alike.
“If ten people use the same prompt and apply for the same job,” he says, “the HR person is reading almost the same thing over and over.”
The problem, he realised, was no longer access to AI. It was differentiation.
Rather than searching for a better prompt, Habeeb decided to build something more personal.
Over several days, he uploaded his work history, CV, portfolio, writing style, career milestones, and professional biographies into a Claude Project, a persistent workspace within Anthropic’s Claude AI that lets users store documents and long-term instructions. He taught the system to distinguish between two different versions of himself: the communications professional he presents to organisations in Nigeria and the SEO writer he introduces to overseas clients. He even trained it to recognise which tone, experience, and CV belonged to each audience.
“I told it to differentiate the two personalities,” he says. “If I ask for the offline person, it gives me that. If I ask for the online one, it gives me that.”
The result is a personalised career assistant.
When Habeeb finds a vacancy, he pastes the job description into a simple web interface he built with Claude’s help. The application sends the request to Claude through its API, rewrites his CV for the role, drafts a tailored cover letter and application email, then estimates how closely his experience matches the position before he decides whether to apply.
His experiment reveals an evolution in how people are using AI.
The first wave of generative AI centred on discrete tasks that people once turned to search engines for. Users asked chatbots to write emails, summarise meetings, explain unfamiliar concepts, or generate ideas on demand. The next wave is becoming far more personal. Rather than treating AI as a tool they consult occasionally, workers are turning it into a permanent fixture in their professional life—one that remembers years of experience, understands individual working styles, and carries out repetitive tasks with increasing context and consistency.
In that future, the competitive advantage may no longer come from knowing how to prompt AI but from teaching AI who you are.
Started with ChatGPT like everyone
Habeeb’s introduction to generative AI was much like everyone else’s. During a three-month internship with a United Kingdom company, unfamiliar assignments regularly landed on his desk. Rather than admit he did not know where to begin, he turned to ChatGPT.
“There was a time during my internship when they would give me a task and expect me to finish it in record time,” he says. “I couldn’t say I didn’t know how to do it. I would just go and disturb the hell out of ChatGPT until I got what I wanted.”
ChatGPT became as much a tutor as an assistant.
“It was like AI was holding my hand.”
By early 2026, he had switched much of his workflow to Claude, whose writing style he preferred. At first, he used it like millions of other people: to edit copy, answer questions, and draft job applications. Then he noticed something unsettling.
Advice on writing the “perfect AI prompts” had become a cottage industry across Facebook, LinkedIn, and YouTube. The same templates circulated endlessly: paste in a job description. Ask the chatbot to rewrite your CV. Generate a tailored cover letter. Repeat.
The results were efficient and increasingly indistinguishable.
Generative AI has dramatically lowered the cost of producing good writing. But when everyone relies on similar models, trained on similar data and guided by similar prompts, quality alone becomes a weaker differentiator. The scarce resource shifts from writing ability to originality.
Habeeb concluded that the problem wasn’t Claude. It was that he was using it the same way everyone else was.
So instead of searching for a better prompt, he decided to build a better memory.
Habeeb’s Eureka moment
The experiment required an unusual amount of patience.
Over three days in May this year, he began constructing what he called My Personal Log, a Claude Project that would serve as a permanent record of his professional life. Into it went years of work experience, portfolios, biographies, writing samples, CVs, and examples of how he naturally introduced himself in different situations.
One instruction appeared again and again.
“Imagine you were me.”
The distinction mattered because Habeeb was not trying to teach Claude a profession.
He was teaching AI about a person.
The distinction illustrates how AI is evolving from a conversational interface into something closer to professional infrastructure.
Early chatbots behaved like consultants hired for a single meeting. Every conversation began with the same introductions because the software remembered almost nothing.
Increasingly, AI companies are encouraging users to do the opposite: create persistent workspaces that accumulate context over weeks or months. Instead of repeating instructions at the start of every session, users build long-term relationships with the software by allowing it to remember their preferences, projects, and ways of working.
Habeeb pushed that idea a step further.
He also wanted to prove that someone with no programming experience could build something like this.
“I know nothing about web development,” he says. “I only prompted the hell out of Claude to show me what to do.”
Working together, they created a lightweight HTML interface.
Open it in a browser, and the workflow appears almost deceptively simple. A job description goes into one box. Habeeb selects whether the application should respond as a communications professional or an SEO writer. Claude’s API then generates a tailored CV, cover letter, and application email using information already stored in the project.
Before producing any documents, the system estimates how closely his experience matches the role. By tech industry standards, the application is relatively simple. Its significance lies elsewhere.
It demonstrates how AI is lowering the barrier between consumers and software creation. A few years ago, building even a basic application required familiarity with HTML, APIs, and debugging tools. Today, people with little or no coding experience are assembling software by describing what they want in plain language and allowing AI to generate the underlying code.
For Habeeb, however, the software was never the point. He was trying to eliminate friction from one of the most repetitive parts of his working life.
The results surprised even him.
After completing the project, he used it to apply for around 15 positions. Eight employers responded. One interview led to a long-term client in the UK who continues to send him work – some accepted, others declined because they were not the right fit.
Compared with the previous years, the difference was unmistakable.
“I used to get feedback from maybe 30% of the places I applied,” he says. “Now I get up to 50%, and I am getting more interviews.”
More importantly, the software changed how he spent his time.
Hours once devoted to rewriting cover letters could now be redirected towards researching companies, preparing for interviews, and deciding which opportunities were genuinely worth pursuing.
The AI has given him back his evenings.
Every task still passes through him
While the improvements have convinced Habeeb that he built something useful, it does not mean that AI can be fully trusted.
Every CV, cover letter, and application email the assistant produces still passes through him before anyone else sees it. Years of working with chatbots have taught him that speed often comes at the expense of accuracy.
“One time it said I had conducted over one thousand interviews,” he says, laughing. “I don’t know where it got that from. I told it not to lie. But it does sometimes.”
The hallucination was harmless. A recruiter would probably have spotted the exaggeration immediately.
But it reinforced an important lesson. The more personal AI becomes, the more responsibility users carry for checking its work.
That is especially true when the system is acting not merely as a writing assistant but as a representative of someone’s professional identity. As more workers teach AI who they are, they are also handing over something more valuable than prompts.
They are sharing years of knowledge.
The workers most likely to benefit from AI may not be those who ask it to think for them, but those who teach it how they already think.
In that future, the most valuable career asset may not be mastering prompts at all. It may be building a digital version of yourself that remembers everything your career has taught you and knows when to put that knowledge to work.
True scale demands moving beyond surface-level integrations to robust execution. We’ve filtered the noise out of Moonshot 2026, optimising the conference strictly for high-calibre connections between startup founders, global financial operators, enterprise leaders and individuals rewiring Africa’s technical frameworks.
Get 20% off Early Bird tickets for a limited time.
