Walk your way backwards

When people ask me where to start in tech, I always tell them: look for what’s relevant right now and walk your way backwards.

This has been the strategy that has worked for me time and time again.

Because if you start with the one everyone is trying to figure out now, you’re all on the same plane. While some people will pick it up faster, you get to start with them and then fill in the knowledge gaps.

I believe that is one of the easiest ways to give yourself a chance of earning in tech within 90 to 120 days.

And there’s a reason why. Think about it. If you decide today that you want to become a software engineer, and you start with the traditional roadmap, you’re going to spend months learning things that people have already spent years mastering.

You have to learn programming, understand databases, figure out frameworks, learn how to deploy applications, and eventually start building projects.

And there’s nothing wrong with that.

But by the time you feel ready to start looking for opportunities, you’re competing with people who have been doing those things for years.

Now imagine you decide to learn something that only recently became relevant.

Something companies are trying to figure out. Something experienced engineers are still experimenting with. Something where the best practices are still being written.

Suddenly, the gap between you and the people who have been in tech for 10 years becomes much smaller.

Not because their experience no longer matters.

But because the problem is new, and everyone is still figuring out the best way to solve it.

And that’s where you want to position yourself.

Start with the problem, then learn the technology backwards

One mistake I see people make when trying to get into tech is that they spend too much time learning without understanding what they’re learning for.

They’ll spend six months watching tutorials, another three months trying to decide what to build, and eventually start looking for jobs.

I prefer the opposite approach.

Find something people are trying to solve right now.

Look at the tools being used to solve it.

Build something with those tools.

Then, whenever you encounter something you don’t understand, go backwards and learn the underlying concept.

If you’re building an AI automation and you discover that you need to understand APIs, learn APIs.

If you need webhooks, learn webhooks.

If you need a database, learn how databases work.

If you’re building an AI agent and discover that you need to understand authentication, permissions, and how different systems communicate, learn those things.

You are no longer learning for the sake of learning. You are learning because there’s a problem in front of you that requires that knowledge.

And because you’re applying what you learn immediately, the knowledge sticks differently.

Eventually, the gaps start closing.

And before you know it, you understand far more than the particular tool you started with.

I’m not saying fundamentals don’t matter. They do.

I’m saying you don’t always have to start with them.

You can start with a practical problem and allow the fundamentals to reveal themselves as you work.

So, what should you be learning right now?

If I were trying to break into tech today, especially with the goal of making my first income within the next 90 to 120 days, these are some of the areas I would pay attention to.

And I’m not recommending them simply because they’re trending.

I’m recommending them because businesses are actively trying to figure them out, the tools are becoming more accessible, and you can start building useful things without spending years mastering an entire field.

According to Upwork’s 2026 skills report, demand grew significantly for AI integration, AI chatbot development, and AI video generation and editing.Upwork, In-Demand Skills 2026, published February 2026. Skills explicitly referencing AI grew 109% year over year against 23% for other in-demand skills. AI video generation and editing led at 329%, followed by AI integration at 178% and AI chatbot development at 71%. Worth reading with its limits in mind: the figures cover Upwork’s own marketplace, measure demand originating in the United States, and some categories grow from a small base.

That doesn’t guarantee you’ll find work. But it tells you where some businesses are already spending money.

Here are nine niches I would explore.

1. AI workflow automation

A lot of businesses still have people doing repetitive tasks that could be automated.

Someone fills out a form. Another person copies the information into a spreadsheet, updates a CRM, sends an email, and follows up with the customer.

You can automate a good part of that process.

Start with tools like n8n, Make, or Zapier. Build simple workflows and gradually learn APIs, webhooks, authentication, and how applications communicate.

You don’t necessarily need to know how to code before getting started.

What you can sell: Lead management automations, CRM integrations, customer onboarding workflows, and automated reporting systems.

2. AI customer support and voice agents

Businesses are experimenting with AI systems that can answer customer questions, schedule appointments, and handle routine conversations.

Think about a salon receiving booking calls, a real estate company managing property inquiries, or an ecommerce store answering questions about deliveries.

You can build agents that handle some of these interactions and transfer customers to a human when necessary.

Start by learning conversational design, AI agent platforms, and how to connect agents to existing business software.

What you can sell: Appointment booking assistants, customer support chatbots, AI receptionists, and voice agents.

3. AI video generation and creative production

Not everyone who wants to earn in tech needs to become a programmer.

Businesses need advertisements, product demonstrations, and social media content. And AI video tools are changing how these are produced.

But here’s where I think people get it wrong.

Learning how to generate a video is not the business. Understanding how to create a video that solves a business problem is the business.

Learn AI video generation, editing, storytelling, and creative direction. Then specialize in a particular industry or content format.

What you can sell: Product demonstration videos, advertising creatives, promotional videos, and short form content packages.

4. AI search visibility

For years, businesses focused on ranking on Google.

Now people are also asking ChatGPT and other AI search tools to recommend products, services, and businesses.

Naturally, companies want to understand how they appear in those answers.

This has created growing interest in AI search optimization, sometimes called GEO.

You can start by learning how AI search systems retrieve and reference information, how traditional SEO applies, and how businesses can improve their online presence.

What you can sell: AI search visibility audits, website content optimization, structured information improvements, and brand visibility reports.

The important thing is to focus on measurable improvements rather than promising that an AI platform will recommend a particular business.

5. AI powered internal tools

I think this is particularly interesting for people who want to become software engineers.

A lot of businesses don’t need complicated enterprise software. Sometimes they just need a simple application that solves one specific problem.

An order management dashboard. A reporting tool. An internal application that helps employees find information.

With AI coding tools, you can start experimenting with these much faster.

And as you build, you’ll naturally learn databases, authentication, APIs, deployment, and how applications work.

What you can sell: Internal dashboards, custom business applications, reporting tools, and simple software integrations.

6. AI knowledge bases and document search

Businesses have information scattered across PDFs, Google Drive, Notion, internal documentation, and old support tickets.

And employees spend a lot of time looking for information they already have.

You can build AI systems that search these documents and answer questions using the company’s own information.

This is where RAG, or Retrieval Augmented Generation, becomes useful.

Start by building a simple document search application. Then learn embeddings, vector databases, information retrieval, and permissions as you go deeper.

What you can sell: Internal knowledge assistants, document search applications, and AI powered company knowledge bases.

7. AI agent evaluation and testing

Everybody wants to build an AI agent.

But how do you know the agent actually works?

Getting an agent to complete a task once is very different from getting it to complete that task reliably a thousand times.

What happens when an API fails? When the agent retrieves the wrong information? When a model update changes its behavior?

Someone has to test all of that.

Learn how to create test scenarios, evaluate AI outputs, identify failures, and monitor agent performance.

What you can sell: AI agent testing, chatbot quality reviews, evaluation datasets, and agent monitoring systems.

8. MCP and AI tool integrations

As AI assistants become more capable, people want them to do more than answer questions.

They want them to access information, interact with applications, and complete tasks.

And that means connecting AI systems to external tools.

The Model Context Protocol, or MCP, provides a standard way to make some of these connections.

You can learn how to build integrations that allow AI assistants to retrieve information from databases, interact with internal systems, and perform authorized actions.

Along the way, you’ll learn APIs, authentication, application architecture, and security.

What you can sell: Custom MCP servers, AI tool integrations, internal agent connectors, and workflow integrations.

9. AI agent security

This is one of the more technical areas, but I think it’s worth paying attention to.

As developers give AI agents more access to files, repositories, applications, and business systems, they’re also introducing new security risks.

Prompt injection, excessive permissions, unsafe tool execution, and sensitive information exposure are becoming important problems.

And many teams are still figuring out how to secure these systems.

Start by understanding how agents interact with tools, how permissions work, and how attacks can happen. The OWASP Top 10 for LLM Applications is a useful starting point.

What you can eventually sell: AI application security reviews, agent security testing, permission audits, and security hardening.

This requires deeper technical knowledge than some of the other niches, so I would approach it as a longer term specialization.

---

Now, I know this is a long list.

But the goal isn’t to learn all nine.

The goal is to find one where businesses already have a problem, where the tools are still evolving, and where you can become useful relatively quickly.

If you’re starting without a programming background, I would look at AI workflow automation, AI video production, or customer support automation first.

If you already know how to code, I’d explore AI powered internal tools, MCP integrations, or agent evaluation.

And if you’re interested in security, AI agent security is an area worth building deeper expertise in.

The important thing is to pick one and start solving problems.

Because knowing how to use a tool isn’t enough.

You need to know what problem somebody would pay you to solve with it.

Now, how do you actually start earning within 90 to 120 days?

Give yourself a simple timeline.

Your first 30 days: Learn enough to build something.

Don’t spend the entire month watching videos.

Find a working example, reproduce it, break it, and understand how it works.

By the end of those 30 days, you should have built at least one small project that somebody else can interact with.

Days 31 to 60: Build something for a specific industry.

This is where I think a lot of people miss the opportunity.

Don’t just build a generic AI chatbot.

Build an appointment booking assistant for salons.

Don’t just build an automation.

Build a workflow that helps real estate agents manage inquiries.

Don’t just generate random AI videos.

Create product advertisements for a particular category of ecommerce businesses.

The more specific the problem, the easier it becomes to explain the value of what you’ve built.

And the easier it becomes to identify who might pay for it.

Days 61 to 90: Start talking to people who have that problem.

Find businesses that could benefit from what you’ve built.

Show them your project.

Explain what problem it solves.

Ask them what they currently do and where they’re struggling.

You might discover that the thing you’ve built isn’t exactly what they need.

That’s fine.

Because now you’re learning from people who could actually pay you, rather than guessing what the market wants.

Keep improving the project based on those conversations.

Offer a small, clearly scoped paid implementation when you’ve demonstrated that it can solve their problem.

Days 91 to 120: Deliver, document, and repeat.

If you get a client, focus on delivering the work properly.

Make sure it works.

Make sure they understand how to use it.

Document the results.

And turn what you’ve learned into something you can show the next potential customer.

If you haven’t landed a client yet, look at what you’ve learned from the conversations and refine your offer.

The goal is not to become an expert in everything within 120 days.

The goal is to become useful enough at solving one specific problem that somebody is willing to pay you for it.

And getting there in 90 to 120 days is possible, even though it isn’t guaranteed.

Do all of this in public

Whatever you pick, don’t do it quietly.

Post what you’re building. Post what broke. Post what you figured out when you fixed it.

Not because you need an audience, but because when you don’t have experience to point to, the next best thing is evidence.

A short video of an agent completing a task.

A breakdown of how you automated something.

A post explaining why the first attempt failed.

Each one is proof that you can do the work.

And people don’t hire from résumés alone. They hire from what they have seen you do.

It also shortens the distance between you and the people who have the problem you’re learning to solve.

Because some of them are already watching.

There’s one more thing people often miss

Your first opportunity in tech doesn’t necessarily have to be a full time job.

It could be a freelance project.

A small consulting engagement.

An automation you build for a business.

A tool somebody pays you to customize.

Or a service you provide every month.

Sometimes people are so focused on getting hired by a technology company that they forget other businesses also have technology problems.

And those businesses don’t always care whether you have a computer science degree, whether you’ve completed five courses, or whether you know the history of every programming language.

They want to know whether you can solve their problem.

Of course, as the work becomes more complex, your technical knowledge matters more.

And that’s exactly why I believe in learning backwards.

Because every project exposes something you don’t know.

Every problem forces you to learn something new.

And every time you fill one of those knowledge gaps, you’re becoming more capable.

The tools that are relevant today might not be the tools everybody uses in five years.

But the knowledge you build while solving actual problems doesn’t just disappear.

You’ll have learned how systems work, how to deal with customers, how to build things, and how to figure out unfamiliar technology.

Those skills transfer.

So when people ask me where to start in tech, I don’t necessarily tell them to go and learn a programming language first.

I tell them to look at what is relevant right now.

Find what everybody is trying to figure out.

Start there.

And walk your way backwards.

Because sometimes the fastest way to learn the fundamentals is to first give yourself a reason to need them.