Work

How I actually work

These situations come from running an independent freelance Generative AI business on Fiverr. Client work itself is confidential, so what follows isn't a gallery of finished output. It's the process behind it — the kinds of situations that come up, and how I've handled them. The specific client, brief, and deliverable have been generalized throughout; the decisions and reasoning haven't.

Working principles

How I begin, and how I keep going

How I begin

By asking what "done" actually looks like before starting anything — a surprising number of problems are really unclear definitions of success.

How I gather requirements

By asking specific, closed questions rather than open-ended ones — "should this be A or B" gets a faster, clearer answer than "what do you want."

How I reduce ambiguity

By stating my assumption out loud before acting on it, so it can be corrected early and cheaply instead of discovered late.

How I communicate

Plainly, and in writing where it matters — a clear "no" or "not yet" is more useful to someone than a vague "maybe."

How I deliver

In checkable steps rather than one large reveal at the end, so a wrong turn is caught early rather than at the deadline.

How I learn and improve

By writing down what went wrong immediately after it happens, while the specific reason is still clear — not from memory a week later.

In practice

Five kinds of situations, generalized

A brief that didn't say enough

A project arrived with a general direction but no clear definition of what a successful result would look like.

Starting work on an unclear brief risks building the wrong thing carefully rather than the right thing quickly.

I asked a short set of specific questions to pin down what "good" meant for this particular request, rather than proceeding on assumption.

I shared a small early version for confirmation before producing the full piece, so any misunderstanding surfaced while it was still cheap to fix.

The brief became clear before real work began, and the final delivery matched expectations on the first pass.

It's faster to slow down at the start than to fix a misunderstanding after the work is done.

A tight turnaround

A client needed a fast delivery, well outside the timeline that would normally allow for a slow, iterative process.

Speed and quality usually trade off against each other, especially without a team to split the work across.

I identified the one element that would be hardest to fix late, and made sure that part was right first, before polishing anything else.

I worked in small, checkable stages and confirmed the highest-risk part early rather than assembling everything at once.

The delivery landed on time without needing a rework cycle afterward.

Speed under pressure comes from a repeatable process, not from working faster in a panic.

A revision that could have spiraled

A client came back after delivery with general dissatisfaction, without pointing to anything specific that was wrong.

Vague feedback can turn into an open-ended revision cycle if it isn't narrowed down first.

I asked which specific part wasn't landing, rather than guessing and revising broadly.

Once the actual issue was identified, I addressed only that, instead of reworking the whole piece.

A single, focused revision resolved it, and the working relationship stayed positive.

Most revisions are a communication problem before they're a technical one.

A repeat client with a different need

A client who'd worked with me before came back with a request that didn't fit the same shape as the previous project.

It's tempting to reuse the previous approach on autopilot rather than treating the new request on its own terms.

I re-scoped the conversation from the start, as if it were a new engagement, rather than assuming the old terms carried over.

Pricing and expectations were reset explicitly for the new request before work began.

The relationship continued rather than being strained by mismatched expectations from a previous project.

Repeat trust is built by treating each new request as its own conversation, not by assuming it.

Running several engagements at once, alone

Multiple client engagements were active at the same time, with no team to divide the load.

Without a shared system, details or deadlines are easy to drop when everything depends on one person.

I kept a simple, consistent way of tracking commitments and deadlines — the same instinct behind writing maintenance procedures, applied to running things on my own.

I sent short status updates on my own initiative rather than waiting for a client to ask where things stood.

No deadlines were missed across concurrent engagements, without a team behind me.

Working alone reliably depends on process discipline, not on effort or memory.

Day to day

Tools, briefly

Generative AI tools for the work itself — including prompt development for AI image and video generation, and models like ChatGPT, Claude, Gemini and Grok for research, documentation and workflow support — plain documentation for keeping process and commitments straight, and ordinary written communication for everything client-facing. None of it is the point — it's what the process above happens to run on.

This is the process. If it's relevant to a role you're hiring for, start a conversation.