Skip to main content
How I Use AI to Run the Back-Office of Busy Entrepreneurs
Top view of a laptop, smartphone, coffee, and glasses on a wooden desk.

You started a business because you had a vision. You did not start it because you love reconciling invoices at 11pm, sorting a 300-message inbox on a Sunday, or rewriting the same caption in four different formats for four different platforms.

And yet, that is where most entrepreneurs end up. Running the business instead of building it.

I work with entrepreneurs who reached that exact wall. They hired me to take the admin weight off their shoulders, and over the last year I have completely rebuilt how I do that work. The short version: I use AI for the mechanical parts, and I stay human for everything that needs a brain with a heartbeat.

I want to walk you through what that actually looks like, because if you are thinking about hiring help (or building your own system), this is the workflow that is working in 2026.

What I actually do for entrepreneurs

Most of my clients are running lean. They are solopreneurs, small agency owners, consultants, coaches, freelancers who grew into a team of three. They do not need a full-time employee. They need someone who can absorb the operational chaos and give them their time back.

The four areas where I move the needle the most:

  1. Bookkeeping and financial admin
  2. Data processing and CRM hygiene
  3. Documentation (reading, summarising, organising)
  4. Social media operations (scheduling, captions, reporting)

Before AI, each of those was a grind. Now, the mechanical 70% is automated, and the judgment 30% is where I add real value. Let me show you what I mean with real examples.

1. Bookkeeping, the part nobody wants to touch

A coach client sends me her receipts at the end of every month. Three years ago, that meant me opening each PDF, typing the vendor, the amount, the category, the VAT. Mind-numbing, expensive for her (I was charging by the hour), and error-prone.

Now: she drops the folder into a shared space. My AI assistant reads every PDF, extracts the data, categorises each expense according to her chart of accounts, and flags anything unusual (a duplicated charge, an unusually high restaurant bill, a missing invoice number). What used to take me six hours takes me forty minutes. I spend that saved time on what actually helps her: calling out expenses that might be deductible but she is missing, spotting a subscription she forgot she had, preparing a clean file for her gestor.

She pays less. I earn more per hour. Her accounting is cleaner. Everyone wins.

2. Data processing, the hidden tax on your time

A consultant client had 2,400 contacts across three different CRMs, a Mailchimp list, and a Notion database. Duplicates everywhere. Titles misspelled. Companies written five different ways ("Acme", "ACME Ltd", "Acme Limited", "acme"). He had been meaning to clean it for two years.

I ran the whole thing through a process where AI merged duplicates, standardised company names, identified the contacts he had not talked to in over a year, and flagged the ones who had recently changed jobs (based on their email signatures in old threads).

The judgment part was mine: deciding which contacts were worth keeping, writing the reactivation email sequence, choosing who to reach out to first. But the grunt work, the part that had been blocking him for two years, was done in a weekend.

3. Documentation, where entrepreneurs drown

Most of my clients get hit with long documents constantly. Contracts from new clients, legal notices, proposals to review, reports from suppliers, terms and conditions from platforms they just signed up for. They know they should read them. They also have twelve other things to do today.

This is probably where AI saves me the most time now. When a client sends me a 40-page contract and asks "is there anything I should be worried about?", I no longer read it cover to cover. I use a tool that converts the PDF to clean text, I ask AI to pull out the non-standard clauses, the payment terms, the termination conditions, the IP assignments, the non-competes. Then I review those specific sections with my eyes, because AI will miss things a human catches.

The result: my client gets a clear summary in hours, not days, and the expensive parts of the contract (the parts that actually need human review) get proper attention.

Important caveat: I am not a lawyer. For anything that carries real legal weight, I recommend my clients consult an actual lawyer. AI speeds up the work. It does not replace professional advice.

4. Social media, the invisible marathon

Scheduling posts is not the hard part of social media. The hard part is writing captions in four different tones for four different platforms, keeping a consistent voice, turning one long piece of content into ten short ones, and reporting what actually worked at the end of the month.

I use AI to do the first draft. I polish it with the client's voice (which I know, because I have been working with them for months). I schedule. At month end, I pull the numbers, and AI helps me spot patterns I would otherwise miss ("posts that ended with a question got 40% more comments"). The client gets a one-page report instead of a spreadsheet they will never open.

The system behind this (and why I am telling you)

A lot of this work got faster when I set up a specific tool called Claude Code on my Mac, combined with a converter that turns PDFs, Word files, and Excel sheets into clean text that AI can read efficiently. That combination is what lets me process a folder of messy documents in minutes instead of hours.

I wrote a step-by-step guide on how to install it, originally for a colleague. If you want to set up the same system yourself, the guide is free to download at the bottom of this post. It is written for people who have never touched a terminal, so do not be scared by the technical words. If you can copy and paste, you can follow it.

When to do it yourself, and when to hire someone

I am going to give you honest advice, because I think most blog posts skip this part.

Set up the system yourself if: you genuinely enjoy tinkering, you have a few hours to commit, and the work you want to automate is repetitive enough that you will use the tool weekly. You will save money and you will learn something useful.

Hire someone (me or anyone competent) if: the setup feels like a distraction from your real work, you do not want to be the person who figures out why the tool stopped working, or you want someone who already knows the judgment part (what to automate, what to leave alone, what patterns to look for). The tool is only half the value. The other half is knowing what to do with it.

Either path works. Just do not keep doing the admin by hand at 11pm on a Sunday, pretending you do not have better options.

Two things you can take with you

Because I want this post to be actually useful, not just something you read and forget:

  1. Free checklist: 10 back-office tasks you can automate with AI this week. A short list of the highest-leverage things to automate first, with examples. Good for entrepreneurs who want to start small.
  2. Free guide: Install Claude Code on your Mac, step by step. The full technical guide (30 to 45 minutes of setup), written for non-technical people. If you want to replicate the exact system I use.

Both are free, both are downloads, both open in a new tab.

👉 Download the checklist (link)

👉 Download the full installation guide (link)

And if at any point you decide you would rather have someone run this for you: get in touch. The first conversation is free, and I will tell you honestly whether I am a good fit for your business or not.