Freelance AI automation for small Indian businesses: a realistic start

Where the repetitive work is

Talk to owners you already know: a family business, a clinic, a coaching centre, a CA's office, a local distributor. Look for tasks with three properties: they happen often, they follow a pattern, and a mistake can be caught before it matters.

Notice the right-hand column. You protect your client, and yourself, by leaving decisions with legal or financial effect to a person.

The discovery conversation

Spend thirty minutes with the owner before quoting. Ask:

1. 1Walk me through the last time you did this task. What did you open first?
2. 2How many times a week does it happen, and who does it?
3. 3What goes wrong most often?
4. 4Where does the information come from — WhatsApp, email, paper, a tally sheet?
5. 5If this worked, what would you do with the time?
6. 6Is any of this information about customers or patients? Who else can see it today?

Write the answers down and send a short summary afterwards. The summary itself is often what wins the job — owners rarely have their own process written down.

Question 6 matters because personal data brings obligations under India's Digital Personal Data Protection Act, 2023 for the business, and you will be handling it on their behalf. Minimise what you touch; see DPDP Act basics for data analysts.

A fixed-scope offer

Keep your first offer to one page. A template:

Project: Purchase-invoice extraction for <Business name>

Problem: About 60 supplier invoices a week are typed into the purchase
sheet by hand (owner's estimate).

What I will deliver:
  1. A process that reads invoice PDFs/photos from a shared folder and
     fills a review sheet: supplier, invoice no., date, taxable value,
     GST, total.
  2. A "check" column that flags rows where the total does not equal
     taxable value + GST, or fields are missing.
  3. A one-page guide for your staff and a 30-minute handover session.

Acceptance: On 30 of your past invoices chosen by you, at least 27 have
all six fields correct, and every row with a wrong total is flagged.

Not included: posting entries into accounting software; handwritten bills.

Timeline: 2 weeks from receiving sample invoices.
Price: fixed, 50% at start, 50% on acceptance.
Running costs: any API or software subscription is paid by you directly;
I will estimate it before we start.
Support: fixes for defects found in the first 30 days.

The acceptance line is the most important part. It stops the project turning into "keep improving it until I'm happy" and it shows the owner you'll measure your own work.

Pricing without guesswork

There is no standard rate, and anyone quoting you a market rate should show their evidence. Build your price from your own numbers:

• Estimate your hours honestly, including discovery, testing on their data and handover. First projects usually take longer than you expect.
• Decide the hourly value you need, then multiply.
• Compare against the owner's side: if the task takes their staff a few hours a week, a price they recover in a few months is easier to accept than one they recover in years.
• Make running costs visible and separate. Owners dislike surprise monthly charges more than a higher upfront price.

The paperwork to check

Before invoicing regularly, get proper advice on your own situation:

• Tax and GST: whether you need to register, and what to charge, depends on your circumstances. Check the CBIC GST portal and a chartered accountant.
• Your employment contract: if you have a job, read its moonlighting, conflict-of-interest and IP clauses first. See the side-project IP check.
• A simple written agreement: scope, payment, ownership of the code, confidentiality, and that you are not responsible for decisions the business makes using the output.
• Accounts and keys: set up API accounts in the client's name where possible, so they own the billing and access if you part ways.

Turning one project into the next

When the first project is accepted, ask for two things: permission to describe it (without the client's data) and an introduction to one other owner with the same problem. Businesses in the same trade often share the same repetitive task, so your second project can reuse most of your first.

The Generative & Agentic AI add-on covers document extraction, retrieval and evaluation — the techniques behind most of these automations. For writing up what you built, see how to write an AI project case study.