AI Workflow Automation
Automate your everyday tasks with AI
Turn repetitive work into workflows you can run again—with Codex, ChatGPT, Claude, and the tools you already use.
Bring a task you keep doing by hand. We’ll map the steps, build a small working version, and test what should be automated and what still needs your review.
Free 15-minute consultation. Coaching scope and price agreed before booking.
One task. A repeatable process.
Start with the work, not the tool
“Every week I combine exports, check the numbers, and write the same report. How can I stop rebuilding it from scratch?”
INPUT
Approved files, a repeatable format, and clear rules.
WORKFLOW
Check the data, prepare the output, and flag exceptions.
REVIEW
You check the result before it is shared or acted on.
Illustrative workflow—not a client result or a promised time saving.
Which task keeps taking up your time?
Start with something frequent, clearly defined, and easy to check. These are possible learning exercises; suitability depends on your files, permissions, and current tools.
Files and documents
Organize downloads, rename batches of files, or prepare documents from approved source material. Preview proposed changes first and keep the originals.
Reports and spreadsheets
Combine recurring exports, flag missing or duplicate records, calculate totals, and prepare a draft summary. Use fixed rules for the calculations and review the commentary.
Routine business admin
Turn approved notes into action lists, prepare recurring project updates, or draft follow-ups. Keep the source attached and review before sending or changing shared records.
Example: your recurring report
An illustrative exercise using permitted CSV exports. The aim is a draft you can verify—not a report that hides problems or sends itself without approval.
Today: repeat the same preparation
Download the approved exports and collect them in a folder.
Copy rows together and try to spot missing or repeated entries.
Recalculate the same totals and reformat the report.
Write a summary, check it, and send it manually.
A pilot: prepare, flag, and review
Run the workflow on copies. It checks expected columns, applies agreed duplicate rules, calculates totals, and saves a draft report plus an exception list. The original exports stay unchanged.
Review flagged records and the summary against the source. Only then decide whether to share it. If inputs change or a check fails, stop and report the problem. A recurring schedule comes after a successful repeatable test.
From a manual task to a repeatable workflow
Work through one useful process with Dejan, using examples you are permitted to share. The goal is something you can understand, check, and run again.
01
Map the task you do now
Show the input, the steps you repeat, and a correct finished example. Identify how often it happens, which decisions need judgement, and what must never change.
02
Build a small first version
Choose the simplest suitable tool. Use sample data or copies, define the output, and add a preview or draft step before enabling changes to real files or systems.
03
Test the awkward cases
Try missing inputs, duplicates, a changed file format, and an interrupted run. Check totals against the source. Flag problems instead of silently inventing values or producing a misleading result.
04
Make the next run manageable
Save the instructions, setup, and checks. Decide who reviews results and how to pause or recover. Test repeated runs before adding a schedule, and avoid creating duplicate outputs or actions.
Keep a short operating guide: where the inputs go, how to start the workflow, what a correct result looks like, and what to do when it fails. Compare the actual time and review effort with your original process.
Choose the right tool for the task
We start with your existing workflow—not a requirement to buy every AI subscription. Sometimes a small script or your app’s built-in automation is enough.
Codex for scripts and repeatable work
Use Codex to help write, test, and refine a script for a clearly defined task, such as validating CSV exports or preparing files. Review the changes and commands within an approved project before running them on real work.
ChatGPT for information and task workflows
Practise turning permitted source material into structured notes, reports, and drafts. For recurring work, check whether Scheduled and the required connected apps are available. Connecting an app does not automatically create a working automation.
Claude Code for custom workflows
Explore scripts and small utilities with Claude Code. Inspect files, follow changes, run checks, and investigate errors. Agree on the input, output, and limits before letting a coding tool modify your project.
Claude for supported work tasks
Explore document and connected-app workflows with Claude, including Cowork where available. Confirm the required access and whether the task depends on local files or cloud connections. Start with low-risk outputs you can review.
Keep fixed rules predictable
Use code or existing app rules for exact calculations, naming formats, and duplicate checks. Add AI where interpretation is useful, such as drafting a summary. Not every step needs an agent.
Decide where it runs
A workflow you start yourself, a scheduled run, and an app-triggered action are different setups. Confirm the runtime, permissions, source access, and failure notifications before relying on unattended work.
Capability references checked on 22 October 2026. Features, access, and scheduling requirements depend on the tool, plan, and workspace settings.
Build the workflow together
Choose private coaching around your own task or a workshop for a team process. We agree the pilot, required access, scope, and price before booking.
1:1 workflow coaching
Request a quote
Work on one recurring task with practical guidance at your level. We agree a manageable pilot before the first session.
Your task and approved examples
Guided build and test
Clear review and recovery steps
Instructions for the next run
Team automation workshop
Request a quote
Map a shared process and explore a small pilot with the people who use it. Company permissions and responsibilities are agreed first.
A shared process and clear owner
Approved tools and sample data
Human review and failure handling
A separately scoped team session
Software subscriptions, usage credits, and hosting are separate. This is training and guided setup—not an unlimited implementation or ongoing monitoring service. Existing Figma package rates do not automatically apply.
AI automation questions
What can run automatically, where Codex fits, and what should stay under your control.
What is AI workflow automation?
Can Codex help automate tasks outside app development?
Do I need coding experience?
Will the workflow run when my computer is off?
Does every step need an AI agent?
How do we protect company or client information?
Will it send messages, delete files, or change live records?
Is this training or a done-for-you automation service?
What affects the price and what should I bring?
Choose your next learning path
This page focuses on work you already repeat. Explore the broader AI training page for tool introductions, prompting, and app-building lessons.
