Delivery systems: projects that run smoothly

0:00 Delivery systems

Clients rarely leave because of one mistake. They leave because of a pattern: missed deadlines, having to chase for updates, surprises at the last minute, work that needs checking twice. The good news is that reliability isn't a personality trait. It's a system. In this lecture, you'll learn a simple delivery system for freelancers and small teams: planning and project boards, a communication rhythm, managing feedback and revisions, quality control, including a QA gate after every AI step, handover and offboarding, and how to handle mistakes when they happen.

0:39 Why it matters

Why does this matter? Because reliability is your reputation, and reputation drives referrals, renewals and rates. A client who never has to chase you will happily pay more and tell others. And in 2026, there's a new reason. AI can speed up delivery, but it also creates new ways to fail: confident errors, inconsistent style, licensing questions and data leaks. Here's the key idea. Design your delivery so quality doesn't depend on remembering. Checklists and rhythms remember for you.

1:13 The pilot analogy

Here's an analogy. Think of airline pilots. They're highly skilled, yet before every flight they run a checklist, and during the flight they report at set points to air traffic control. Why? Not because they're forgetful, but because the stakes are high and routines catch the rare mistake that skill alone misses. Your delivery system is the same. A project board is your flight plan. Weekly updates are your radio calls. A QA checklist is your pre-landing check. And an autopilot, like AI, is useful, but you still check the instruments and keep your hands near the controls.

1:56 Plan and communicate

Let's build the system. First, planning. Break each project into milestones and tasks on a board, in Notion, ClickUp, Trello or Asana, with owners and dates. Include client tasks too, like supplying content or approving drafts, because client delays are the most common reason projects slip. Next, a communication rhythm. A weekly written update on the same day each week: status, done this week, next week, what you need from them, and any risks. Use one agreed channel per client. And short async videos, for example with Loom, are great for walking through drafts without scheduling a call.

2:39 Feedback and QA

Now feedback and revisions. Consolidate feedback: ask the client to gather all comments from their team into one place, in one round, by a set date. Give them a structure: what's working, what must change, and what's a nice-to-have. Tie revisions to your contract's definition, so a revision is a change within the brief, not a new concept. And keep versions labelled clearly. Then quality control. Before anything leaves you, run a QA checklist: matches the SOW, facts checked, brand and style followed, accessibility basics, lawful claims, licensed assets, working links and permissions.

3:19 Worked example 1: Tom in Bristol (illustrative)

A simple worked example, illustrative. Tom, a freelance developer in Bristol, kept missing deadlines because clients sent feedback by email, WhatsApp and phone, and content arrived late. He changed three things. A shared board where client tasks, like send product photos by Tuesday, sit next to his own. A Friday update with a needed-from-you section. And one consolidated feedback round per milestone, in a shared document. Within a month, his projects stopped slipping, and when they did, the board showed exactly why, which made those conversations easy rather than awkward.

3:58 AI in delivery

Now AI in your delivery workflow. It can speed up research, first drafts, transcripts, resizing, code scaffolding and test generation. But build it in deliberately. Mark every AI step in your standard operating procedures, with the approved tool, the saved prompt and the inputs allowed. Add a QA gate straight after each AI step: fact check, brand and style check, licence check for images and code, and a legal claims check in sensitive sectors. Keep a short error log. When something goes wrong, note why and fix the prompt or the process. And always respect client restrictions agreed at onboarding.

4:41 Worked example 2: Nadia in Lahore (illustrative)

Now the realistic scenario, illustrative. Nadia runs a three-person content studio in Lahore producing blog posts and newsletters for UK software companies. She introduced AI drafting to handle volume. Within weeks, a client spotted an invented statistic in a published article. Embarrassing. Nadia responded well. She apologised, corrected the article the same day, and explained the fix. Then she added a QA gate: every figure must link to a source, checked by a second person. She logged the error, updated the drafting prompt to forbid statistics without sources, and told clients about the new check. The client renewed, citing how she handled it.

5:26 Watch me: the QA checklist

Watch me run the QA checklist on a deliverable. It's a landing page for a clinic. Does it match the SOW item? Yes, one page, English and Arabic. Facts and names? I check the doctors' names and qualifications against the client's list. One is misspelled. Fixed. Brand and style? Correct colours and tone. Accessibility: alt text on images, readable contrast. Claims: the draft says guaranteed results, which is risky in healthcare advertising, so I change it to wording the client can support. Licensed assets: photos from the client and a licensed library. Links and forms: tested. Two issues caught before the client ever saw them.

6:12 Handover and mistakes

Next, handover and mistakes. At the end of a project, hand over properly: final files in agreed formats, a short guide or recorded walkthrough, credentials and access transferred or removed, and a summary of results and recommendations. Then ask for feedback and a testimonial. And when you make a mistake, and everyone does, follow a simple pattern. Tell the client early, before they find it. Take responsibility without excuses. Fix it, or propose options. Explain what you'll change so it doesn't happen again. Handled well, a mistake can actually strengthen trust.

6:52 Common mistakes

Let's list the common mistakes. No single board, so tasks live in your head. Updates only when the client chases. Feedback from five people in five channels. Unlimited revision rounds. Sending work without a QA check. Letting AI output go out unreviewed. Ignoring client restrictions on AI or data. Overbooking, because you didn't plan capacity. And hiding mistakes, which turns a small problem into a lost client. Plan capacity honestly: know how many projects you can run well at once, and leave a buffer for the unexpected.

7:30 Recap and try this now

Let's recap. Reliability is a system: a board with client tasks included, a weekly update on the same day, consolidated feedback, defined revisions, and a QA checklist before anything is sent. Build AI into delivery deliberately, with marked steps, a QA gate after each, and an error log. Hand over properly and handle mistakes openly. Your try this now: set up one board for your current project with client tasks included, send your first weekly update using the template, and add the QA checklist to your process today. Next, we'll get you paid, including internationally.