AI is My New Favourite Teammate

No Code vs AI Agents

Last week I created an end-to-end marketing campaign using ChatGPT. This took time to set up, but saved me the annual subscription cost to at least one software service. What surprised me, however, was how much fun it was interacting with AI.

The campaign setup involved Python code running in Selenium, detailed knowledge of how AI augments Building Information Modelling and preparation of infographics. I am a capital markets specialist whose coding experience ends with simple Excel macros. All of this is new to me.

AutoGPT points out that AI agents are proficient at “in between tasks”. These are sub routines necessary to get work done, but not complete workflows. To delegate to AI while delivering outcomes, you need it to be expert, correct errors and work in the background.

The type of tasks AI agents perform well include web search and scraping, filling out forms, and tailoring your C.V. to job applications. Standard, repetitive tasks may be automated with no code applications. Repetitive tasks with changing data require coding. But this does not mean that you need coding skills.

What AI Can Do Today

Over the past two weeks I summarised how to run an AI project in a series of LinkedIn posts. By far the most popular was the last post about AI governance. The full guide is here.

People are looking for guidance on how to bridge the gap between AI’s potential and what they can do with it today. They are hoping regulators can tell them. A better answer is it’s a new, fun teammate.

Recent research from Cornell University revealed that human-AI teams communicated almost half as much again as humans alone. The AI handled much of the drafting and refinement work, while humans concentrated on content creation. This mirrors my experience, where focusing on outcomes rather than routines makes work enjoyable.

Creating My Campaign

There are over 40,000 commercial construction companies in the UK. Most are small and have no AI budget. I needed to find out from Companies House, which of them has at least £25 million of revenue.

First up, I had ChatGPT write a Python script to identify commercial construction companies. We then searched the records to see which companies published full accounts. The third step was to download these records.

ChatGPT realised that Companies House filing histories don’t link directly to pdf accounts. We must navigate an interim page. Only then can we plug in the revenue extractor, download and order the data, and export it to my CRM.

You can do this task with SaaS such as Endole or DueDil. My ChatGPT subscription costs less however, and does much more. On this project alone I had it list the benefits of enhancing BIM with AI, to ensure I didn’t miss anything, and mock up infographics to send to prospects.

ChatGPT is My Favourite New Teammate

Despite having no coding background, I now have a series of Python scripts running in the Selenium environment, with advanced options in JavaScript. I understand what the code does because I specified its function. I relied on AI to correct its own errors and figure out why Companies House API was unresponsive.

While I am delighted by the outcome, it was my interaction with ChatGPT that was most revealing. I text back and forth with colleagues around the world a lot, on email, Teams and WhatsApp. Our conversations are interrupted by time zones, conflicting priorities and frustrations when people don’t understand each other.

In contrast, ChatGPT is a patient and constructive conversationalist, as well as an expert in what I want doing. Our interactions feel like talking to my favourite colleague, who is always there when I need them. It adapts to my dumb questions and adjusts recommended actions as it learns how I like to work.

I have the commercial subscription to ChatGPT, as the free version was limiting my conversations. I read that Claude is better for coding and Gemini 2.5 Pro is the best experience. I can’t be bothered to keep switching however, and am delighted with my results.

The campaign will launch once we’ve ironed out an issue with our CRM. This software is on borrowed time given the growth of agentic AI workflows. In the meantime, if you know a largish, UK construction company, which wants to accelerate its BIM with the latest NVIDIA kit, spot design mistakes as they happen, and use a digital twin for real-time collaboration between engineers and site workers, please send them my way.

Questions to Ask and Answer

1. What are my repetitive, standard tasks that can be automated with no code apps?

2. What are my repetitive tasks handling changing data that require agentic AI?

3. Do I know any large construction companies I can refer to Simon?

Here are 3 ways I can help

1. Book a consultation to talk about AI.

2. Read the deep dive into data science.

3. Explore use cases using accelerated computing.