In this article, I’ll talk about how I’ve seen Ai catch up staggeringly quickly in the area of Delivery, with relevant coherent & useable output from LLM’s like ChatGPT. This is not a guide on what Ai is, nor a walkthrough or tutorial on how to use. Go elsewhere for that.

I wanted to write very specifically about something I’ve witnessed over just the last quarter. A phenomenon. If you’ve been living under a rock, you may have not heard of the already-upon-us Ai revolution, & I’ve been prodding at its underbelly for quite some time as a tech professional. As a novelty with huge potential, I’ve enjoyed playing with it & observing how society takes it on. The exponential progress of GPT & the misunderstanding by the general public on what.it.is, particularly entertaining.

My work as a consultant has meant I’ve got to experience how many different companies use, interact & talk about Ai along with whether it will fit into their business. Up to three months or so ago, it was still all very speculative. Like ‘Agile’ Ai is literally talked about by everyone (now including me) but it’s often-pontificating whimsical bullshit designed to sell some kind of product offering rather than anything insightful. Opinions on Ai then are a bit like arseholes. Everyone has one!

Experiencing a new organisation just recently & I’m amazed at how ‘use ChatGPT for it’ is a commonplace response. Not only for development though, where I’ve spoken directly to engineering teams who use it to augment & better their code but to everyone. People are using it not just to make their emails better, more succinct, write their CVs or even write test plans, project plans & test scripts but they are using it as a genuine bonifide talent replacement. I’m not sure how I feel about that. Yet.

I was until this quarter a believer in not relying on a tool without first mastering the skill to do that thing ‘manually’. I’m proud of my experience in what I do & how to apply it but I have been starting to test the latest v4 of GPT to compare it side by side with my work. In fact, at the time of writing this, I’ve signed up to an twenty bucks a month subscription to GPT4x. That’s how serious I’m taking it.

It’s always difficult to explain to people what I do. Sometimes even I don’t know how to best explain life as a Delivery Manager working in tech. I’ve honed so much good practice & process over the years as I’ve built my experience, I can often roll this out without much change project to project, company to company.

I’ve now starting using ChatGPT to do things I could quite easily do myself but now with a perverse interest of whether it can do it better or more useful, or quicker. Yes, definitely the latter!

It excels for me in writing test approaches, acceptance criteria & DoR’s (definition of Ready) and can give me a quick guide into an approach to solve a project problem, issue or snag. I’ve started putting draft emails through it & asking it to curate a response, better, more succinct, more professional. Like a PA. The game changer however is using plug ins and integration tools to make my interaction with an Ai more frictionless. Up until these integrations I’d have to open a browser window, log into openai, and craft a prompt. Allowing the Ai access in-line with my information I want to process, questions I want to ask, plans I want to make allows an incredibly powerful tool I’m worried people will rely on. Very quickly however I’m starting to struggle to find issues with what it (the Ai) recommends or suggests. I’m waiting for the catch but I fear there isn’t one.

Clients & businesses are starting to consider different things. My friend Paul runs a Ai start up called unfold.ai offering counsel & support for organisations who don’t know where to go to start exploring Ai. Clients are starting to want more & more data rationalisation & consolidation projects so they can a) get their data in one place and b) use data science augmented by powerful Ai to start getting actionable insights on that data. I’ve worked on this last year;

• Data warehouse technology
• LLM and NLP products (especially in the customer service space for the latter)
• Ai driven development using tools like Copilot & Code Whisperer from Amazon    
• Cost reduction – its definitely coming – why employ someone to do a bunch of tasks when someone who can drive Ai & has only enough contextual enough understanding as necessary can be hired for far less. Not direct replacements for whole engineering teams, but Agile coaches, Engineers who suck, yeah maybe you’re in trouble.
  

The cost of advancement.

The issue with Ai, a bit like having anything nice, is it will get abused. Over the years I’ve dealt with client’s shadow IT. Where fantastic tools such as Microsoft’s PowerApps puts development power into the hands of ill-trained exec. In comes the consultancy team to untangle it all. A new PMO office set up to do Agile without the experience. Queue a sea of post-it notes & a badly implemented Atlassian suite. And now, Ai. I can’t even begin to think about how this is going to get abused. I mean, have you used software recently? Any? I recently installed the BA app for my upcoming jaunt to Silicon Valley. Its garbage, loading up the web page into a mobile compliant frame which passes you straight to the website anyway. Europcar’s booking system & badly integrated website is woeful (I tweeted them here). Don’t get me started on the lack of joined up’ness of NHS systems. I’m not saying Ai is responsible for any of that yet – buts it’s a bad state of affairs already. And then? We’ll start using Ai to write great hunks of code with constantly demotivated & de-powered Engineers who become more prompt pushers than code crafters. I think we all need to strap in for a very bumpy road when it. Comes to software engineering.

What Delivery Managers can do with Ai

Working with other DMs there’s a common theme (sometimes compounded by the organisational culture they work in) in that I’ve found many people don’t say ‘I don’t know’ & recently I was trying to explain to someone the complexities of Monte Carlo simulation. Whilst not for this post, in a nutshell a Monte Carlo simulation is using a mathematical model (often done using Python but I do have a similar thing in Excel) to predict potential results of an unknown. The folk I was explaining this to immediately said, ‘can I have your spreadsheet Mario’ & I explained how they could learn quite quickly how to do this using ChatGPT.  At the time I said that I didn’t know if ChatGPT could do it, so I asked it….

Steps to Simulate Monte Carlo in Excel for Estimating π

1. Create Columns for Random X and Y Values: In columns A and B, you'll generate random numbers between 0 and 1 for each point's X and Y coordinates. Use the formula =RAND() in each cell for this. You might want to fill this down for a large number of rows; let's say 1,000 for starters.
2. Calculate the Distance: In column C, calculate the distance from the origin (0, 0) to the point (X, Y). The formula for this is =SQRT(A2^2 + B2^2). Again, fill down.
3. Check if the Point is Inside the Circle: In column D, write an IF formula to check if the point is inside the circle. This will be true if the distance is less than or equal to 1. Your formula could look like =IF(C2<=1, 1, 0). Fill down.
4. Count the Points Inside the Circle: At the bottom of column D, sum up all the 1s. This gives you the number of points that are inside the unit circle. Let's say this cell is D1001.
5. Calculate π: Finally, use the formula to estimate π. Create a cell for this, let's say E1, and put in the formula =(SUM(D2:D1001) / 1000) * 4.
6. Iterate: To make the simulation more robust, you can recalculate several times (Excel recalculates when you press F9) and average the estimates.
7. Increase the Number of Points: For a more accurate result, you could use more points, just bear in mind that Excel has row limits and performance constraints.

Office integration

Without doubt it’s the clearest most succinct instructional to build a basic template in Microsoft Excel – my flabber was appropriately ghasted. With each major app, (in particular within the Microsoft Ecosystem) there’s a pretty easy way to integrate ChatGPT. In Excel, you add an ‘add-in’ and authenticate via a secret key your Chat GPT account & then GPT can interact directly with your data!

Other Ai tools available

If you spend a bit of time to look, there’s a few clear integration possibilities for not only major apps on your computer, but also on your phone too. We’ll talk about security in another post, (maybe) but its staggering what you can do with GPT on-the-go.

If you’re looking for Ai tools to play with, here’s a list of nine I’ve played with which are awesome (I think);

1. ChatGPT – The big Momma of Ai
2. Kickresume – Improve your CV & built cover letters
3. Scribble Diffusion – Transform your sketches nto refined images
4. Forefront – Ai assistant to leverage chat & internet browsing
5. Descript – An audio & video editing tool with transcription capabilities
6. ChatPDF – An Ai tool to interrogate PDFs
7. Bard – Googles ChatGPT competitor that understands images
8. QR Code Art – This is awesome, create custom artwork using QR codes, great for branding
9. Decktopus – Create better slide decks & presentations

Thanks to @therundownai for the over-all list on Twitter

Examples of how using Ai can save a Delivery Manager time!

I’m doing a presentation in a couple of weeks about Lean, its one I’ve done before & it’s fairly well trodden. However, I could of used AI to complete the spine without me having to think. Using one of the tools listed above, Decktopus, I was able to give it a topic, a theme, a style, key tenets I wanted to cover & the time I had to deliver it & it created a presentation for me. Amazeballs

Estimation & story creation

Estimation & story writing. Its such a ball-ache, we have spreadsheets, online tools, we do Fibonacci, planning poker, don’t even get me started on story points & t-shirt sizes. It’s the bain of a Delivery Managers life. That & writing user stories. Keeping them succinct, relevant & meaningful is an artform in it’s own right. I fed GPT some requirements the other day & it came out with some beautiful stories & AC (Acceptance Criteria) I could barely write better myself

What makes it powerful is it’ll write totally feasible Epics & User Stories effortlessly & with context;

In summary

Look, I’m not here to pontificate on the pro’s, con’s & cultural impact on Ai, nor its compounded improvement in short time, nor the creation of some Skynet-esque singularity event that will ultimately lead to the human races demise, I’m just saying that as a Delivery Manager, there’s never been more reason to put AI into your tool chest. Listen to a bit more in-depth discussion on using Ai over on the Delivery Manager Daily podcast & follow me on X.