Can We Use AI To LEAD Software Development Teams?

We often speak about how AI can help, or hurt, software developers... but how can it influence the leadership of software development teams? In this clip from Dave Farley's Engineering Room podcast, Gary Gruver discusses the possibilities and what the future might hold for technical leaders.
Listen to the full episode HERE ➡️ open.spotify.c...
-
🗣️ THE ENGINEERING ROOM PODCAST:
Apple - apple.co/43s2e0h
Spotify - spoti.fi/3VqZVIV
Amazon - amzn.to/43nkkRl
Audible - bit.ly/TERaudible
-
🙏The Engineering Room series is SPONSORED BY EQUAL EXPERTS
Equal Experts is a product software development consultancy with a network of over 1,000 experienced technology consultants globally. They increase the pace of innovation by using modern software engineering practices that embrace Continuous Delivery, Security, and Operability from the outset ➡️ bit.ly/3ASy8n0
-
#softwareengineer #developer

[@THE_NE0]

A lot of talking but no answer to the question ?

[@madlopt]

@@THE_NE0 that's the usual thing on KZbin, they making videos to answer on some important question for you but they don't know the answer. And anyway you will watch all of them, that's why they making it )))

[@AUNEDJ]

What were you expecting from AI related content? 😂

[@ulrichborchers5632]

Dave, I admire you for being able to stay calm when being confronted with this madness, and then putting things into perspective in such a polite way.

[@matthewtrow5698]

Surely if AI can lead software development teams then at that point AGI must have been reached and human software developers will be redundant. A Lead developer or a project lead with technical knowledge is light years ahead of what AI can currently do for software engineers. At this level, it's not so much about the code, it's about ... so many things. The architecture, the business logic, the lack of business logic, being a conduit between the engineers and a project owner - the list goes on. Until AI is capable of communicating at that level, the question posed in this video is science fiction.

[@br3nto]

11:42 AI tools aren’t in the right place to achieve the goals that both of you are looking for. AI tools in the IDE is really the wrong place for them. Instead, they should be in git repositories and issue tracking systems. Then it could be the goal to define requirements, and have AI tools to help incrementally change the system to meet the requirements. People in the loop can validate or direct the changes. Requirements can incrementally change, and the AI tools identify and close the requirements gaps. Maybe even suggest 10 different alternatives for consideration. Devs can provide feedback on the alternatives, and combine the alternatives. Etc etc. there is still this fundamental problem of making sure that everyone is on the same page and working in the same direction and making the right changes though. The other gap is that we don’t have a cross-boundary multi-language “compiler” that can detect error states in distributed systems yet. That is , a compiler that can create an AST for any distributed or monolithic system.

[@Kitsune_Dev]

Can you talk about Technical debt and how to reduce it? dealing with this right now at work 😢

[@ContinuousDelivery]

I did a video a while back on Technical Debt - kzbin.info/www/bejne/Z36loX6Vhd14pLc There are also a couple of clips from Adam Tornhill's Engineering Room episode which might be useful.

[@br3nto]

5:35 I like that point Dave. What we actually want to know is: what are the gaps between what the system does and what we want the system to do, and what effort and cost is required to close those gaps. Leaders of organisations want to be able to steer the ship, and to do that they need to know how to correct course and what efforts and resources will get them to the new place quicker. Actually, it’s probably not just leaders that want that, rather everyone wants that.

[@br3nto]

The other point is that leadership teams often don’t really care about the small details of how. Their focus is more on things like: how can we increase our market share, how can we increase profit, how can we attract bigger customers with larger growth potentials, what market segment niches should we be expanding into or getting out of?

[@AlwaysHCYT2]

when there is too much boiler plate code to write, the answer is not: you can use AI to increase your productivity, the right answer is: it is time to think about a new library, framework or tool!

[@AlwaysHCYT2]

I don't remember CEOs and other managers yelling: "this is the future" when IDEs were introduced. and the step further with productivity was probably bigger than the AI that help you to write unit tests! (the introduction of TDD was something big... but it didn't need billions of investments in hardware, so, it was not interesting for CEOs and managers!)

[@stephendgreen1502]

Fortunately for a seasoned dev or experienced lead dev, AI does not replace experience, your most valued asset. It tends to not offer solutions to a broad understanding of the codebase and tech stack or platform. It offers a solution to a prompt or a narrowly scoped immediate problem.

[@tsomerville1970]

Aider-chat helps incremental. And BDD / Gherkin working super well for non-visual. Just pay attention to the scenarios And yes, a good convo to explore super helpful before hitting go Yes to not replacing programmers, and making is more helpful

[@nexovec]

The person that made this up should get a patent on no managerial vision.

[@JohnCross-e3w]

There's a problem with the idea that "AI wont replace software engineers, it will just make them more productive".... If software engineers are suddenly more productive, companies wont need as many software engineers. I'd love that to mean that we all start work 3 days and getting paid for 5 days... but I suspect that isnt what will happen.

[@sergeybush4253]

incrementalism - perfect point!

[@JavierBonnemaison]

Betteridge's Law of headlines stands strong.

[@llothar68]

Project managers are for sure easier to replace by AI then programmers.

[@agileduck5773]

I'm sad to say that an AI has a chance of being more useful than most of the SCRUM matsers I've worked with

[@Algorerhythm]

We may be reading much more code than we write.

[@phatster88]

From Low Code to No Code to Fake Code

[@mooooooooooomooooooooooomo6829]

answer is no?

[@denislitvin2535]

@@mooooooooooomooooooooooomo6829 😂😂😂

[@denislitvin2535]

I have same question!

[@JunYamog]

I am surprised that neither SEEM to have used aider or even just Claude’s project feature. What they are imagining to help developers is already being developed. It seems they have only used the chat interface on LLMs. Although not perfect there are now specific tools being developed for iterating through an artifact rather that generating in 1 big bang.

[@mktatyt]

9:23 Isn't that what's called temperature in the OpenAI API? If you set that to 0 you get reproducible results.

[@brownhorsesoftware3605]

Please stop leaving out the COST of AI while you are trying so hard to make it useful for something, anything. The inputs to AI are not just prompts - the energy it uses is massive not to mention the human toll in job loss.

[@SimGunther]

Why would the leaders lead themselves out of a job? TLDW Ay Eye would make developers more productive at making worse software over time

[@jvf890]

No