AI Sped Up the Code; The Bottleneck Moved.
Earlier this month I called a friend who’s a CTO for some market research. He said something crazy to 2020 me: “I threw away the entire sprint process.”
What happened
His team started where the whole industry started, using AI to write code faster. Coding was the hardest part of development, so it made sense. It worked, and then it created a new problem. In his words: “We started coding crazy, but then we flooded the QA … and then product management and design just became the bottleneck immediately.”
The slow part moved. His takeaway: “It wasn’t about using AI to code faster. It was about using AI to develop software faster.” That meant looking at the whole chain instead of the one stage with the shiny tool.
How he mapped it
He listed each stage of how his team builds software, added a stage for security, and ranked every one on four levels:
- Manual: no AI at all
- AI-assisted: a person drives and AI helps
- AI-directed: AI drives and a person guides
- Autonomous: you give it a goal and walk away
He also ranks the space between levels, because it’s hard to call any stage fully done. “You want to be realistic with yourself.” The map showed him where his team was far behind and where he could stop spending energy.
Then he looked outside engineering.
Software doesn’t ship by itself. Sales enablement, customer success, and go-to-market all have to keep up, and he says they’re light years behind his team. Their answer is usually more meetings. His is “Meetings are the death knell of all of us.” He wants routine work automated, and people alerted only when something needs a human.
He spends about half his week on this: AI strategy, reworking plans, and explaining AI to other parts of the company.
Where agile fits
I’ve spent 13 years in agile and change work. Agile principles remain relevant, but I think how we do things is changing aggressively. A team that has to pull in its whole year twice is a team that needs to look at its cadence again.
Why this is a change management problem
Somewhere in that call I said, “everything is change management because everything’s changing,” and I’d say it again. The tool is the easy part. Getting engineering, product, sales, and support to work from the same picture of where work is stuck takes more. Part of what I do is help people see that they’re probably not talking about the same thing when they talk about value.
Try this with your team
- List the stages of how your team delivers work.
- Rank each one as manual, AI-assisted, AI-directed, or autonomous. (Be honest.)
- Look for where work is piling up now. That’s your new bottleneck.
- Show the map to the department next to yours and ask how they’d rank themselves.
Speeding up one stage only helps if the next one can take the load.If you want a read on where your whole organization stands, take the
EVOLVE Snapshot: https://evolve-snapshot.iconoclastltd.com/
If you’d rather talk it through, grab 30 minutes with me: https://tidycal.com/sarah-iconoclast/30-minute-meeting