Steve Taplin “sells” engineers for a living.
But lately, he’s telling most of his clients the same thing…
“You don’t need more of them.”
Taplin runs a software engineering firm. His clients call him when something is broken — seven years into a roadmap, nothing shipping…
Much of the time, it’s a CEO who wants to know what’s wrong with the engineering team. In more and more cases, after spending a few days on site, he ends up giving them the same answer: engineering isn’t the problem.
That is an unusual thing to say when you sell engineering capacity for a living.
But that’s what makes Steve so trustworthy.
If engineering isn’t the problem — then what is?
Here’s how Steve puts it: Engineering is the only function that really adopted AI. The coding assistants work. Teams are faster. The gains are real.
But shipping software takes two things: deciding what to build, and building it. AI made the building fast. It did nothing for the deciding.
The bottleneck didn’t go away. It just moved upstream, to product.
I’ve seen this in my own work: a CTO runs a genuinely good AI program. Then their engineers run out of things to do. Velocity gains only make a difference when there’s a next thing in the queue.
It’s easy to clock a review queue back up. But no one’s measuring when a product team fails to make a decision.
Product may need to catch up. But there’s still plenty of opportunity for Engineering to push forward.
We discuss what those opportunities are and how to execute on them — including these three items that are worth getting started on this week:
Review your backlog. When Steve ran this exercise with a recent client, the measure that came back was 14 years. That number is not an engineering capacity problem…we discuss what it is and what comes next.
Cap how much of your code AI writes. With insider access across dozens of organizations, Steve’s found there is something of a “magic number.”
Flip how you train engineers on AI. For the vast majority of teams, AI writes code → Engineer tests it. Steve’s teams flip it. And get consistently better results.
Also in the full episode:
Why I don’t believe AI costs are ever coming down. We were sold this exact promise fifteen years ago and it wasn’t true then either.
The tech debt surprise: why some engineering leaders are suddenly less worried about their legacy code than they were a year ago.
And then there’s the central debate — and I want you guys to weigh in on it…
Steve says product is starved. It never got the tools engineering got. He wants the person closest to the customer writing the requirements.
I think the opposite. In a $100M company, most of product management is coordination. AI now writes the requirements and builds the prototypes. What’s left of the job is talking to customers and picking which problems to solve…let engineering absorb the rest.
Tell me what you guys think — settle it for us.
Steve, thanks for joining me.
This was a particularly fun conversation — and I suspect we’ll be revisiting some of these core discussion points as things unfold.
About Steve
Steve Taplin is founder and CEO of Sonatafy Technology, a software delivery firm with engineering teams across the US and Latin America, which means he spends most of his working life inside other companies’ delivery problems.
He is co-author of The Backlog Illusion, written with Sonatafy CTO Chris Horvat, and author of Fail Hard, Win Big.
Steve also hosts Software Leaders Uncensored — a podcast that features close to 200 conversations with CIOs, CTOs and engineering leaders.
Connect with Steve on LinkedIn: linkedin.com/in/stevetaplin
