Stop Burning Cash, Start Delivering
Every engineering leader I talk to has the same question about AI: "Is it actually making an impact?" Not "could it?" or "will it?" - but is it delivering value right now? Here's what I'm seeing in the trenches.
Text Generation Actually Works
PR descriptions, documentation, code search, predictive typing - AI is crushing it here. At LinearB, we also use AI to improve our workflows, such as automatically generating iteration retro summaries; this delivers real value. No magic, just concrete improvements to standardization and workflows.
Code Generation? Hold Your Horses
Test creation and code generation are still experimental. Your mileage will vary dramatically based on your architecture and tech stack. We recently sat down with gen AI practitioner Birgitta Boeckeler, who's been testing these tools in legacy systems, and her findings confirm what I'm seeing across our community.
My takeaway?
Measure everything. Your business doesn't care about AI buzzwords — they want ROI.
Here's Your Playbook
For mid-size and enterprise teams:
- Give your team space to explore. The proven use cases are there: standardized PRs, documentation, code search. Let your engineers experiment with these tools today. Not tomorrow, but today. That’s how you get immediate productivity gains while experimenting to find your best use cases.
- Pick concrete wins you can deliver today. Your execs see the AI line items piling up on the budget. They're not asking for moonshots - they're asking why they're bleeding cash on AI tools. Want to change that conversation? Start rolling out gen AI for the proven use cases and tie adoption to efficiency metrics like cycle time and planning accuracy. That's immediate ROI they can see.
- Monitor the real impact, not just the hype. Your entire delivery pipeline needs to adapt to AI-generated code. Track metrics like code review bottlenecks, change failure rate, and rework rates. Early data shows gen AI can actually hurt these metrics if you're not careful. Want transformative AI? Look beyond code generation to your entire delivery pipeline.
Here's the truth: Your business is watching every AI dollar that walks out the door.
They don't really care about your cutting-edge tools or agent-based whatever. They care about next quarter's numbers. Start with standardization wins you can measure in dollars and cents.
Once you've proven AI can pay for itself, then you can talk about the bleeding edge stuff. Look, AI isn't going anywhere. But if you want to keep experimenting with the cool new toys, you better show concrete value with the basics first. Start with what works, measure religiously, and use those wins to fund your next moves.
Everyone's chasing AI that can write better code. I'm betting on AI that understands your entire business - code, JIRA tickets, deployment data, support queue, all of it.
That's not just cool tech, that's transformative value.