You have 1 article left to read this month before you need to register a free LeadDev.com account.
Estimated reading time: 7 minutes
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
After a year of DirectorPlus, one theme keeps surfacing: AI strategy. In this special edition, engineering leaders reflect on what it really takes to make AI tooling deliver.
We’ve published DirectorPlus for a full year, interviewing leading directors and technical execs from renowned software-heavy companies, including GitHub, Spotify, Yelp, Intuit Mailchimp, Netlify, and others.
And no matter the topic, AI is the elephant in the room. In 2025, AI-assisted development has permeated new tooling, feature updates, reports, and overarching company strategies.
What have we learned? AI is a game-changer, but it’s not magic. It takes intentional strategy and technical discipline to get the most out of working with AI agents in a software setting.
So, before you head into summer, here’s a check-in with past DirectorPlus guests on how AI thinking is evolving.
Outcomes with AI are a mixed bag
While engineering leaders are actively pursuing many new investments, from platform engineering to developer metrics, legacy migrations, and others, AI continues to dominate the conversation.
“Folks on the DX engineering team are daily AI users, and we started experimenting really early,” says Laura Tacho, CTO at developer productivity company DX. At large, she acknowledges that the AI hype has been near deafening:
“It’s not hard to go on LinkedIn and see a headline about 30% of code being written by AI, or that it’s going to replace developers outright,” Tacho says. But, actual success with AI in engineering is a different story. “Within real organizations, the results are really mixed.”
The industry is discovering there is an art to writing effective prompts and avoiding technical debt and security bugs when using AI coding assistants, let alone knowing how and when to deploy them.
Directors turn AI mandates into reality
This hasn’t dissuaded executives from chanting the AI cheer at full volume. In the last year, we’ve witnessed top-down internal AI mandates from Shopify and Amazon and big AI usage statements, like Microsoft’s claims that up to 30% of its code is AI-generated.
For Tacho, executive AI orders may feel controversial now, but will eventually feel like a no-brainer. “I’m not put off by these mandates at all. I think in 20 years, we’ll look back at these mandates like some of the executive messages that came out in the 1990s, saying that the business needed to embrace the internet,” she says.
But not everyone agrees mandates are the right move – and flashy statements alone won’t drive results.
“This sounds like executive theatre,” says Kent Wills, senior director of engineering at Yelp. “Communicate outcomes, don’t dictate tool usage. If AI helps, great. If not, people are still on the hook for real results.”
In other words, leaders should stress the end objectives the organization is trying to achieve, rather than stringently force certain tools. And at the end of the day, AI doesn’t replace the responsibility of humans to deliver, either.
To get those results, directors must translate vision into reality. “This is where directors come in,” Tacho adds. “It’s your job to track adoption, impact, look for populations that need more support, and figure out what you can learn from folks who are heavy users.”
One way to do this is by grounding the AI vision in practical, team-level enablement. As Sabrina Farmer, CTO at GitLab, puts it: “I’m committed to ensuring that our engineering and development teams can increase their time spent on strategic work by integrating AI into their workflows.”
“We take a deliberate approach to AI adoption, providing our teams with the necessary resources and guidance to support their use cases,” she adds, no matter if they’re using AI-enhanced code analysis solutions, testing frameworks, or project management systems.
GitLab’s approach, rooted in collaboration and skills-sharing, helps upskill the entire engineering organization – from developers to quality engineers and infrastructure teams – bringing that high-level AI vision to life.
More like this
It takes teamwork to deliver on AI
For Gus Fune, CTO at BÆRSkin Tactical Supply Co., the biggest wins lately have helped boost team productivity with AI.
The first way they’ve achieved this is by consolidating developer workflows. Fune calls it “adopting a one-fleet paradigm,” in which the entire development team works only with one set of tools, engineering conventions, and frameworks. This has reduced vendor costs (since the team has been able to consolidate vendors, renegotiate contracts, and leverage discounts) and has the side effect of consolidating documentation and processes, positioning them nicely for AI tools.
“Whenever we hook in an AI tool or agent, the entire context is full of examples and guidelines on how to follow,” Fune says. It’s still a work in progress, with some legacy areas left to consolidate.
The second way BÆRSkin Tactical Supply Co. has reaped rewards is through investing in people skills. This has involved holding workshops and activities to improve collaboration and communication. Encouraging people skills not only benefits how the team collaborates, but it also boosts how teammates communicate with AI.
“This is leading to a better engineering organization, but also it’s helping developers who are not yet productive with AI tools,” says Fune. For some, AI still feels too alien, or their communication style doesn’t mesh with it. Encouraging knowledge sharing has helped Fune’s teammates develop better working habits and deliver better prompts. “After a few months, I’m seeing considerable results,” he says.
Experiment collaboratively and re-evaluate
Savvy leaders aren’t handing out AI tools without guardrails. They’re encouraging experiments, collecting feedback, and tracking impact.
For instance, Fune takes a deliberately loose approach to AI tooling budgets. His team tests many tools, from free trials to paid plans, to see what actually delivers. This is helping them stay current with the trends while remaining mindful of what gets added to the tech stack. “We tested everything out there in small and bigger plans to test and validate things. The more we test, the more we validate our current approach.”
Wills at Yelp also emphasizes the importance of tooling guidance, recommending the following: “To help teams explore using AI, share what tools fall within company policy, make it easy for people to use it in their day-to-day, and celebrate what’s working.”
Even with decisive executive-level messaging, it’s up to internal engineering leaders to enact AI initiatives. Tacho encourages training and experimentation. It also requires constant monitoring: knowing the team’s baseline performance and how AI is affecting end business outcomes.
She also advocates for structured rollouts and identifying and doubling down on methods that have proven to move the needle. This will entail looking for signs that AI tools are improving outcomes. As Tacho expands:
“Are your teams shipping more? Is quality the same? Are you spending less time on maintenance? Is the code becoming harder to read?”
Innovating niche tools with AI
Engineering leaders must walk the talk, too. Fune encourages directors to lead by example. “I am the first one using a lot of AI tools, recording my process, writing notes, and sharing with the team,” he says.
AI can also be a boon for organizations with the grit to develop and maintain their own internal applications, possibly removing the need for (some) pricey third-party software-as-a-service solutions. For instance, at BÆRSkin, engineers are utilizing AI to build out minor internal tools – either tailored to specific company workflows or replicating common tools that are easily available via third-party services, he says.
As an example, Fune cites replacing DORA/SPACE metrics tools with AI-generated alternatives. They’re using AI vibe coding tools like Replit and Bolt to quickly construct these sorts of internal applications. “We are always a ‘build before buy’ organization, so this helped us a lot.”
Spotify, too, is seeing major gains from AI-powered internal tooling. “At Spotify, finding information has consistently been among the top three concerns in our engineering satisfaction survey,” says Pia Nilsson, senior director of engineering and head of platform developer experience at Spotify.
“One of the ways we’ve addressed that is with AiKA, our ‘AI knowledge assistant’,” she adds. AiKA pulls information from technical documentation, Slack channels, organization charts, and other internal knowledge bases.
It complements their existing systems and has encouraged a culture of up-to-date, accurate documentation. “It’s turned into a powerful feedback loop for knowledge sharing,” adds Nilsson.
Spotify has since made AiKA available to enterprise customers via Spotify Portal for Backstage – with support for OpenAI, Gemini, Bedrock, or any LLM of their choosing. Over 1,000 employees use AiKA every day, Nilsson says. “All Spotify employees, not just engineers, have access to it. Instead of searching everywhere, they just ask AiKA.”
New York • Oct 15 & 16, 2025
Full agendas for LeadDev’s New York events are up! 🎉
The outcomes: more with less
Across the industry, many are working with tighter IT budgets. So, how should engineering leaders protect or prioritize high-impact work when resources are limited?
Tacho views data-driven approaches and continuous improvement mindsets as critical to maintaining an edge. But it’s not just about the numbers – it’s about upkeeping morale amid tense times. “Leaders who excel during these times don’t treat uncertainty and pressure as morale-killers,” says Tacho. “Instead, they’re the raw materials for building something better.”
To keep this morale, Fune recommends being upfront about industry changes in the macroeconomic sense and retaining a collaborative spirit.
“Keeping the team updated, but involved in making decisions themselves on what they want to test while keeping their expectations open, really helps out,” says Fune.
