By Michelle Gill, Sr Director of Engineering, Core DevOps, GitLab
Assembling an AI team should start with hiring people who demonstrate natural curiosity, grit, and technical versatility across AI, ML, and software engineering. When you onboard the right talent, your team will be equipped to navigate the bleeding edge of technology, maintain expertise, stay current on advancements, and separate signal from hype.
Surprisingly, after successfully hiring these experts and establishing a clear center of AI excellence, the real challenges begin.
The Skills Paradox
In my experience, the same traits that make talented AI engineers invaluable also make them nearly impossible to coach. Ten experts means ten brilliant approaches to every problem, leading to ten debates you will need to mediate before any solution can be shipped. The irony is that these are precisely the people you want on your team. They contribute a depth of experience that is difficult to find. However, their strong opinions often create disagreement loops and competing solutions where everyone presents a viable path, yet you still have to choose one direction to pursue.
All of this can impede development velocity. In today’s AI landscape, if a project takes longer than two months from initial concept to production, it’s already obsolete. Not all engineers can sustain this relentless pace, balance shipping code with staying up to date on research, and remain aligned with objectives when the direction keeps shifting. As a leader, you must keep the team moving quickly, make decisions that avoid endless feedback loops, and continuously assess whether you still have the right talent in place. In this environment, the leadership frameworks that work for traditional engineering teams ultimately break down.
Here’s what delivers results.
Four Frameworks for Consensus and Velocity
Start by addressing the basics. Flatten your organizational structures so that unnecessary layers don’t turn decisions into multi-week exercises. Shorten your timelines to match the pace of innovation, using the pressure of a looming deadline to identify when to fail fast, when to upskill existing talent, and when to provide an off-ramp. Set unapologetically high standards for adaptability and on-time delivery.
Once you’ve accomplished that, give your experts these four decision-making frameworks to move at the pace of AI.
- One DRI governs every decision. After soliciting input from the team, one person makes the call. Discussions should be timeboxed with defined success criteria. No parallel debates should take place in various channels.
- Differentiate ideas from execution. Once a decision is made, commit to that single direction for a fixed period. During that time, questioning the approach becomes temporarily suspended. Theoretical debate can continue indefinitely without boundaries. Therefore, set a course and gather real-world data before considering a change.
- Rely on evidence to pivot, not just a new idea. The bar for success does not need to be perfect; it can simply be “better than before.” If a new approach demonstrates improvement in your evaluation metrics, give it serious consideration. If it doesn’t, abandon it immediately.
- Engage your experts using their language. When you’re communicating with people who conceptualize in terms of model architectures, embedding dimensions, and evaluation frameworks, don’t force every conversation into business metrics and OKRs. Business impact is vital, but these are technical people solving technical problems. Speak technically when necessary, speak strategically when appropriate, and understand the difference.
These frameworks will help you deliver secure software faster and make better decisions. However, the reality is that even perfect implementation of these systems won’t change the environment. Competitors will announce breakthrough features every few weeks, and rival companies will continuously try to recruit your best engineers. The frameworks enable velocity. Retaining your talent safeguards your competitive position.
Maintaining Momentum
Building the team represents only one-third of the battle. Managing them constitutes the second third. Then, after brilliant hiring, established frameworks, and consistent shipping, you actually have to retain them.
Offer them problems that are worth solving. A lack of compelling vision, unappealing issues to solve, and nonstop debates are fatal to AI team engagement. Show your engineers how their work connects to a greater purpose, and make decisions that enable them to actually execute it.
Establish a path for career progression. AI roles have become complex faster than most companies’ career frameworks have evolved. Define what senior AI leadership looks like in your organization. Develop clear advancement opportunities with milestones that recognize both technical depth and strategic impact. High-performing talent will choose organizations where they can see themselves growing, not just working.
Emphasize continuous learning. The opportunity to work on the cutting edge, learn constantly, and stay ahead of the curve drew them to your team in the first place. Create space for it by allowing time for conference attendance, research, and experimentation. This isn’t a nice-to-have perk. This shows how your high-performing team remains effective in a field that is continually undergoing reinvention.
Technology will continue to advance, regardless of our readiness. The models will continue to improve, and your competitors will keep shipping. Yet human ingenuity is what will ensure your company excels. You win when you have engineers who move fast without sacrificing quality, leaders who align brilliant minds without crushing their creativity, and teams that ship consistently in an environment built for chaos. Support them, and you will remain on the bleeding edge of innovation.
##
ABOUT THE AUTHOR
Michelle Gill combines strategic vision with hands-on expertise to build and guide high-performing engineering teams within DevOps and the rapidly evolving fields of AI and data science. Currently at GitLab, she drives innovation in AI-powered workflows that enhance efficiency and reduce cycle times across their DevSecOps platform. With direct experience in software engineering, Michelle is passionate about helping teams develop skills they need in order to increase impact on their company’s success, and has employed numerous methods of trial and error, servant leadership, and old-fashioned hard work over the last 15 years to realize this progression at scale.
