Jake Hookom, EVP of product at DevOps and version control company Perforce says he has been studying the rise of AI and copilot assistant tools among the software application development community… and some key trends are emerging. His team is hearing from users that copilot-type tools deliver developer productivity gains of between 50-70%. But what deeper signs and trends do we need to be aware of here?
It sounds like good news. Hookom suggests that while this demonstrates the potential of AI to transform workflows on an individual level, the Perforce team (who work across a diverse range of developer teams, but are known for penetration within the games development industry and others in particular) say that they are also hearing that this does not necessarily translate across the enterprise.
Developers might be writing code quickly, but the quality and visibility gaps widen across teams and siloes grow. Trust, transparency and integration lag behind. Add AI into the mix… and an already broken process gets bigger, or, at the very least, fails to create the hoped-for value and alignment with business goals.
Implementers become orchestrators
“That’s the risk with broken processes, but there is also opportunity, thanks to the way developers are shifting from being implementers into orchestrators who review and manage output; team leaders who were managing teams of five humans now supervise the equivalent throughput of 50,” said an upbeat Hookom, in a briefing. “So, given the headspace and time that this liberates, what if, in addition to individual impact, development teams also switched their focus more to how AI, particularly to enable DevOps or DevOps workflows, could positively impact the rest of the organisation and drive that elusive value? After all, the point of DevOps itself was always to eliminate siloes and remove barriers. This is why our recommendation is to start building a shift-up mentality.”
We might define shift-up here by saying that it is what AI should be doing anyway i.e. elevating the thinking and conversation above implementation to the wider impact. So, not just writing code faster, but also focusing on the quality, security, compliance and the implications for other parts of the business chain. If there were ever a time to start thinking about the connective tissue between different functions, the Perforce man says it is now.
With AI, the suggestion here is that collaboration across the delivery chain evolves from handoffs to partnerships. So how does it work?
“First, the developers overseeing the work of both AI and fellow humans need to apply solid, modern architectural practices to guide, review, and govern their work, playing a critical role in maintaining the quality of change… while ensuring that the trust and transparency of AI are present. I also believe this is why the tools and solutions that we (tech companies) create need to be designed for “human-in-the-loop” scenarios. However, this is just the start: the deeper, more impactful efficiencies of AI workflows reside in the relationship between business analysts, testers, infrastructure and data management functions,” said Hookom.
Testing vs. development
In practice, this might mean looking at the relative state of testing vs. development… and indeed the release process… and vice versa: “If I deliver this code and it goes over to the QA team, what happens?” or, “Once that software has been deployed, what’s the feedback loop to catch issues early?”
Hookom explains that developers cannot carry this responsibility alone: the supporting tools at DevOps teams’ disposal need to shift-up too. For example, he says that if AI generates 3,000 lines of code, a companion tool could summarise and test that output to help humans better interpret, understand and prioritise next steps. Tools will be able to not only find where the problem is, but also fix it and ensure it never happens again, or remove unnecessary processes altogether.
“Get the shift-up mentality and supporting cast of tools in place and the AI workflows that developers use help them provide demonstrable value to the broader organisation: this is where the true multipliers lie, weaving AI into the connective tissue of the entire DevOps delivery pipeline. Developers also get some superpowers and arguably a kind of promotion. Executed well, rather than being a shaky step into the unknown, shift up could be a decisive step towards the future,” stated Hookom.
A manifesto for shift-up
Perforce provides a “manifesto” of sorts for the shift-up that can now happen inside software application development teams and departments as follows:
- Velocity must come with trust – We need to help developers to move fast and stay secure and compliant, but without becoming security experts. Unlike shift left, shift-up doesn’t have to mean risking creating visibility and quality gaps.
- The drive to simplified workflows – Let AI handle the grunt work and remove repetitive tasks because that’s what it’s good at i.e. people are the arbitrators of what is important.
- Champion natural interfaces – Engineers describe the outcome they want and let the system figure it out (side note: developers have to move from prompt engineering to context engineering).
- Fewer context switches – Consolidate fragmented tooling so work happens in one place, not across ten, reducing friction for users constantly having to switch screen views.
- Augment, not overload – Give developers and everyone working a way to bring applications to market superpowers without giving them more to manage. We all manage too much as it is: it’s time to make AI work for humans.
Lead image: Wikimedia Commons, bottom image ChatGPT.