AI's impact on engineering teams is real, but how is it changing right now?
We brought together a group of tech leaders to discuss what they’re actually seeing in their teams and had an insight into the situational reality of running an engineering team that’s going through this shift.
The consensus was not that we’re in a doomsday situation, and not a utopia either, we’re somewhere in the middle. And engineers are moving faster than most organisations are built to handle.
In this article, we delve into some of the key takeaways from the discussion including managing expectations, how roles are changing and where the real return on AI investment is.
AI is reshaping how engineering teams operate, but not always in the ways leaders expect. Drawing on insights from a panel of tech leaders, this article unpacks the real, situational shifts happening inside organisations adopting AI at scale.
You'll learn why senior stakeholders now expect 100x gains from AI (and why that expectation often outpaces reality), how the software development lifecycle is moving from rigid handoffs to whole-team collaboration, and why ‘context velocity’ is replacing story points as a measure of team performance. We also explore how new roles like the Forward Deployed Engineer are emerging, why curiosity may matter more than seniority in determining who thrives, and what Atlassian's ‘fragmentation tax’ reveals about the gap between AI investment and actual results.
Whether you're a CTO, engineering manager, or team lead trying to navigate this shift, this article offers practical, experience-based takeaways on managing expectations, restructuring teams, and proving AI's genuine business value.
A year or two ago, we expected to see 10x returns from AI. Now, senior stakeholders are expecting much more and that expectation gap is creating real friction inside organisations.
Even measuring AI's impact is proving difficult. A study by METR from 2025 found that AI tools actually slowed experienced developers down. They have since had to redesign their experiment, noting that developers are likely faster now but that reliable evidence of how much remains elusive.
Another challenge arises when companies try to bolt AI engineering to processes that are broken or outdated platforms and systems that are in need of modernising. This can then amplify existing problems and create more work for teams. Our guide on AI-First Software Modernisation explains how organisations can prepare their software, engineering practices and operating models for effective AI adoption. It introduces Codurance’s approach to building AI-ready systems and AI-accelerated engineering together, so AI can be adopted safely and connected to measurable business value. Read the AI-First Software Modernisation guide here.
The core argument here isn't really about AI at all. It's about good engineering discipline: domain-driven design, test-driven development, and designing for flow. Organisations that have invested in solid engineering practices and repeatable, maintainable ways of working are the ones seeing genuine throughput gains from AI. Speeding up a broken system doesn't make it efficient, it just breaks things faster.
Moving beyond individual productivity, there is a more structural shift happening in the SDLC.
The death of handoffs
The traditional model where the product team writes the spec, UX create the wireframes, and the engineers build it. This model is orderly, but now looks slow in comparison to what can now be achieved with AI. The new team flow that is emerging where teams collaborate on the ‘why’ as much as the ‘what’. They ask ‘why are we building this, what are the trade-offs, and what happens if we don't build it at all’.
Moving from story points to context velocity
Teams have long been measured on story point velocity (how many tickets have moved across the board). But in this new era of AI, context velocity has become more important - how quickly a team can genuinely understand what the organisation is trying to achieve and where they can add value. This shift is a move from output-tracking to results-focused.
The rise of the system orchestrator
With more people being able to do tasks that fall outside of their remit thanks to AI tools, a new risk has emerged that everyone assumes they can cover every role.
AI has already created a new wave of job titles: prompt engineer, GTM AI Engineer and classic roles with ‘AI’ simply bolted on. Some jobs have been rewritten entirely and some roles are a new invention, however in some cases an existing job description has merely been updated to include the word ‘AI’.
What’s more interesting is what’s emerging at the senior end - the emergence of the Forward Deployed Engineer - with roles being advertised at both OpenAI and Google. This is essentially a technical consultant role where the engineer is embedded within a clients’ organisation to build and deploy products for real-world production. The skills required for a good Forward Deployed Engineer are the same skills that have always made a good engineer - strong communication, sharp analysis and a genuine problem solving ability.
AI has undoubtedly had an impact on the entry-level job market too. With many repetitive tasks now automated and the rising expectations that everyone should be AI fluent, this is putting a squeeze on those seeking early-career engineering roles.
Despite not having the years of experience under their belts, junior talent has an unexpected advantage through not carrying the baggage of old ways of working. A suitable approach that benefits both parties is to pair a junior with a senior, and the junior often asks “why are we doing things this way?”, which is a genuinely useful challenge to what can often be legacy approaches and ways of thinking.
The concern arises more when we look at the mid-level engineers. These people have built their careers around specific tools and processes and can be facing the steepest adjustment curve as they don’t have the fresh eyes of juniors or the strategic vantage point of seniors.
These themes extend into hiring too. We’re seeing a return to human-in-the-loop recruitment where behavioural-based interviewing, observing how candidates work through problems with AI and in-person interviews are introduced.
Atlassian recently conducted research into the state of teams and from this study, coined the term ‘fragmentation tax’. This describes a significant and often unacknowledged gap between how much organisations are investing into AI and the results that they're seeing.
Ai is moving fast, but not always in the right direction, leading to duplicated effort and rework. However, some organisations are seeing genuine value, but it often comes down to whether teams are questioning what they’re building and why before diving in and creating new products and features.
There’s also a maturity curve in how leadership treats AI investment. Boards are expecting leaders to track ROI and tie AI initiatives directly to core business metrics, meaning that tech leaders increasingly need to speak the language of the board by providing concise value statements backed by data. You can read more about how to speak the language of the board in our article The Value Multiplier for CTOs: Navigating the Shift from Technical Delivery to Commercial Impact.
At Codurance, we help organisations adopt AI in a way that delivers measurable business value. Our approach brings together AI-ready systems and AI-accelerated engineering, so teams can move faster without amplifying the problems already holding them back. Whether you're under pressure from the board to show returns or trying to make sense of how roles and ways of working are changing, we work alongside your teams to build the foundations, practices and confidence needed to make AI work in practice.
We help organisations:
If you'd like to find out more, get in touch with us today.
What is the ‘fragmentation tax’ in AI adoption?
The fragmentation tax is a term coined from Atlassian's research into the state of teams. It describes the significant, often unacknowledged gap between how much organisations invest in AI and the actual results they see, typically caused by duplicated effort and rework when AI moves fast but not in the right direction.
Why isn't AI alone fixing broken engineering processes?
AI amplifies whatever process it's applied to, good or bad. Bolting AI engineering onto outdated systems or broken processes tends to create more work, not less. Genuine gains come from organisations that have already invested in solid platform engineering and disciplined practices like domain-driven design (DDD) and test-driven development (TDD).
What is a Forward Deployed Engineer?
A Forward Deployed Engineer is essentially a technical consultant role where an engineer is embedded within a client's organisation to build and deploy products directly for real-world production. The core skills required (strong communication, sharp analysis, and problem-solving) are the same skills that have always defined good engineers.
What is ‘context velocity’ and how is it different from story points?
Context velocity measures how quickly a team genuinely understands what the organisation is trying to achieve and where they can add value, rather than tracking how many tickets move across a board. It represents a shift from output-tracking to results-focused team performance.
Who benefits most from the AI shift? Juniors, mid-level, or senior engineers?
According to industry leaders, junior engineers often benefit from not carrying the ‘baggage’ of legacy working methods, while senior engineers bring strategic perspective. Mid-level engineers who've built careers around specific tools and processes, may face the steepest adjustment curve.
How is hiring changing because of AI?
Organisations are returning to human-in-the-loop recruitment, including behavioural-based interviewing, observing how candidates work through problems using AI, and reintroducing in-person interviews. This is aimed at assessing judgement and adaptability, not just technical output.
How should tech leaders measure the ROI of AI investment?
Leaders are increasingly expected to tie AI initiatives directly to core business metrics rather than relying on enthusiasm about the tools themselves. Boards now expect concise, data-backed value statements that demonstrate a clear return rather than assumed productivity gains.
Natalie Gray is Director of Marketing & Growth at Codurance, a global AI-first software engineering consultancy that supports businesses and their investors to drive value through building and modernising sustainable software and platforms at all stages of the investment lifecycle. With more than 20 years in the tech industry, Natalie believes the power to innovate is accelerated when people are able to try new ideas, fail fast and collaborate. This is why, to her, community is at the heart of her approach to aligning business and technology outcomes. Outside of her commercial role, Natalie runs a number of tech meetups, is an advocate for women in tech and enjoys spending time with her family or protecting her 800 day streak on Duolingo.