For our third CTO Breakfast Roundtable, we brought together CTOs, Heads of Engineering, Product leaders, and investment partners for a direct, practical discussion: What will our engineering teams actually look like as AI becomes part of everyday work?

Setting the Scene

Participants shared why this topic matters so much now.

AI coding tools like GitHub Copilot, Cursor and custom LLMs have moved quickly from interesting prototypes to being used by engineers every day. But the conversation is moving beyond “how do we use this tool” to bigger questions:

• How does this change who we hire and train?
• Do we need fewer engineers? Or just different skills?
• How do we keep security, quality, and compliance in check when code is being generated so fast?
• How do we manage the inevitable hype internally?

As one person put it: “Software engineering has always been a story of more abstraction—from assembler, to higher-level languages, to cloud, and now to prompting and orchestration. This is just the next layer.”

But that doesn’t mean it’s simple.

A Concrete Example from Damilah

By way of an example Aleksandar Karavasilev, CTO at Damilah explained how they deliberately picked a small but real innovation project to see what using AI end-to-end could achieve.

The project? Helping the University of Economics in Skopje improve exam scheduling—a task that normally would take a full team 2–3 months.

Damilah’s approach was to use AI in every step:

• Scope: The team interviewed university staff, recorded the sessions, transcribed it, and immediately ran them through ChatGPT and Perplexity to summarise and define scope in one go.
• Architecture: Using Napkin AI, they generated a solution architecture in about 30 minutes.
• Code: A single developer and an architect used prompting tools to generate the working code.
• Testing: About an hour to validate and check requirements.

Total time? Around six hours from interview to functional prototype.

Aleksandar pointed out it wasn’t a throwaway demo. The university is adopting it for real use, with just a bit more refinement.

His takeaway:

“Instead of 2–3 months with a full team, we did 90% of it in six hours with a tiny team. You can’t do that everywhere, but in the right context it’s game-changing.”

Where AI Delivers Value

This example opened up a lively discussion about where AI tools are already proving valuable.

Several attendees talked about how AI tools are cutting down the time for:

• Discovery and inception: moving from a full sprint to a few hours.
• Prototyping: empowering product managers and designers to spin up clickable prototypes in hours instead of weeks.
• Automating boring tasks: writing repetitive tests or scaffolding code.

The phrase “fail fast and see what flies” came up more than once.

But there was strong agreement the gains aren’t universal. Greenfield projects and quick experiments are ideal for these tools. Brownfield, compliance-heavy, or legacy work? Much harder.

The Buck Still Stops with Humans

Nobody in the room pretended that AI was a silver bullet.

Especially for financial services and banking teams, AI-generated code can’t just be merged in. They described strict security and compliance requirements that demand human review in order to consider:

• Performance problems or concurrency bugs.
• Security vulnerabilities that AI won’t catch.
• SOC 2, auditability, and client expectations.
• IP ownership issues when vendors claim rights to generated code.
• The opaque, “black box” nature of LLMs.

The message was clear: “The buck stops with the human.”

Voices around the table agreed that guardrails and discipline were essential. They discussed needing:

• Strong CI/CD pipelines with security checks.
• Policies defining acceptable AI usage.
• Clear processes for reviewing and approving code.

Shifts in Roles and Skills

One big theme was how AI changes what engineers actually do.

Several participants described the rise of what they called the Product Engineer:

“People who solve business problems end-to-end. Tech is just the means to an end.”

Others pointed to the emergence of super-productive seniors:

“Seniors can use AI to do 5x the work they used to. But that creates a problem—how do we train the next generation?”

The worry is that AI takes away many of the simple, repetitive tasks juniors used to learn on. Without those, how do they build the foundational knowledge needed to review, design, and maintain complex systems?

Voices around the table debated how to handle this:

• Some suggested not letting juniors use AI too early, to force them to learn fundamentals.
• Others felt curiosity should be encouraged but structured, with real mentorship and deliberate training plans.

New Roles and Skills Emerging

There was broad agreement that AI is creating demand for new kinds of roles:

• Prompt Engineers: People who know how to craft effective, context-aware instructions for AI tools.
• Model Integrators: People who understand different models and can chain them together to solve real business problems.
• AI Governance Leads: Defining what’s allowed, securing approvals, and managing risk.

One of the attendees warned:

“If you don’t have guardrails, you’re just setting yourself up for chaos.”

Keeping Product, Design, and Engineering Aligned

A big part of the conversation turned to how AI is changing relationships between teams.

Product managers and designers now have access to tools that let them build rich prototypes in hours.

That’s great for speeding up feedback loops, but it also risks creating silos:

“We spent years getting product, design, and engineering to work together in discovery. Now they’re off doing their own thing again.”

Participants talked about the need for:

• Clear, shared OKRs to keep teams aligned.
• Involving engineering early in discovery, even if AI is making everything faster.
• Creating review processes that make sure those prototypes, if they’re not going to be thrown away, are secure, performant, and maintainable.

As someone summed it up:

“AI won’t fix your process. If you have a crap process, it’ll just make the chaos go faster.”

Managing the Business Hype

Several participants brought up the challenge of managing internal expectations.

“Every CEO is watching videos on LinkedIn about AI and then turning to the CTO saying ‘make this happen.’”

People shared practical advice:

• Channel enthusiasm into small, targeted pilots that deliver real wins.
• Educate business stakeholders on what AI can and can’t do.
• Avoid “solutions looking for a problem.”

One private equity partner noted:

“We see a growing gap between AI-native startups that are lean and fast, and legacy companies buried in compliance and old systems. This is just cloud and Agile all over again, but on steroids.”

What Will Teams Look Like?

Toward the end, the group talked about what their teams might look like in a couple of years.

Ideas included:

• Smaller, more focused teams with tighter domains.
• Closer ratios of product managers to engineers.
• New roles such as Product engineer, Prompt Engineer, Model Integrators, etc
• More diverse skills on teams, balancing deep tech expertise with communication, design, security, and understanding customers.

The collective takeaway was that the industry is currently in transition as AI technologies are emerging and evolving and is yet to reach a steady state.

AI presents a generational opportunity to redefine software engineering productivity, team dynamics, and product development. However, its successful adoption depends on careful governance, cultural alignment, and sustained investment in human expertise.

Engineering teams will not disappear but will evolve—becoming more impactful, cross-functional, and focused on high-value problem-solving. The organisations that succeed will be those that balance technological acceleration with human-centred development, ethical oversight, and continuous learning.

The roundtable underscored both the optimism and caution required to navigate this AI-driven transformation.

A Straight-Talking Conversation

As with every CTO Breakfast, this session was defined by openness and honesty.

Voices around the table challenged assumptions, shared real-world experiments, and talked about the cultural changes needed to make AI work in practice.

The clear takeaway?

AI tools will make some things easier. But they also make the hard stuff—like aligning teams, maintaining quality, building skills, and staying focused on the customer—even more important.

As one participant put it:

“Technology is the easy bit. It’s people and process that make or break you.”

Thanks to everyone who joined us.
We’ll keep these discussions going, helping engineering leaders navigate the real changes AI is bringing to our teams and organisations.

As AI continues to transform software development, engineering leaders are rethinking how teams work, how tools are used, and how success is measured. At our latest CTO Breakfast Briefing, technology executives gathered to explore these shifts, share emerging practices, and examine the growing influence of agentic systems on the software development lifecycle.

One of the highlights of the session came when Aleksandar Karavasilev, CTO at Damilah, shared the results of internal AI experiments. These revealed practical gains and sparked further discussion around how AI is reshaping team workflows, engineering oversight, and productivity.

AI tools accelerate development, but measuring impact remains complex

Aleksandar opened the session by presenting findings from structured experiments using tools like GitHub Copilot and Cursor. One example showed how engineers used Cursor to analyse over 900 SQL procedures, reducing what would typically take weeks into a few days. Across 80 engineers, the company recorded over 400–500 hours saved in just three months.

While the productivity gains were clear, attendees agreed that traditional metrics—such as story points or cycle time—struggle to reflect the real impact of AI assistance. Some have started running side-by-side comparisons using past sprint data. Others rely on direct feedback from engineers to assess where AI adds the most value.

Agentic AI: Practical applications already in use

The discussion moved beyond basic AI tools to focus on agentic AI—systems that operate autonomously and carry out multi-step tasks with limited human input. Attendees shared concrete examples already in use:

1. Cursor AI for Refactoring SQL

Used to analyse and refactor legacy SQL codebases. In one case, over 900 stored procedures were reviewed using Cursor, accelerating output by up to 40x. The tool helped not only with syntax cleanup but also with identifying duplication patterns, unused procedures, and potential areas for optimisation.

2. Product Development Agents

AI agents that help automate early-stage product work, such as persona creation, solution mapping, and documentation. One attendee described using agents to rapidly develop user personas by aggregating public data, aligning it with internal user journey maps, and drafting functional product outlines that served as a starting point for cross-functional discussions.

3. Multi-Agent Systems (MAS)

Autonomous agents working in coordination—one generating code, another testing it, a third documenting it—mimicking the functions of a development team. In one example, a multi-agent system designed a basic feature end-to-end, then returned human-readable notes and integration steps, helping the engineering lead decide whether to implement or modify.

4. Legacy System Modernisation

Agentic tools helped parse and understand legacy systems (like COBOL), reducing dependency on domain-specific knowledge held by a few individuals. A participant shared how their team used AI to map out the dependencies across an aging monolithic application, flagging risky elements and suggesting components that could be containerised or rewritten.

5. Security-Wrapped Agent Frameworks

To meet regulatory requirements, some attendees described wrapping agents in code that tracks data flow for full traceability. These wrappers were used to ensure that LLM-based agents operating in sectors like healthcare and finance logged all activity, allowing compliance officers to audit and validate their outputs against policy.

6. MCP Servers for Workflow Orchestration

Using Model Context Protocol, attendees built workflows where agents autonomously accessed APIs, gathered data, and triggered actions without manual steps. One use case involved pricing intelligence: the AI queried multiple APIs for pricing data, summarised the output, sent updates via internal messaging tools, and triggered reports—automatically and on schedule.

Reviewing AI-generated code requires human oversight

Participants expressed caution about relying too heavily on AI-generated code. Some described scenarios where AI wrote code, created the tests, and approved the changes—resulting in logic flaws that went unnoticed.

In response, organisations have implemented stricter review processes:

• Senior engineers are required to approve all AI-generated pull requests
• AI-generated reviews are disallowed
• AI support is used during reviews, but not in place of human reviewers

Human accountability and measurement challenges

Participants also discussed optimal approaches to handling bugs in AI-generated code – as, they generally agreed, no AI tool is capable of producing perfect code.

In particular, “code bloat” was highlighted as a major issue. In other words, as AI operates so rapidly, it’s easy to generate huge amounts of code – and inevitably the more code that exists, the more bugs there will be.

Because of this, there was a consensus that:

• It’s essential to ensure human involvement at every step of the development process.
• “People must remain accountable for what they produce,” said one attendee.
• Code reviews – both by humans and tools such as CodeRabbit – are particularly important.

Clear rules and guardrails enable safe adoption

Governance came up as a priority. Attendees explained how their organisations define usage boundaries and assign accountability:

• Technical leads or architects determine how and where AI tools can be applied
• Guardrails are introduced to prevent misuse in code, testing, or deployment
• Some use lightweight wrappers to log all AI input and output, enabling auditability

This helps organisations embrace AI without losing control over quality, process, or compliance.

Privacy, security and regulatory requirements shape implementation

Security and compliance featured heavily in the conversation. Organisations are under pressure to ensure that AI tools don’t expose sensitive information or create audit gaps. A few attendees discussed:

• Creating data wrappers to monitor agent activity
• Preventing tools from leaking PII
• Building traceable workflows that can meet third-party audit standards

These measures allow AI to be used in sensitive industries like finance and healthcare.

Managing AI Costs and Infrastructure choices

As AI becomes more integrated into workflows, its operational cost has come under scrutiny. Attendees noted the rising expenses of using large language models through public APIs and the unpredictability of long-term pricing.

Alternatives currently under consideration include:

• Hosting lightweight private LLMs
• Running AI models locally on edge devices
• Replacing SaaS-based solutions with internal AI-driven systems built using MCP

These strategies aim to make AI adoption more sustainable while maintaining data privacy and performance.

AI is redefining engineering workflows

Attendees agreed that AI is changing more than just speed—it’s influencing how engineers work. By offloading repetitive or time-intensive tasks, AI enables developers to focus on problem-solving and innovation.

However, participants raised concerns about the learning curve for junior engineers. If AI handles too much, new developers risk missing foundational knowledge. Several organisations now combine AI tooling with mentorship programmes and hands-on training. One attendee noted that AI should be treated as a peer contributor: fast, efficient, but imperfect—always requiring human oversight.

Rethinking SaaS business models in the age of AI

Attendees discussed how AI—particularly LLMs and generative tools—are forcing product and commercial leaders to reconsider SaaS business models. One speaker described the tension between using AI to accelerate a content-driven business while also facing existential risk from the same technology. As AI agents shift how people search, consume, and generate content, traditional SEO-led monetisation strategies may become less effective. Another participant noted that the ease of creating AI-powered workflows has made it harder to differentiate products, urging businesses to focus on their unique value propositions. Several attendees agreed that pricing models would need to evolve—moving away from flat-fee SaaS toward usage-based or outcome-driven approaches—especially as LLM compute costs rise and multi-agent architectures become more common.

Final Thoughts: Success requires structure, not just tools

AI tools have moved from experimentation to execution. They’re helping engineering leaders solve long-standing problems, accelerate timelines, and explore new ways of working. But the session also made it clear that success depends on structure—on clear governance, trusted oversight, and shared learning.

Our next CTO Breakfast will focus on how organisations scale their AI practices across teams, while maintaining quality, security, and cultural cohesion.

We hope you’ll join us.

AI tools are capable of dramatically accelerating the pace of software development—sometimes by up to 20x. Our recent CTO Roundtable discussion explored the possibilities and how to reap the benefits.

Software developers can make huge productivity gains through the adoption of AI tools – sometimes with speed increases of up to 20x. That was the outstanding conclusion of our recent CTO Roundtable, Exploring AI-Driven Productivity in Software Development, hosted by Iain Bishop and Aleksandar Karavasilev, CEO and CTO of Damilah Technology.

Kicking off the event, they presented the findings from a series of controlled experiments they had conducted at Damilah, which demonstrated that time savings of up to 2x are quickly and easily achievable by using AI tools, in particular for tasks such as:

• Quality analysis
• Coding
• Unit testing

Furthermore, when engineers already experienced in using AI performed the set tasks, they were able to operate 4x to 6x faster.

While Iain and Aleksandar highlighted these impressive gains, the discussion revealed even greater possibilities. A start-up founder explained how combining a series of AI tools for different stages of development enabled him to deliver productivity improvements of up to 20x. To do this, he used:

• Perplexity for market research
• Replit and Lovable for building apps
• Tabnine for auto-completion of code

This illustrated how companies willing to fully embrace AI across the development lifecycle can potentially achieve extraordinary results – for example, by enabling them to rapidly put new product ideas in front of users as basic MVPs. As a result, this founder now plans to launch at least three start-ups within a year.

Increasing dynamism

This kind of accelerated development cycle can usher in a whole new era of dynamism in the market, it was concluded—the main points being:

• Start-ups are usually lacking capital.
• AI enables them to develop an application at pace and at low cost, before significant sums of money need to be invested.
• This helps to de-risk new ventures and make it easier to attract funding.

But even companies who take a steadier or more cautious approach to AI-adoption, it was agreed, can still make immediate productivity gains. This could be, for example, by starting with ChatGPT, then progressing to more sophisticated tools such as GitHub Copilot or Cursor.

Overcoming barriers and managing adoption

The conversation also turned to barriers to AI adoption. Several of those present described a certain amount of reluctance or resistance among larger enterprises, often due to concerns relating to company policies and regulatory matters.

However, it was noted that simply banning the use of AI tools can lead to “shadow AI”, where developers, in particular, will find workarounds in a constant pursuit of innovative ways of working.

Several attendees agreed that the solution to this is not to block, but rather to encourage the use of specified AI tools with some strict guardrails in place to ensure certain boundaries are not crossed, particularly with regard to security, privacy and IP protection

Addressing recruitment challenges

Questions were also raised about whether developers should be allowed to use AI during the hiring process, as often this could mask an applicant’s true capabilities when it comes to coding and testing. However, as one participant pointed out, companies should be looking to hire employees who are adept at using AI, as having an AI-proficient workforce will be necessary to maintain competitive advantage in the future.

Another attendee recommended a solution to the hiring problem: they deliberately asked applicants to use an AI tool which they knew would produce a certain bug in its code. They were then able to check whether the applicants had detected the bug, and therefore whether they were capable of testing and fixing code without the use of AI.

Furthermore, as one participant noted from their own experiences, many developers will decline to work for a company that denies them the opportunity to use AI tools to enhance and accelerate their work.

Encouraging adoption and keeping pace

One recommendation to encourage those more hesitant about AI, alongside formal training, was to hold informal “brown bag” sessions. The participants from Damilah talked about their own experiences of this – holding the sessions once a fortnight, and inviting anyone from their own company or their client base to share their experiences and knowledge of AI tools.

The importance of keeping on top of rapid changes in the quality of AI tools also emerged. Several participants pointed out that some, which didn’t perform well a few months ago, are now proving to be highly valuable – software-testing tool CodeRabbit being a good example of this.

Human accountability and measurement challenges

Participants also discussed optimal approaches to handling bugs in AI-generated code – as, they generally agreed, no AI tool is capable of producing perfect code.

In particular, “code bloat” was highlighted as a major issue. In other words, as AI operates so rapidly, it’s easy to generate huge amounts of code – and inevitably the more code that exists, the more bugs there will be.

Because of this, there was a consensus that:

• It’s essential to ensure human involvement at every step of the development process.
• “People must remain accountable for what they produce,” said one attendee.
• Code reviews – both by humans and tools such as CodeRabbit – are particularly important.

Measuring success

The roundtable also discussed the challenges related to measuring the gains achieved by AI. The key points raised were:

• Software development, by its very nature, involves performing a different task for each new project.
• Therefore, the only way to effectively measure productivity gains is to have two teams performing the same tasks in parallel – one with the use of AI tools, and the other without, to act as a control group.
• AI tools, used in the right ways, can clearly unlock major productivity gains – but many businesses do not have the time or resources to conduct such controlled experiments to measure the gains with any accuracy.
• In light of this, the tests conducted by Damilah have proved invaluable.

Sharing knowledge and expertise

Our roundtable participants agreed that keeping up with the lightning pace of change in AI is almost impossible. It therefore requires individuals and companies to work together to share knowledge and experiences for the advantage of all.