Fifty people watched the panel discussion in London in May 2026 on the theme of Engineering 2028: A Leadership Masterclass. Chaired by Iain Bishop, the panel included: 

• Lucinda Faucher, VP of Product at Valve (platform for sales, marketing and distribution for flexible workspaces) 

• Alttaf Hussain, Director of Engineering & AI Innovation at Yoti (platform for privacy-focused identity and age solutions) 

• Vishal Manani, CPTO at Eckoh (platform securing sensitive data in contact centres) 
• Chris Parsons, CTO and independent consultant, writer & AI strategist 

Where to start with AI orchestration ​ 

A theme that continued throughout the session through to the final questions was how to get teams on board with AI orchestration. Vishal Manani gave the example of his software engineering colleague, describing his role as similar to an artisan baker’s. The colleague was loathe to become a machine operator making white sliced bread. While this example is of someone who respects their trade and the outcomes from his work, the panel knew of others who simply enjoyed coding for the sake of coding.  

Leaders need to understand the motivations of individuals in their teams to bring them on the AI journey. As stated in our Engineering 2028 report, the demand for software is growing with the rise in productivity gains from AI. However, no one shied away from the fact that some people will lose their jobs. Those who are using AI now have a greater advantage and will succeed longer term.  

Engineers’ roles are changing rapidly. Job descriptions still haven’t changed but engineers must accept that they will spend most of their time refining agents rather than writing features directly. Those who refuse this kind of role will eventually need to look to other professions, as traditional engineering roles disappear.  

This doesn’t mean giving up. Leaders can have a huge impact on adoption rates. Encourage those who have embraced AI early on to become AI champions in the team. Given them the licences they need to try out what they want to do. Then get them to share their successes. This kind of positive reinforcement is a tried and tested method of increasing technology adoption, and it works with AI too. 

Speed vs quality 

The tension between speed vs quality with AI is widely discussed. The software development lifecycle (SDLC) is changing, with bottlenecks moving from coding to reviewing.  

“We are in a point in history where you can go from vibe coding to production at blistering speed.” – Alttaf Hussain, Director of Engineering, Yoti 

Especially in private-equity backed companies, AI is being jumped on to pick up the pace of software creation. However, the panel was concerned that what comes before and after code generation has not kept pace. Requirements, review and governance are still reliant on humans. With AI possibilities of creating more code than ever, organisations need to consider that there is still a barrier to achieving results at speed, unless we want sub-standard products. 

Human oversight will still be critical for the foreseeable future, as AI’s limits with judgement are clear. One example shared by Vishal Manani was of an AI agent writing brittle test and then changing the application’s code to make the test pass. This caused an outage. You can speed up one area of software development, but it will result in other areas slowing down. 

The human moat 

When asked about what stays with us, the humans, the panel agreed on one area above all: accountability.  

“When things go wrong, who owns it, monitors it, reviews it?” – Lucinda Faucher, VP of Product, Valve 

Chris Parsons referenced Tristan Harris’ documentary as an extreme example of where lack of accountability can lead to with AI. If companies argue that they are not responsible for someone’s “AI psychosis”, despite creating the AI caused it, who takes responsibility for disastrous consequences?  

Traditionally, the person who writes the code takes responsibility for its accuracy. With AI, the lines are blurred. It can build fast, but it might build entirely the wrong thing. Again: the SDLC bottleneck is simply moved – not eliminated. AI needs regulatory compliance and security oversight built in from the start, with a “human in the loop” throughout the SDLC.  

Measuring success  

According to Lucinda Faucher, nothing has really changed when it comes to measuring success. Metrics like revenue generation and customer satisfaction is what any business and its stakeholders will care about. Ultimately, what we build needs to solve our customers’ problems. Otherwise, what’s the point?

A question came in from the audience around token usage being used as a measure of success by Meta, and whether the panel saw that as a good metric. Chris Parsons disagreed – proxy metrics may be useful to measure steps to getting to where you need, but technical metrics don’t have the same meaning any more. Bottleneck identification is more critical to address than lines of code or token counting.  

Five top tips to implement AI orchestration   

The panel’s recommendations around AI orchestration were relevant to any organisation. Here are the five top tips we summarised from the event: 

1. Start with low-risk, controlled workflows in non-safety-critical areas, as AI is still not ready to review AI-generated code without humans in the loop
2. Build evaluation frameworks before scaling AI adoption, as AI output is non-deterministic, and decide “what good looks like” before implementation
3. Create an AI champion programme, led by your early adopters, to bring more of your people on the AI journey
4. Focus on your requirements quality and formalising human checkpoints to speed up the SDLC  
5. Establish clear governance and accountability structures with your leaders and stakeholders from the start, as you can expect things to go wrong sometimes 

Following the event, several people felt that the conversation was just getting started. How do we persuade our leaders and stakeholders that we may need to slow down to go faster? How do we get consensus on “what good looks like” before implementation? If you’re a technology leader who would like to discuss practical next steps around leading human + AI teams responsibly, follow our Linkedin and subscribe to our email newsletter to hear about our events before anyone else. 

To introduce the Byte, I shared findings from our Engineering 2028 survey. Almost all survey respondents were already using AI. Half reported the greatest benefits in AI being increased speed and innovation.  

I shared how productivity expectations are rising more sharply among teams that have more AI experience. This lies in contrast to the usual technology “hype cycle” trends, where increased utilisation leads to disillusionment.  

Teams tend to start using AI tools following a co-pilot model. They use them to accelerate the way they have always worked, but as experience increases, they find new ways of working to fully utilise the AI tools. 

There has always been latent demand for more software in product software companies. AI removes the barrier to product creation by creating code at lightning speed, leading to increased demand. 

This all paints a rosy picture… so what’s going to slow us down?  

• Capability gap: Engineering is faster than ever, making leadership harder. AI adoption is a change management process.  

• Operating model gap: In such a fast-changing environment, standards and process can’t keep up. There are fragmented tools and shadow AI usage.  

• Governance gap: Compliance and security concerns, IP leakage and other risks are emerging. Accountability and ownership are unclear.  

We can’t simply lock down AI – so how do we, as CTOs and technology leaders, control what’s going on?  

“The problem with autonomy without alignment is that it doesn’t scale.”

The path to engineering orchestration​ 

As competition, stakeholder expectations, and advances in AI technology increase, we need to facilitate the safe adoption of AI technologies. We need to create an environment where we won’t get in the way of progress and increased productivity. But we need to take safety, compliance and consistency seriously. 

Technology leaders need to change the way we work: we need to move to a model of engineering orchestration. This means using a platform that allows us to manage the use of agentic AI workflows while retaining oversight and control over what is produced.  

The human controls agentic workflows across every aspect of the software development lifecycle (SDLC). All standards, such as technical, regulatory and design systems, are included in the orchestration platform. This is the shift from doing work to designing systems that do work.  

Changing skills priorities  

As engineering teams become orchestrators, team sizes will reduce and individuals will become more versatile. New roles will appear such as Product Engineer – engineers who understand customer/end-user needs, as well as being able to orchestrate the solutions. Other Engineers need to develop system thinking. They need to understand the bigger picture around architecture, non-functional requirements and design patterns. 

As development cycles shrink, miscommunication will have an even greater negative impact, so leaders need to consider cultural differences, as well as time zone and proximity even more carefully.  

“Being fluent in AI is becoming as important as being able to type on a keyboard.” 

Skills needed are evolving. Prompt engineering is a basic need for people involved in software engineering. A key differentiator will be commercial understanding: really understanding the value we bring to our companies and customers. The best engineers will be creative and curious – exploring what’s possible and approaching problems from a different angle. 

The “human moat” surrounding orchestration and governance is a set of unique, irreplaceable human skills: strategy, leadership, empathy, ethics and creativity. 

Our role as leaders is to hire for, grow and retain these skills, understanding that our own role is being redesigned. We need to: 

• Move on from managing tasks to designing systems 

• Raise AI fluency across our organisations 

• Strengthen governance without slowing innovation 

• Build product-focused, commercially aware teams 

What AI orchestration looks like in practice 

Our CTO, Aleksandar Karavasilev, demonstrated to the CTO Craft community Damilah’s multi-agent platform (DMAP) to bring AI orchestration to life. Here are the principles behind our AI orchestration platform: 

1. Start with context, not prompts 
   DMAP begins with an explicit project context: the technology stack, governance constraints, standards, and which models handle which tasks. Every agent working on your project operates from the same rules.  

2. Run work through a structured workflow
   DMAP supports agentic workflows within all stages of the SDLC, such as requirements, design, development, testing and validation.  

3. Humans in the loop by design 
   Work can be paused, reviewed, approved or rerouted at any stage. You define where human sign-off is required, what a pass looks like, and who is accountable.  

4. Every decision is auditable 
   DMAP keeps a full record of what ran, what was produced, what a human approved, and why anything was rerouted.  

5. Cost management 
   With model costs changing often, it is critical to monitor token usage and costs. DMAP includes this on a dashboard for full visibility and control. 

Take the next step towards AI orchestration 

Those who are continuing to add AI on top of their existing SDLC, relying on experienced people to absorb the extra review and governance load, will soon struggle. Those who redesign how work flows through the system entirely will remain competitive.  

At Damilah, we are helping companies understand the barriers to AI orchestration. If you are looking into making the change, get in touch today to set up a conversation with me and the team.  

Iain Bishop
CEO, Damilah

Watch the recording

CTO Craft Con London felt different this year. A year ago, discussions were around AI pilots, with people getting excited about the companies who were taking the next step with AI.

This year, there was a mix of nervousness around the dramatic changes we are already seeing and competitiveness amongst those who were steaming ahead (primarily AI-native start-ups). 

We partnered with CTO Craft to survey technology leaders about how engineering might look in two years’ time, leading to the report, Engineering 2028: Leading Human + AI Teams Responsibly. What struck us at the London conference was how closely the report aligned to what so many people told us about their experiences. 

“The shift in mindset over the past year has been striking. At CTO Craft 2025, many leaders were still cautious about AI and how it would impact our industry. This year, everyone agreed that AI is already changing the industry, and at speed. After only a year, we’re mostly hearing about how companies are using AI in practice.” – Aleksandar Karavasilev, CTO at Damilah

Headcount is not shrinking as AI increases output 

According to our Engineering 2028 report, increased AI-enabled productivity drives increased demand for more software. Ideas that were sitting in a “maybe later” pile are becoming viable and roadmaps are rapidly stretching. Companies need to maintain their headcount to keep up with the latent demand for new features and products.

The CTOs we met at CTO Craft were from a range of companies. Those at investor-backed software companies largely agreed with what was said in the report, as they face increased pressure to deliver AI-based solutions. At more established or non-software companies, issues were still around legacy systems preventing advancement with AI, meaning humans are more in need than ever. And at the opposite end, there were software start-ups with a couple of people able to create a brilliant product from scratch with AI, which would have previously taken months and a critical mass of software engineers. 

Speed is shifting the bottleneck in the SDLC 

“For so long, engineering has been the bottleneck in the software development lifecycle (SDLC). As that constraint starts to ease, pressure moves into other parts of the lifecycle, especially discovery, alignment, and review.” – Iain Bishop, CEO at Damilah

In Iain’s talk at CTO Craft, he highlighted the findings of the report Engineering 2028 around AI transforming the SDLC. New features or products are possible to generate in a fraction of the time with fewer people involved, but they are not perfect and not necessarily built for purpose. So, while AI rapidly accelerates code creation, removing a major bottleneck in the SLDC, when it comes to code review, we see new delays. 

Moving towards AI orchestration 

This displaced bottleneck is changing how software engineering teams work every day and how they are structured. In our Engineering 2028 report, organisations earlier in their AI journey reported using AI to help engineers move faster on individual tasks. Organisations further along the journey are redesigning the teams altogether.

“Many of the presentations I saw at CTO Craft validated and reinforced our view that the future of engineering is orchestration. Companies are increasingly using AI platforms to support how work flows through teams.” – Iain Bishop.

Software engineers need to become “AI orchestrators”, setting tasks for multiple agents. The need for human review of AI’s output increases, as well as checking agents’ processes throughout the development process. 

Governance must keep pace with AI 

An interesting talk from Sonar compared the difference between different LLMs when it comes to creating code. Newer models create significantly more lines of code. And the more code you produce, the more bugs will be produced. Sonar proved this and, while the bugs were minimal, they do stack up when it comes to writing millions of lines of code in seconds. This means newer LLMs inevitably produce more bugs. 

So, speed is improving, but quality is not automatically keeping pace. This backs up what respondents felt in our survey with CTO Craft, where only 8% are seeing AI generate higher quality outcomes. More code means more to review, so governance needs to accelerate in line with the speed of AI’s coding. Governance needs to be built into how teams operate from the start. 

What AI cannot replace

As Iain said in his talk: “AI builds the how. We own the why.”

The combination of judgement, strategy, empathy, and a deep understanding of both the customer and the business determines whether good work gets built at all. After the Women in Tech breakfast, Damilah’s Head of Marketing, Julia Valentine, reported:

“One woman I spoke to said that ‘soft skills’ sounds like ‘easy’, i.e. the opposite of hard. But the one thing AI can’t do is ‘soft skills’. People who excel at soft skills (these days, often women) will be more desirable in tech leadership than those with ‘hard skills’, such as coding. Could this be an opportunity to seize gender equality in tech?”

AI is highly effective at execution, but it does not understand context in the way people do. It does not carry accountability, nor does it understand what matters most to the business or customers.

As Hywel Carver, CEO at Skiller Whale, put it: “The one thing AI cannot do that humans can is empathy.” 

How engineering teams are evolving  

As AI takes on the kinds of tasks engineers used to focus on, engineers now need to get a better grasp on product, even transforming from software engineers into product engineers. This means understanding what both customers and what the business want to achieve, connecting technical decisions to commercial outcomes. 

Leaders need to understand this shift to develop the right people in their teams, as well as hire the right people into new roles that may arise. CTOs’ and CPOs’ roles are becoming more closely aligned, and we are seeing the rise of the CTPO role.

As Iain put it in his talk, “Engineering in 2028 isn’t a tooling upgrade. It’s a leadership redesign.” 

The organisations that come out ahead over the next few years will be those that have genuinely rethought how their teams are organised around them: moving from the focus on AI tools to AI orchestration. 

They’ll be able to test those products quickly, then easily adapt them based on insights gained and iterate at pace.

Companies will be able to rapidly adjust their products, create multiple variants, make huge pivots – or anything in between – in response to changes in market conditions and new opportunities, all the while creating outputs of higher quality.

And all this will be possible at significantly reduced cost – thus negating the need for large amounts of high-risk, early-stage funding. Instead, investors will be able to place their bets on an MVP, or adaptation of an existing product, that is already proven to be successful and scalable.

In all of the above, the potential value will be created by the smart integration of the right AI tools to accelerate and refine processes at every step.

However, this isn’t the distant future. In many cases, it’s already happening.

Yet for many firms – particularly those backed by private equity (PE) investors – the reality isn’t always matching up to the vision.

So why does this gap exist?

Overcoming the blockers to AI-driven value

Because a typical PE investor will be looking for a substantial return over a period of three to five years, many initial investments will have been made prior to the explosion of generative AI tools that are now massively disrupting almost every industry. Usage of these tools may not have been part of the original investment thesis – but they absolutely need to be now.

However, too many organisations are failing to keep ahead of the game and unlock the enormous potential value that AI tools can deliver.

In our experience, the main blockers are as follows:

1. Time constraints

We are seeing many companies suffering from a simple lack of bandwidth to explore how best to leverage the capabilities of AI.

This is often compounded by intense pressure to deliver results against the original investment thesis. Pressure builds on top of pressure as new opportunities for growth are discovered or acquisitions are made which then need to be carefully integrated. This can leave little room for reflection or experimentation with new tools.

Add to this the rapid pace at which AI tools are constantly evolving and improving, and it can seem almost impossible to keep up – let alone move ahead of the game – when it comes to understanding the best ways to deliver practical, value-driving applications of the technology and successfully roll them out across the organisation.

2. Hype versus reality

As has always been the case whenever a new, heavily hyped technology floods the market, most AI tools are currently not mature enough to deliver fully on the promises made by their vendors. There is no doubt whatsoever that AI is a game-changer. But being able to work out the difference between the sales pitch and the practical reality can be challenging and requires deep expertise.

3. Security, legal and compliance challenges

Legitimate concerns exist around issues such as security, regulatory compliance and IP-protection – many of which are yet to be clarified and resolved. While most AI tools offer, for example, zero data retention and assurances around IP, understanding and mitigating these requires time, experience and focus – for example, in ensuring the tools are correctly configured to be fully compliant.

Overall, it’s vital to address these issues. And soon.

If your company takes too long to release new products or new features, you may quickly find yourself in trouble as your time to value becomes severely eroded. Your new competitors – who could be almost anyone armed with the right AI tools and the ability to use them effectively – can already enter the market and pull the rug from under your feet, able to adapt to market demands and seize opportunities far faster than companies with legacy platforms and products.

Learning from best practice

So how can you start turning AI into tangible value?

We believe that working with a skilled and knowledgeable partner is a powerful first step. Ideally, you would choose one that is constantly tracking the latest advances in AI technology and how they can be applied– safely and legally – to deliver greater value in the shortest possible timeframes.

They would also be able to consult on best practice, drawing on their own experiences in the field alongside those of similar clients they have supported.

The aim would be for them to help you deliver high-quality outcomes using the most effective AI-enabled processes and techniques.

To address legal and security issues, it’s essential to have an AI policy in place that creates the appropriate guardrails for safe and compliant usage of AI. Above all, we strongly recommend that absolutely everything always remains subject to human accountability – so have your people reviewing and refining every AI output at every stage of your processes. Here at Damilah, we do – and always will.

Last, but not least, there needs to be a shift towards outcome-focused roles. In other words, enabling AI tools to handle more of the laborious, time-consuming, detailed technical work – thus allowing a skilled human workforce to maintain oversight while concentrating on value creation and strategy.

New pricing models and investment strategies

And here’s one further thought. It’s likely that agentic AI tools will soon become prevalent in many organisations, which may have a profound effect on commercial pricing models.

Instead of the traditional pay-as-you-go or seat-based SaaS pricing structures, we may soon find outcome-based pricing models becoming the norm – that is, where fees are based on successful delivery and results.

This, in turn, would require PE houses to review – and radically adapt – their investment strategies, resulting in major impacts on the organisations they back.

It’s too early to predict precisely how this AI-powered future will unfold. But one thing is certain: all organisations – and particularly those funded by PE – need to remain fully alert to the fundamental shifts that are occurring and be nimble enough to rapidly adjust. Those that don’t risk becoming dead in the water.

We can help you transform your AI vision into genuine value . To explore the possibilities, get in touch now.

Iain Bishop, founder and CEO, Damilah

 

We therefore ran a controlled experiment, pitting a team of AI-assisted software engineers and quality analysts against a ‘human-powered’ group who were only allowed to use their own brains.

Here’s what we did and what we discovered…

How we ran the experiment

In all, a total of 52 of our people took part in our experiment, over a series of two hackathons.

We split them into two groups:

• One assisted by AI tools, with 30 engineers and 8 automation quality analysts (QAs)
• A control group that was purely human-powered, with 12 engineers and 2 automation QAs – allowing us to baseline the potential gains of using AI

We gave each group three hours to perform the same task.

For the engineers, the task was:

• Develop a .NET Web API to dynamically process mathematical expressions
• Implement a custom PEMDAS-based algorithm for expression evaluation, without using any third-party libraries
• Write unit tests to validate the functionality of the solution
• Test the implementation using five provided edge cases

And for the automation QAs:

• Write three automated functional test scripts for specified scenarios
• Test these functions using the provided test web shop application

The tools used by the AI-assisted team were:

• GitHub Copilot
• Cursor IDE (using ChatGPT)
• Qodo (was Codium)
• Tabnine

The results

We expected the results of this experiment to be positive in favour of the AI team, but we were still amazed by the difference the AI tools made in terms of enhancing speed and quality.

Here are the overall outcomes we recorded… 

On average, when compared to our human-powered team, our AI-assisted engineers were able to:

• Complete the coding nearly 2x (44%) faster

• Conduct the unit tests just over 2x (51%) faster

• Cover nearly twice as many (83%) more edge cases

And when we compared our fastest human-powered engineer with our fastest AI-assisted engineer, the results were even more impressive: the AI-assisted engineer was nearly 5x (78%) faster.

We also compared a human-powered engineer with one who already had experience using AI tools (in this case, GitHub Copilot). We found that:

• For the coding, the AI-powered engineer was 4x (75%) faster

• For the unit tests, the AI-powered engineer was 6x (83%) faster

This demonstrated to us that, as our team of engineers become more experienced with AI tools, our productivity gains will increase even further.

For the QAs, we also saw a significant improvement in the times it took the AI-assisted analysts: the average overall time was just over 2x (54%) faster with AI.

And, as with the engineers, we compared the two fastest times, and found that the first AI-assisted QA to complete the task was 9x (89%) faster.

A comparison of AI tools

We also aimed to make some comparisons between the four different AI tools that we used in the experiment, in particular with regard to user experience, productivity gains, and security and IP protection.

For user experience: GitHub Copilot came out on top. Our developers rated it as a robust and mature tool, suited for .NET application development. It offered consistent suggestions and responses as well as strong context management. Cursor and Codium came in joint second place.

For productivity gains: Cursor came out on top, allowing our team to be 3.2x (69%) faster than human-only developers when it came to completing the full task. GitHub Copilot was in second place, making the team 2.7x (62%) faster.

For security and IP protection , we found the following:

• GitHub Copilot transmits code snippets from the integrated development environment (IDE) to GitHub in real time to generate relevant suggestions. Once the suggestion is created, both the prompt and the suggested code snippet are immediately discarded—but note that this is only the case for the Business and Enterprise licence options.

• Cursor provides a Privacy Mode that can be activated during the onboarding process, ensuring that no code is stored on their servers or by their sub-processors.

• Qodo: Paid licence user data is not used to train its AI models. The data is deleted from their storage after 48 hours. Also,
they provide an option for a zero-retention policy, where data is removed immediately if users specifically request.

All three tools are certified for SOC2 compliance.

(Note that we didn’t assess Tabnine as we felt the model wasn’t mature enough and its users struggled to complete the task.)

Conclusion and our next steps

Our experiment made it clear that AI could offer us some huge benefits in productivity and quality. In every aspect of the tests we conducted, from coding to unit tests to automated test script production, there was a clear time saving – in most cases very significant. It will also enable us to improve the quality of our outputs, as AI-generated code was able to give us broad edge case test coverage.

Furthermore, we expect our efficiency gains to improve further as it is clear that development speed increases with experience when it comes to using AI tools. We found that just a short amount of training significantly accelerates outputs.

As for the future… Our product owner colleagues also ran an experiment to understand how AI can accelerate and improve the product discovery process. We are now looking at how we can use their AI-generated requirements as prompts to build applications – ultimately with the possibility of using AI-assisted processes from an initial description of requirements right through to final outputs.

Meanwhile, right now, we’re already starting to reap the benefits of AI-assisted development with some of our clients, delivering even greater value for them.

If you’d like to find out more about how AI-assisted software development can benefit your business, get in touch now.

Aleksandar Karavasilev, CTO at Damilah