Last week I had the good fortune of attending Dublin Tech Summit (DTS). The two day conference was filled with lots of discussions about AI, quantum and the opportunities and challenges for European businesses. Big name speakers from the likes of OpenAI and Anthropic were present.
Below are my six key takeaways, grouped by theme. As you read them, please bear in mind there is always going to be some hyperbole and exaggeration at conference talks; it benefits people to big up their industry. The core concepts shared in our “AI for Business” livestream series still ring true; start with a problem you want to solve, look at your processes from end-to-end and then look at tools that can help you.
Agentic AI:
You’re probably all sick of hearing about AI at this stage, but I must cover the topic as the majority of talks at DTS discussed AI, with many delving into Agentic AI; the hot new thing on the block.
AI versus Agentic AI
AI-powered co-pilots have become commonplace. OpenAI’s CFO Sarah Friar claimed that 40% of production code in Walmart is now co-piloted or written solely by a co-pilot and that OpenAI is now at the level of a graduate programmer with three or four years’ experience. This doesn’t feel that new.
However, agentic AI goes beyond co-pilots. Co-pilots require humans to be kept in the loop but agentic AI can connect apps and workflows together without the need for humans. To harness agents, you need to have good documentation, well structured data and accessible APIs. This advancement brings with it both exciting and daunting possibilities.
Disruption
One area that could be heavily disrupted by agents is design, especially in a B2B context. If agents will be interacting with other agents and humans feature less in the cycle, your space for UX and UI design will be a lot smaller and your user journeys will become a lot shorter. Aesthetics could become less important as the UI becomes simply an agent orchestration layer. Data strategy could become more important than product design.
The best way to prepare for potential disruption is through constant learning. At DTS, attendees were advised to focus on deepwork monotasks rather than multi-tasking.
Another area ripe for disruption could be the very popular SaaS (Software as a Service) business model. SaaS tools are task focused but agentic AI is outcome focused. Soon we might have a stack of agents to call on rather than a stack of technology tools.
The exciting opportunity is that smaller teams can become a lot more potent and powerful. See the “Investing in AI” section for more on this.
Not there yet
Fergal Reid, the VP of AI at Intercom made the point that automating jobs end-to-end is still really hard. While co-pilots are good at writing code, Fergal questioned what percentage of a developer’s job is spent actually writing code versus other tasks like thinking about the problem to solve and debugging the code they have written. This feeds into another limiting factor for AI adoption: it’s difficult for employers to quantify the value of AI.
Fergal believes that while agents could reach a point where they can automate jobs end-to-end, especially if the breakthroughs in the sector continue at the current pace, they’re not there yet.
Pulse check from the floor
One thing that stood out to me from talking to attendees who are experts in the AI space, rather than the big name speakers, was that we may actually be underhyping the power of the technology itself. A common thread I heard was that the technology is actually more powerful than we realise, we just haven’t harnessed it properly yet. People don’t realise how powerful the models have gotten and it’s easy to have opinions that are a year or two out of date. For example, if you tried an image generation tool a couple of years ago and found it frustrating or inaccuracte, it might be worth checking it again to see if improvements have been made in the intervening two years.
I can’t validate this as I’m not a technical expert in the field, but I did hear this from several attendees who are experts in different areas of AI and machine learning.
Investing in AI:
I attended several talks to get an idea of the funding landscape for founders, and again AI dominated these discussions. Last year, one third of VC capital went to AI companies. Many VCs believe we are witnessing a seismic shift comparable only to the internet itself, and bigger than blockchain, fintech and even the advent of smartphones. This shift creates opportunities.
Firstly, small teams can be far more powerful if they harness AI and agentic AI correctly. They can do more for less and they can do it faster. For example, Loveable grew to \$50 million in ARR (Annual Recurring Revenue) with just 30 employees.
This has a knock-on effect on how VCs fund companies. The VCs I spoke to believe that the days of arriving with just a pitch deck to fund a Version 1 are over. You should now have some form of working platform or prototype in place.
There was also a theory that the size of rounds will change drastically. While a seed round may remain at around the €500k mark, the next step will likely be in the vicinity of €30 million rather than €3 million. That in-between space could be squeezed out because concepts and markets take less time and resources to prove.
The market is also volatile. Chegg, a homework-helping Ed tech company, was used as an example of a company whose value hemorrhaged with the release of Chat GPT-3.5 in November 2022. By November 2024, Chegg’s stock price had fallen 99%, primarily because of competition from ChatGPT (although it is worth saying that it had begun to fall from its 2021 high even before the release of ChatGPT).
In the age of AI, market share is much, much harder to defend, especially for GPT wrappers. Speed of execution is essential to stay on top of the pile.
However, the funding playbook hasn’t been completely torn to shreds. VC’s will still look at three core elements of your pitch:
- What is the problem?
- Is there a big enough market for it?
- Can your team execute on it?
Some things never change!
Quantum is coming:
Quantum is most definitely coming. While most of the quantum talks at DTS were off-broadway in smaller rooms, they were standing room only.
Probably my favourite talk of the conference came from Brendan Barry, founder of quantum company Equal1.
Brendan ran through some recent breakthroughs that have seen the hype around quantum begin to grow again. The ability to manipulate electrons is now at production scale, rather than something only seen in university labs. Pure silicon has recently become available and the D-Wave paper from March 2025 was the first to demonstrate quantum supremacy on useful, real-world problems. While the jury is still out on just how seismic a breakthrough this is, it’s definitely a step forward.
Building better manufacturing and fabrication technology is a bottle neck for the industry. With quantum you’re working on the atomic level, so exceptionally precise tooling is required, as are purified geranium and silicon. These are both tightly controlled substances and special licenses are required to import, export and transport them. The qubits in a quantum computer need to be cooled to a temperature colder than space and designing technology to cope with that is extremely challenging.
But these blockers will likely slow down development rather than halt it. And the potential use cases for quantum computing are exciting, from better drug creation (by understanding how molecules will combine) to less energy consumption (the Equal1 quantum computer uses the same amount of energy as a hairdryer to run). It may even be possible to train full AI models in minutes by eliminating the need for back propagation (in theory at least).
These things are still quite a way away. For now, quantum computers can do some things that a classic computer can do and they can do them slightly faster, slightly better or with slightly less energy. But this industry is changing quickly. Watch this space.
Regulation V Innovation:
With about one third of the groundbreaking EU AI act now in force, a debate raged over the course of the conference about Europe’s stance on regulation, particularly in the AI space. Everyone seemed to agree that it was not binary and that a balance between regulation and innovation needed to be struck. But there was a clear schism between pro-regulation and anti-regulation camps.
For those opposed to strict regulation, many expressed a concern about regulating something before we even know what it is capable of or how it will be used. A comparison was made to adding a seatbelt regulation to cars before the Model T was even invented. This camp questions whether regulation can even achieve the goal of safer technology.
Furthermore, in the wake of the US lifting regulations around AI at the federal level, they feel that Europe will regulate itself away from the cutting edge of the technology, losing our place at the table in the process. The feeling is that you can be innovative or regulated but not both and that innovation is more impactful in the long run.
The arguments against this stance are also numerous. Firstly, Europe has a largely unregulated defense sector. Despite this, we are even further behind the US in this space, so less regulation does not guarantee more success.
Secondly, they argued that self regulation never works (and to be frank, I agree completely with this). Governmental level regulation can even the playing field, creating the rules by which everyone needs to play.
Having rules and regulations in place can create new industries, as evidenced by the numerous booths at DTS filled with companies focused on AI regulation, compliance and safety. Being compliant can become a competitive advantage and companies in Europe should be used to this since the introduction of GDPR legislation which did not cause Europe to fall behind in other areas of technology.
Both sides seemed to agree that the timing of regulations is key to their success and that good governments are taking a proactive stance. In Ireland, both the National AI Strategy and guidelines on the responsible use of AI in the public service sector have been launched recently.
Regulatory sandboxes, like those operated by the ICO the UK, could be a great path forward to guide and encourage companies to release new tech safely. This could lead to AI safety by design.
Infrastructure:
Again, this was a common theme right across both days of talks. The pressure on infrastructure like water, energy and transport to continually feed the technology beast, creates both major challenges and opportunities for governments and companies alike. Solving these challenges could leave a nation at the forefront of the new AI world. These opportunities exist not just for AI related businesses, but for adjacent industries like energy and manufacturing who could solve these problems and break open bottlenecks.
Either way, governments need to take the maintenance and improvement of their national infrastructure very seriously.
Humanoid robots:
I’ve got to say that I’m not convinced on this one, but I did attend a very interesting talk on the subject. China, thanks to some heavy government investment, is surging in front in this space. While there are only about five hundred thousand humanoid robots in existence currently, China is aiming to have around forty million by 2030.
Unlike LLMs, which have lots of data available to be trained on, these robots need to be trained in real world scenarios because there is no easy data to input. This means training takes longer. It also conjures up quite a funny image of a load of robots banging into each other in a warehouse somewhere as they clean, cook dinners and do other household tasks.
While automation and robotics will undoubtedly be helpful in people’s lives, the humanoid aspect is that part that is yet to convince me. For example, we already have robotic lawnmowers which are specifically designed to complete that task. A humanoid robot pushing a regular lawnmower around a garden seems laughable.
But given the heavy investment going into the space I’m going to keep an eye on it to see if I will be proved wrong.
6 cool companies:
Finally, here are six cool companies that stood out to me from the conference that you may not have heard of. Definitely work checking out!
- Audire.ai - Uses powerful AI to complete quality assurance checks on 100% of business interactions (the current standard for quality assurance checks is around 5%). This helps teams to deliver a higher standard of customer service and compliance.
- Clear.ip - A really clever use of AI in conjunction with human patent attorneys to help startups patent their idea for a fraction of the normal costs. A great shout for any founders out there!
- Motoklik - Another cool use of AI, this time for thrill seekers. Motoklik uses AI to tell motocross riders how to set up their suspension properly, allowing them to ride with maximum confidence.
- FoodCloud - Food waste is a huge contributor to energy consumption and waste, a key topic in the discussions around infrastructure. FoodCloud is on a mission to change that. Their platform reduces food loss and waste through surplus food redistribution and I have been a big fan of theirs for some time.
- Jentic - An agentic company worth checking out. Jentic allows you to plug your AI agents into a vast open repository of API operations and workflows with a single turn-key integration.
- Equal1 - The world’s first rack-mounted silicon quantum computer. It is definitely worth checking out if you’re interested in the quantum space.