The gap between AI ambition and AI delivery is the defining business challenge of this decade. Most businesses understand by now that AI is not optional. What far fewer have figured out is how to move from a boardroom demo to a system that is running, useful, and actually valuable.
In May 2026, the Rise of AI Conference gathered in Berlin for its tenth edition. Over 300 of the most serious AI thinkers, builders, and policymakers spent two days working through the hardest questions the industry faces. We followed the summit closely, and what struck us is how many of the issues discussed reflected exactly the world we are trying to navigate.
The four lessons from Berlin
For headline readers, here is the short version. Each point is unpacked further down.
- The integration layer is everything - the winners are not the ones with the best models, but the ones with the best plumbing.
- The maturity gap is real and growing - pilots are being launched faster than the foundations they depend on are being built.
- Governance is a delivery requirement - responsible AI is a precondition for durable delivery, not a constraint on it.
- Sovereignty matters beyond Europe - owning and governing your AI is a universal question, not a European one.
1. The integration layer is everything
Leaders from Siemens, IBM, and Deloitte were explicit: the organisations seeing real returns from AI are those that invested in connecting it to real business data, real workflows, and real decision-making processes. Not the organisations with the most sophisticated models, but the ones with the best plumbing.
This resonated. The work we have done rebuilding The Surgical Assistant; a surgical staffing platform, on a modern AI-augmented stack only paid off once it was wired properly into the day-to-day reality of theatre scheduling. The model matters far less than the system it lives inside.
Sophisticated model
No integration → no returns
INTEGRATION LAYER
AI Model · capability
Business Data
real returns
Workflows
real returns
Decisions
real returns
2. The maturity gap is real and growing
Gartner predicts that over 40% of agentic AI projects will be cancelled by 2027 and finds that 85% of AI projects fail due to poor data quality. Organisations are launching pilots faster than they are building the foundations those pilots require. The reasons AI projects fail are almost never technical. They are organisational, architectural, and cultural.
We see this routinely. The discovery conversations that go well are the ones where the client is willing to look honestly at their data, their workflows, and their internal readiness before we touch a model.
40%
of agentic AI projects
cancelled by 2027
SOURCE: GARTNER
85%
of AI projects fail
due to poor data quality
SOURCE: GARTNER
3. Governance is a delivery requirement
With the EU AI Act's high-risk classifications rolling out progressively, and voices from the European Parliament and the United Nations in the room, the summit was clear: responsible AI is not a constraint on progress. It is a precondition for it. Governance must be by design, not added afterwards.
In financial services and healthcare work we have been part of, data ownership, access controls, and audit trails are decisions made on day one. Not because a regulator is watching, but because anything else creates rework and risk later.
Oliver Brouckaert unpacked this in more depth in Data Security Architecture: Why Protecting Your Data Is Not Just Compliance - governance and data security are architectural decisions, not compliance checkboxes.
THE WRONG MODEL
Compliance checkbox
Treated as an audit task, retrofitted onto systems that were never built to support it.
THE RIGHT MODEL
Governance by design
Built in from day one
- ✓Data ownership
- ✓Access controls
- ✓Audit trails
- ✓Responsible AI
4. Sovereignty matters beyond Europe
The question of building AI that organisations own, govern, and can trust is not a European regulatory concern - it is a universal one. For South African and UK businesses building for international markets, understanding where AI governance is heading is not optional.
When we have built AI-driven systems for clients like TTT Financial Group, governance has been a core design requirement from the start, not a compliance afterthought.
Owned. Governed. Trusted AI.
A universal requirement, not a European one.
EU
AI Act
UK
AI regulation forming
South Africa
POPIA + AI policy
The era of AI experimentation is ending. The era of AI delivery is here. That is what Berlin said in May 2026.
The case for patience - and why we disagree with it
There is a reasonable argument for slowing down. The regulatory environment is still forming. Integration patterns are still maturing. Many organisations have already burned significant budget on AI pilots that delivered nothing. Why not wait for the landscape to settle?
We understand that position. The caution is not irrational. But the history of technology does not reward it.
The organisations that waited for cloud infrastructure to stabilise before migrating found themselves years behind competitors who moved early and learned as they went. The same pattern played out in mobile, in e-commerce, and in modern application platforms. In each case, the penalty for waiting was not just being slower - it was watching others develop capabilities, relationships, and institutional knowledge that cannot be replicated from a standing start.
AI will follow the same curve. The goal is not to move recklessly. It is to move with intention, with the right architecture, and with partners who have already done this.
The risk is not adoption. The risk is adoption without foundation.
Where this leaves your business
The market is separating. Not dramatically, not overnight, but steadily and in one direction. On one side are the businesses that have started building: clean data, integrated workflows, AI embedded in delivery, capable partners alongside them. On the other are the businesses still treating AI as a future project, something to revisit once conditions are clearer.
The gap between those two groups is not yet insurmountable. But it is growing every quarter. AI capabilities compound on top of strong foundations. Organisations that have spent the last year building those foundations can now move at a speed that businesses starting from scratch will struggle to match.
The question is not whether to engage with AI. That question has been answered. The question is whether your business is building the foundation that will make your AI investment deliver.
How riivo fits in
When it comes to the implementation of AI powered software, riivo specialises in bridging that gap between ambition and delivery. It is a gap we have been navigating internally and with clients long before it became the dominant conversation in international AI circles.
The hardest part of most AI conversations is not the technology. It is knowing where to start. What to build first. What your current systems can support. What the right investment looks like for a business of your size and complexity.
Our discovery process is designed to solve exactly that.
References
- Rise of AI. Rise of AI Conference 2026. 2026. https://riseof.ai/conference-2026/
- Gartner. Predicts 2025: AI Agents and Advanced Automation. 2025. https://www.gartner.com
- Deloitte. State of Generative AI in the Enterprise. 2025.
- riivo. The Velocity of Certainty. 2025. /blog/the-velocity-of-certainty