In preparation for the major AI event taking place in Luxembourg next week, we spoke to the keynote presenter Laurens Vanryckeghem, on what impacts he expects to see caused by AI in the next 20 years. His comments will inspire and challenge many assumed norms in the tech world.

Unmissable reading.

 

AI in 2026 and Beyond

 

There are many conflicting opinions in the market about AI. What 3 impacts do you expect to see this year?

1. The SaaS model starts cracking, and “context” becomes the new competitive advantage.

We’ve been renting software for 20 years. Pay per seat, per month, for access to someone else’s workflow. But here’s what’s changing: AI agents don’t need dashboards. They don’t need user interfaces. They talk directly to APIs, databases, and increasingly to each other. When the interface layer becomes irrelevant, SaaS collapses into what it always was underneath: a database with some functions attached.

Microsoft’s VP of Business Applications recently predicted that traditional business apps will become “the mainframes of the 2030s.” IDC estimates that by 2028, 70% of software vendors will have abandoned pure seat-based pricing. The shift is already underway. So what replaces SaaS as the thing companies compete on? Context. Not raw data, but structured, relational, actionable knowledge about your business, your customers, your industry. A generic AI agent
can draft an email. An agent loaded with deep context about your client history, deal stages, communication preferences, and competitive landscape can draft the right email at the right moment.

That difference is where value lives now.

2. Agents will start talking to each other, and we’ll need entirely new infrastructure to support that.

Google launched its Agent2Agent (A2A) protocol in April 2025 with over 50 partners, including Salesforce, PayPal, and Atlassian. It’s now housed under the Linux Foundation. Anthropic released the Model Context Protocol (MCP) for agent-to-tool communication. Together, these are becoming the HTTP of the agent economy, the foundational protocols that allow autonomous systems to discover each other, negotiate tasks, and collaborate.

This year, we’ll see the first real-world deployments of agent-to-agent workflows in enterprise settings. Procurement agents negotiating with vendor agents. Compliance agents auditing transaction agents. The plumbing is being laid right now.

 

3. The gap between “AI-ready” and “AI-curious” companies will become visible in their results.

Companies that spent the last two years structuring their organizational knowledge, building clean data pipelines, and training their teams to work alongside AI will start pulling ahead in measurable ways.
Companies still debating whether to “implement AI” will realize the gap isn’t about having access to the same models (everyone does), it’s about whether your organization has the context infrastructure for those models to actually deliver value.

 

Many companies have been discussing digitalization for some time. Is the current debate around AI implementation any different from this?

Yes, and the difference is fundamental.

Digitalization was about moving existing processes onto screens. You took a paper form and made it a web form. You took a filing cabinet and made it a shared drive. The workflow stayed the same, it just got a digital wrapper.
AI is about eliminating the workflow entirely.

Think about it this way: digitalization asked “How do we make this process faster?” AI asks “Why does this process exist at all?”

When you digitalized procurement, you built an online portal where people could submit purchase orders, get approvals, and track deliveries. With AI agents, the procurement agent detects you need something based on usage patterns, finds the best vendor by querying other agents directly (using protocols like A2A), negotiates terms, and surfaces a decision for human approval. The process isn’t faster, it’s fundamentally different.

There’s another critical distinction. Digitalization was about tools. AI is about context. A digitalization project asked: “Which software should we buy?” An AI implementation asks: “What does our organization actually know, and is that knowledge structured in a way that intelligent systems can use?”

That’s a much harder, much more strategic question.

The companies that treat AI implementation like another digitalization project, buying tools and bolting them onto existing workflows, will be disappointed. The ones that understand it’s about building a rich context layer that agents can draw from will see transformational results.

 

Some companies are still at the early stages of identifying the right processes to implement AI. In your experience, what are the pain points where AI can create the most value?

I’d reframe the question slightly: it’s not just about which processes to automate. It’s about where contextrich decision-making is currently bottlenecked by human bandwidth.

Cross-system workflows where people are the glue. In most organizations, there are people whose entire job is moving information between systems, translating context from one department to another, making judgment calls that require knowledge spread across multiple tools. Sales ops pulling data from the CRM, cross-referencing it with marketing data, and preparing reports for leadership. That person isn’t doing “data entry,” they’re providing context translation. An AI agent with access to the same underlying data and a well-structured context layer can do this faster, more consistently, and at scale.

Anywhere the organization’s knowledge lives in people’s heads instead of in systems. The senior account manager who “just knows” which clients need a check-in call. The operations lead who can predict supply chain issues because they’ve seen the patterns before. That tacit knowledge is incredibly valuable, and it’s also incredibly fragile. It walks out the door every time someone leaves. AI doesn’t replace those people, but it can capture and scale their context so the organization isn’t dependent on individual memory.

Vendor and partner interactions that follow predictable patterns. RFPs, contract reviews, supplier negotiations, compliance checks. These are high-value but highly repetitive interactions where both sides follow established patterns. This is exactly where agent-to-agent communication will create enormous efficiency gains. Your procurement agent talks to your supplier’s sales agent. Both carry context about historical terms, pricing, and compliance requirements. The negotiation happens in minutes instead of weeks.

The common thread: the biggest value isn’t in automating simple tasks (though that’s useful). It’s in making your organization’s collective context accessible and actionable in ways that weren’t possible when it was locked in individual brains, email threads, and siloed tools.

 

There seem to be many new AI startups in the market. Should companies buy or build their own solutions?

The honest answer is: the question itself is becoming outdated. The real question is “What context do you need to own, and what can you rent?”

Here’s why. The cost of building software has dropped dramatically. AI coding agents can now replicate the functionality of many simple SaaS tools in hours. The build-vs-buy math that kept companies subscribing to dozens of tools is shifting fast. You’re going to see organizations, especially those with some technical capability, start migrating parts of their SaaS stack to internally built, agent-powered alternatives. Not because it’s trendy, but because it’s cheaper and more tailored.

What to build (or own): Anything that touches your core competitive context. Your customer intelligence layer. Your domain-specific knowledge base. Your proprietary workflows. These are your moat. If you outsource this to a vendor, you’re handing them your most valuable asset, and you’re one price increase or API deprecation away from a crisis.

What to buy: Infrastructure that requires extreme reliability and scale, things like payment processing, security, high-availability databases. You’re not going to replace Stripe with an agent. Also, genuinely specialized tools where the vendor has built deep domain context you can’t replicate (yet). What to watch out for with AI startups: Many are what I’d call “wrapper companies,” a thin layer of prompt engineering on top of a foundation model. These are the most vulnerable businesses in the market right now. The moment the underlying model improves (which happens every few months), their value proposition evaporates. The AI startups worth partnering with are those building proprietary context layers, unique datasets, domain-specific training, or critical infrastructure for the agent economy, things like agent authentication, context structuring, or orchestration platforms.

The biggest risk isn’t buying vs. building. It’s failing to own your context. Whether you buy or build, make sure your organization’s knowledge, relationships, and decision-making intelligence stays yours. 

That’s the asset that appreciates. Everything else is a commodity.

Closing thought

We spent 20 years paying to rent software. The next 20 will be about who owns the richest context. Models are commoditizing. Interfaces are disappearing. What remains is what your organization uniquely knows, and whether that knowledge is structured in a way that intelligent systems can actually use. Context is the new currency

 

If you want to joint us on the event: “Not just talk. Real-world AI business cases.”, register here !