Sequoia Capital recently published a remarkable thesis: the next trillion-dollar company will be a software company posing as a services company. Not a better tool. Not a smarter copilot. A company that simply does the work.
That thesis was written from Silicon Valley. But the implications may be even greater for the Netherlands, an economy where 77% of GDP comes from services and where labor shortages are structural across virtually every professional sector.
At Laava, we see this every day. We build the AI systems that actually perform work inside our clients' systems. Not chatbots in a sandbox, but agents that process invoices, perform KYC checks, and generate reports. This article translates the Sequoia thesis to the Dutch market and shows why the opportunities here are particularly large.
From tool to work: the core argument
Every AI founder asks themselves the same question: what if the next version of the model makes my product obsolete? If you sell the tool, you're in a race against the model. But if you sell the work, every model improvement makes your service faster, cheaper, and harder to compete with.
The difference lies in what Sequoia calls "intelligence" versus "judgement." Intelligence is the application of complex but known rules: writing code, processing invoices, drafting contracts. Judgement requires experience and intuition. Deciding which product to build, when a risk is acceptable, how to advise a client.
AI has reached the threshold where most intelligence work can be performed autonomously. Software engineering crossed that threshold first. At Anthropic, more tasks are now initiated by agents than by humans. But it's coming for every sector.
Copilots and autopilots
A copilot sells the tool. An autopilot sells the work.
Until recently, AI models weren't intelligent enough to operate independently, so the logical approach was: build a copilot, put AI in the hands of a professional, and let them decide what to do with it. Think of Harvey for law firms or Rogo for investment banks. The professional is the customer, the tool makes them more productive.
Today, models are intelligent enough that in some categories, the best starting position is an autopilot. The customer doesn't buy the tool, they buy the outcome. And the budget for work in any sector is many times larger than the budget for tools.
The higher the share of intelligence work in a field, the faster autopilots will win.
Why the Netherlands is particularly interesting
The Sequoia thesis is universal, but three structural characteristics make the Dutch market particularly fertile ground for AI autopilots.
1. A services economy par excellence
The Netherlands is a services economy in the most literal sense. According to CBS, 77% of total economic output is services. Commercial services alone represent 54% of added value. Four out of every five hours worked go to services.
For comparison: for every euro spent on software, six euros go to services. In the Netherlands, that ratio is likely even more skewed, given the size of our consulting, financial, and business services sector. The IT services market alone accounts for over €19 billion, management consulting for more than €5 billion, and business process outsourcing for nearly €8 billion. And those are just three segments.
2. A structural labor shortage
The Dutch labor shortage is not cyclical, it's structural. Until mid-2025, there were more vacancies than unemployed people, peaking at 97 vacancies per 100 unemployed. While the labor market has cooled slightly since then, structural shortages in professional services remain significant. The shortages are sharpest in precisely the sectors where autopilots can have the greatest impact:
- Accountancy is facing a global crisis. The Netherlands is losing accountants while demand grows. The RA qualification takes years, starting salaries lag behind tech and finance, and 75% of CPAs are approaching retirement. PwC has openly acknowledged that AI will take over part of the work.
- IT and business services: more than half of companies in these sectors regularly face staffing shortages.
- Legal services: the market is candidate-driven. Corporate Counsel and Compliance Managers are scarce.
This labor shortage is not a problem that can be solved with more recruitment. It's a structural argument for automation.
3. High AI adoption, low AI impact
Here's the paradox. The Netherlands leads in AI adoption: 95% of organizations use AI, the adoption rate of 49% is well above the European average of 42%, and every four minutes a new Dutch company starts using AI.
But only 5% actually derive value from it.
That gap, 95% adoption versus 5% impact, is precisely the gap between copilots and autopilots. Most Dutch companies have implemented AI as a tool: a chatbot here, an assistant there. But the work doesn't fundamentally change. The professional still does the work, just now with an AI tool alongside them.
Autopilots close that gap. They don't deliver the tool, they deliver the outcome.
The Dutch opportunity map
If we translate the Sequoia analysis to Dutch sectors, which are most ripe for AI autopilots? We look for three signals: a high share of intelligence work, existing outsourcing, and a structural labor shortage.
Accountancy and audit
The Dutch accountancy sector is perhaps the most obvious target. The labor shortage is acute, the profession is aging, and the lion's share of the work (processing entries, performing audits, preparing tax returns) is intelligence work. Complex rules, but rules nonetheless.
Globally, the Big Four have already invested heavily in AI. Deloitte launched Zora AI for invoice processing and trend analysis. PwC invested more than $1 billion in AI capabilities. KPMG uses AI to scan millions of journal entry lines. A former PwC board member predicts that 50% of roles in audit, tax, and advisory will be automated within 3 to 5 years.
But here's the opportunity for autopilots: the Big Four sell tools to their own professionals. Nobody yet sells the outcome, simply closing the books, directly to the CFO of a mid-sized company. That's exactly the shift Sequoia describes.
At Laava, we've already put this into practice. An accounting firm with 50+ SME clients implemented our AI Agents for automated invoice processing and data extraction. The result: a 71% reduction in administrative processing time, with significantly higher accuracy. That's not a copilot. That's an autopilot doing the work.
Compliance and KYC
Know Your Customer processes are almost purely intelligence work: searching public sources, checking sanctions lists, building risk profiles. The rules are complex, but they are rules. In the Netherlands, this is extra relevant given the strict supervisory environment of DNB and AFM.
Dutch fintechs like Duna (€40.7 million in funding) and Fourthline are building automated KYC platforms. But most compliance departments still work with manual processes or expensive external consultants.
Laava's Smart KYC Analyst is an example of the autopilot approach: an autonomous agent that collects data from all agreed-upon sources, interprets it according to a risk framework, and produces a structured risk profile. The human analyst retains final responsibility but reviews one summary instead of dozens of raw sources.
Insurance
The Dutch insurance market is highly fragmented, with thousands of small brokers all following the same process. The broker's work in standard commercial insurance is largely intelligence: comparing quotes, filling out forms, administering policies.
Fewer than 25% of insurers have achieved extensive digitalization. The rest struggle with legacy systems and fragmented data. That's not a technology problem. It's a market opening for autopilots that can take over the entire placement process.
IT managed services
Every SME outsources its IT. Patching, monitoring, user provisioning, alert triage: intelligence work that repeats across thousands of identical environments. The current software layer (think ConnectWise, Datto) sells tools to the MSP. Nobody yet sells "your IT just runs" directly to the business as an outcome.
Recruitment and HR
The staffing market in the Netherlands is one of the largest services sectors. The top of the funnel (screening, matching, outreach) is pure intelligence work. An SME services company with 120 employees implemented Laava's AI Agents for screening applications, scheduling interviews, and guiding onboarding, integrated with existing HR systems. AI now handles 80% of routine tasks and saves recruiters 15 hours per week.
The outsourcing wedge
Sequoia's playbook for autopilots is elegant in its simplicity: start where outsourcing already exists.
If a task is already outsourced, you know three things. One: the company accepts that this work can be done externally. Two: there's an existing budget line that can be cleanly replaced. Three: the buyer is already purchasing an outcome. Replacing an outsourcing contract with an AI-native service provider is a vendor switch. Replacing staff is a reorganization.
This is particularly relevant for the Netherlands, where the BPO market alone is worth nearly €8 billion. Many SMEs already outsource administration, IT management, and customer service. Those outsourced, intelligence-heavy tasks are the perfect wedge for autopilots.
Start with the outsourced, intelligence-heavy work. Nail the distribution. Expand into the insourced, judgement-heavy work as AI evolves.
From adoption to impact: where Laava fits in
Most AI implementations in the Netherlands don't fail because of bad technology. They fail because AI is implemented as a standalone tool instead of as an integrated part of the work process.
That's exactly why Laava focuses on deep integration. We don't build chatbots. We build production-grade AI systems that integrate directly with our clients' existing stack: ERP, CRM, ITSM, warehouse systems. Our agents perform actions in your systems, not just suggestions. With approval gates where needed, full audit trails, and the ability to roll back if something goes wrong.
That's the difference between a copilot sitting next to you and an autopilot doing the work.
The three layers of our approach align seamlessly with the autopilot thesis:
- Data layer: making data discoverable, reliable, and governed so AI can use it safely. Connecting knowledge across documents, tickets, code, and databases.
- Agent layer: task-specific agents that coordinate with systems. Not a generic chatbot, but agents with bounded responsibilities, explicit tools, and guardrails.
- Integration layer: embedding LLMs and agents into existing tools and APIs. Not a new system on the side, but intelligence at the heart of the existing workflow.
The convergence
Today, there's a clear distinction between intelligence and judgement. But that distinction is shifting. As AI systems collect proprietary data about what constitutes good judgement in their domain, the boundary moves. Copilots and autopilots will converge.
But the starting position matters. Companies that begin as autopilots build the data from day one that will eventually enable them to handle judgement work too. Companies that start as copilots will need to make a difficult transition later, cutting their own customers out of the work in the process. That's the innovator's dilemma in action.
For the Dutch market, this means: the companies building AI agents now that actually perform work, not just advise, are building a lead that becomes increasingly difficult to close.
What this means for Dutch businesses
If you run a business in the Netherlands, there are three concrete takeaways:
Look at your outsourcing budget. Every euro you spend on external services for intelligence-heavy work (administration, compliance, IT management, first-line customer service) is a candidate for an AI autopilot. Not in five years. Now.
You won't solve the labor shortage with more recruitment. Demographics are working against you. The solution isn't finding more people for the same work. It's needing fewer people because AI takes over the intelligence work. The professionals you have can focus on the judgement work that adds real value.
Adoption without integration is money wasted. The Netherlands scores 95% AI adoption but 5% value realization. The difference isn't the tool, it's the integration. AI that sits alongside your processes delivers little. AI that lives inside your processes delivers everything.