An AI strategy for management starts not with tooling but with business value: decide where AI strengthens your competitive position, choose build, buy or partner per use case, and steer on board KPIs. MIT shows that 95% of GenAI pilots deliver no measurable profit-and-loss impact. The leadership that does see returns builds a portfolio of pilots with a roadmap.
The difference between scattered experiments and competitive advantage comes down to being steered from the top. Research by MIT shows that 95% of GenAI pilots achieve no measurable impact on the profit-and-loss statement. Not because of poor tools, but because of a missing strategy. The leadership that gets returns does three things differently: it ties every AI investment to a measurable business goal, it builds a portfolio of pilots instead of one hopeful experiment, and it keeps control modular so no vendor holds the organisation hostage. At Movir, a disability insurer within Nationale-Nederlanden with 350 to 400 employees, the journey began in the management team. Broad training, ten internal champions and department-level implementation followed. Sales processing went from an hour to a minute, sixty times faster, with positive ROI within four months. That is not luck. It is a strategy that starts with value, not with a tool.
Start with business value, not with tooling
Most leadership teams start wrong. They buy a license, roll out a chatbot and hope for productivity. The CBS AI monitor 2024 names the biggest barrier in the Netherlands: 75% of companies lack experience. You don't buy that experience with a subscription. You build it by tying AI to four sources of value: speed, cost reduction, scale and new propositions. At Reditus, a marketing agency with 10+ employees, finding affiliates went from sixteen hours to thirty minutes per client, 36 times faster, a 95% saving. That is not a stray use case. It is a process acceleration visible straight in the margin, with ROI within three months.
Concretely: a leadership team that starts with value first asks which process costs the most money or time. At Reditus that was manually searching for affiliates, sixteen hours of work per client. By automating that process with n8n and Claude, 95% of that time disappeared. The idea went from sketch to live in four months. No upfront tool hunt, but a process that hurt so much the value was measurable straight away.
The question at the board table is not "which AI tool do we buy?" but "where does AI strengthen our competitive position the most?". SPAIK, founded in 2024 in Amsterdam by Jochem van Laren, Thijs Bongertman and Jan Bolle, positions itself as an AI Adoption Partner for established companies from 10 million euro in revenue, focused on back office and marketing. To see the pattern from start to finish, read why 95% of AI implementations deliver no return.
Build, buy or partner: choose per use case
Not every AI application deserves the same approach. Leadership makes a choice per use case between building it yourself, buying it off the shelf or developing it with a partner. Build yourself what touches your competitive advantage and where your data is unique. Buy what is a commodity. Partner where you lack speed and capacity but want to keep ownership.
A useful rule of thumb: only choose build if the application touches your distinctive capability and you hold unique data. A fraud model at Nationale-Nederlanden belongs in build, because the data and risk profile are unique. A standard meeting-notes assistant belongs in buy. And building a department-wide workflow, such as the sales processing at Movir, often belongs in partner: speed and capacity from outside, ownership inside. Mix up these three tracks and you build too much yourself and buy away too much.
Euphoria Mobility, a mobility SaaS with 60+ employees, chose partner-driven building: a hackathon with 25 engineers produced six working prototypes in two days, three of which went into production, with 12 times ROI in one month. The alternative, building everything from scratch, would have taken months. Forrester shows why the partner choice so often goes wrong: 83% of buyers are dissatisfied with their agency on something, often because the pitch team does not become the day-to-day team. To avoid that pitfall, our buyer guide on the best AI implementation agencies for SMEs helps you ask the right questions.
AI governance light for 50 to 500 people
Governance sounds like a brake. For a mid-market company it is in fact the accelerator that keeps experiments in production. You don't need a compliance department of twenty people. You need three things: clear guardrails, an owner per AI application and a fixed feedback loop on quality. At the fraud department of Nationale-Nederlanden, SPAIK trained 60 people in AI fundamentals, then ran a Champions training, and then those champions themselves set the guardrails and requirements for responsible AI use. Governance emerged bottom-up, carried by the people who work with it.
The risk and compliance framework: the EU AI Act
The EU AI Act is no longer optional. Article 4 has required AI literacy since February 2025: employees who use AI must understand what they are doing. That is exactly why training is the foundation of governance, not an afterthought. SPAIK's AI Fundamentals training scores 9.1 out of 10 across more than 700 evaluations, runs for six weeks and delivers an EU AI Act-compliant certification, with 4+ hours saved per participant per week. Here, literacy and return are the same investment. How to approach this across the organisation is covered in the pillar on AI training for teams.
Avoid vendor lock-in: modular and tool-agnostic
The biggest strategic mistake is locking yourself into one vendor. Build everything inside a single closed platform and you pay the price later in negotiating power and migration costs. That is why SPAIK works modular and tool-agnostic: n8n for workflow automation, Supabase as the database, Claude and Claude Code for the language models, and at Movir for example Adobe Journey Optimizer for the department-level implementation. No single layer is irreplaceable. If tomorrow's language model changes, you swap that one layer, not your whole stack.
For leadership, modularity means one concrete requirement: every AI building block must be replaceable without tearing down the rest. That is not a technical luxury but a strategic insurance against price increases and forced migrations. How to get from a successful pilot to a scalable production environment without lock-in is described in from AI pilot to production.
Board KPIs and the role of the director
AI is not an IT project you delegate. The director owns the AI strategy, just as they own the financial strategy. That means KPIs at board level: not "number of prompts used", but hours saved, lead time per process and revenue from new AI propositions. At Movir the figures were clear: sales processing sixty times faster, ten times as much content, ten internal champions. Those are numbers that belong in a board report.
The director's role is threefold: they decide the sources of value, they assign ownership and they protect the budget from fragmentation. SPAIK calls Movir the model: training, capability building and a long-term partnership, the so-called land-and-expand. CEO Maurick Schellekens steered from the management team, not from the sidelines. With 35+ implementations, 700+ people trained and an average satisfaction of 9.1 out of 10, SPAIK keeps seeing the same thing: where leadership owns it, returns appear; where AI stays with the enthusiasts, they don't.
From pilot portfolio to roadmap
A single pilot is a gamble. A portfolio of pilots is a strategy. Leadership picks three to five use cases with measurable value, runs them in parallel and promotes what works onto the roadmap. At Euphoria Mobility that produced six prototypes in two days, three of which made it to production while the other three were killed early and cheaply. That is how you build a substantiated multi-year roadmap instead of a wish list.
The practical entry point depends on your starting position. An Inspiration Session from 3,000 euro gets leadership aligned. AI Fundamentals at 6,500 euro for ten people lays the literacy foundation. The Kickstart at 14,500 euro builds a first working application in four weeks. An ongoing adoption track (indicative price 10,000 to 20,000 euro per month, minimum six months) anchors the roadmap. All of this fits within the broader context of AI implementation in SMEs and the state of AI adoption in the Netherlands.
At SPAIK we back every figure in this article with a client, a source or a date. Read our editorial policy to see how we secure that.
