Custom AI solutions for your business

Who builds custom AI solutions, and what's the independent consultant's role?

Dave Willemstein doesn't build or deliver AI solutions, but helps decide which solution makes sense, guides the choice of a solution or build partner, and tests from the client's side whether the result works. Anyone who wants to know where the organisation stands first gets a reasoned picture from the AI Readiness Assessment in 2-3 weeks, for €7,000. The build sits with the client's own team or with a partner the client contracts directly.

Is custom development always necessary for AI solutions?

No. For AI solutions, a well-configured standard solution is often the fastest and cheapest route. Under the rules of thumb, custom development fits when the solution touches on why customers choose the organisation, when the data or way of working is distinctive, or when no standard package gives the control that's needed.

Can the AI Readiness Assessment also be run for a single department?

Yes. The AI Readiness Assessment can be run smaller, for a single department or process area, at a lower investment.

Is there a margin on developers suggested for AI solutions?

No. Where it helps, the consultant suggests developers for AI solutions that he knows, with no margin on their rate. The builders come from the client's own organisation or from a party the client chooses themselves.

Source: rates and typical timelines from Dave Willemstein. Updated: October 2026.

What are AI solutions?

An AI solution is an application in which artificial intelligence solves a concrete problem in your work: pulling information out of documents, answering questions based on internal knowledge, sorting requests, or making predictions. The word “solution” matters here. An AI solution isn’t just the software; it’s the whole of the model, the data, the integrations, the working process, and the people who use it.

That distinction explains why many AI projects succeed technically and still deliver little in practice. The model works, but nobody has thought through who checks the outcome, what happens when it’s unsure, or how it fits into the daily routine.

I’m an independent AI consultant and I don’t build AI solutions myself. I help you decide which solution makes sense, help choose a solution or build partner with you, and test from your side whether the result works. The building itself is done by your own team or a partner you contract directly.

Examples of AI solutions

Most AI solutions that are actually used inside organisations fall into a handful of categories:

• Document processing. Pulling data from invoices, contracts, applications or case files and preparing it for review.
• Knowledge assistants. Staff find answers in policies, handbooks and past cases without searching themselves.
• Classifying and routing. Automatically sorting incoming emails, tickets or reports and routing them to the right person.
• Summarising and drafting. Preparing reports, quotes or write-ups based on existing information.
• Forecasting. Estimating demand, churn or risk based on historical data.
• AI agents. Systems that carry out steps in your software themselves, within agreed limits.

Two cases show what a solution like this looks like in practice. In KYC document analysis for a financial services firm, the challenge was mostly in the working process: the system proposes, the analyst decides. In the internal AI assistant for knowledge workers, the technology was ready quickly, and most of the work went into adoption and a handful of clear working scenarios. More recognisable applications by department are covered in a separate knowledge-base article on AI examples for business.

Custom AI solutions versus standard

Not every question calls for custom AI solutions. Often a well-configured standard solution is the fastest and cheapest route. A few rules of thumb:

• Choose standard when several vendors offer the same thing, the task isn’t unique to your organisation, and you mainly want value fast.
• Choose custom when the solution touches on why customers choose you, your data or way of working is distinctive, or no standard package gives you the control you need.
• Choose a combination in most other cases: standard building blocks, with a custom-built layer for your rules, data and integrations.

How you assess standard software and what integration involves is covered on the page about choosing, integrating or building AI software.

From idea to go-live

An AI solution moves through five steps from idea to daily use. At each step, it’s clear who does the work.

1. 1. Choose the bottleneckWhich problem, for whom, and how do we measure success? You and me together
2. 2. Decide the routeStandard, custom or a combination; set requirements for data, privacy and the AI Act
3. 3. Choose a solution or build partnerCriteria upfront, tested on your own data; you contract directly
4. 4. Pilot on real dataBuild partner or your own team builds; I check against the agreed measures
5. 5. Go-live and supportWorking process, adoption and ownership sit in your organisation; measure, adjust or stop

What you can expect from me over that project are concrete outputs: an assessed list of use cases, a specification with success measures, a reasoned selection recommendation, and testing of what’s delivered. The builders come from your own organisation or from a party you choose yourself; where it helps, I suggest developers I know, without any margin on their rate.

The step after the pilot, from a working prototype to use across the whole organisation, is often the hardest. That’s about working processes, adoption and governance; more on that under AI implementation. A broader overview of the development journey is on the AI development page.

AI solutions for small and medium-sized businesses

The same principles apply for smaller businesses, but with less room for experiments. A failed project weighs more heavily in a smaller organisation. Three pieces of advice:

• Start with the software you already have. Many packages include AI features that, with the right setup, cover a good part of the need without custom work.
• Pick one bottleneck with a clear owner. One process done well beats five done half well.
• Only have something custom built once standard demonstrably falls short. Custom work brings ongoing support costs, and you need to be able to carry those.

More on AI for smaller organisations is on the AI for small business page. Want to know where you stand first? The AI Readiness Assessment gives you a reasoned picture in 2-3 weeks, for €7,000. For a single department or process area, it can also be run smaller, at a lower investment.

Looking mainly for someone with technical depth? A separate article on this site sets out the different AI-development profiles and which one fits your question. And if you’d like to talk through a concrete bottleneck, book a strategy call.