What can AI do in work preparation for construction and installation?

What does a work planner do?

A work planner makes sure a project is ready to be carried out. The assignment is translated into what is needed on the building site: people, materials, drawings and a schedule. That applies to a contractor just as much as to an installation company.

In construction and building services, it usually comes down to these tasks:

• Reading the specification and drawings. What exactly has been asked for? Which materials, which requirements and which agreements?
• Estimating. Estimating quantities, materials and hours, often together with the cost estimator.
• Purchasing. Ordering materials from the wholesaler or supplier, at the right moment.
• Scheduling. Coordinating crews, fitters, equipment and subcontractors.
• Preparing the execution. Working drawings, job sheets and instructions for the building site.
• Keeping the project file up to date. Changes, extra work, photos and the documents for the handover.

Much of that work is reading, comparing and retyping. Laying a specification next to a drawing. Copying a materials list into an order. Passing a change on to three parties. AI can help with that. With calculating, estimating and deciding, much less so. Examples for the whole of construction, from enquiry to handover, are on aiopdebouw.nl (in Dutch). This article is only about work preparation.

What can AI already do in work preparation?

At the moment, AI mainly helps with work involving text and documents. This is generative AI: software that reads and writes text. These are the tasks in work preparation where it fits best:

• Summarising a specification. AI lists the requirements, materials, deadlines and conditions. It also lists what differs from how you normally work, and which questions are still open.
• Asking questions of your own documents. For example: "Who supplies the lighting?" You get an answer with the place in the specification where it says so. This is called an AI assistant on your own documents.
• Preparing orders. From your estimate and your schedule, AI makes a draft order per supplier or per phase. You check it and send it.
• Comparing delivery notes and invoices with your order. AI reads a delivery note or purchase invoice, including a photo of a paper one. The software shows what differs. You can read more about this in reading invoices automatically.
• Processing job sheets and snagging points. A fitter records a voice message about what he did or what is still outstanding. AI turns it into a job sheet or a list of snagging points. See the digital job sheet.
• Drafting messages. If the schedule shifts, AI prepares a message for the subcontractor or supplier. You send it.

These tasks have three things in common. They come up often, they consist mainly of reading and typing, and a person can check the result well. So automating work preparation does not mean here that software takes over the work. It means the work planner has to search and retype less.

What does that look like in a project?

This is how it can look, from assignment to handover. It is a made-up example of an installation company that receives a new assignment. Note the right-hand column: it shows what the work planner keeps doing.

You do not have to do all the steps at once. Start with the step that costs you the most reading or typing. For example the specification, or retyping job sheets and delivery notes.

Can AI read a specification? And a drawing?

AI is good at reading a specification. Reading a drawing only partly works. That is because a specification is mostly text, and a language model is strong at text. Such a language model is also called an LLM.

With a specification, AI can summarise, search and answer questions. You can also put two versions side by side and ask what has changed.

Drawings are a different matter. Some AI models can look at a drawing and tell you what is on it. But copying measurements, numbers and symbols exactly is not yet reliable enough to use without checking. A misread measurement does not stand out in a summary. In your order, it does.

That is why the rule is: quantities come from your drawing software, your BIM model or your own measurement, not from an AI answer. AI may help to search and check. For example by flagging that a part is in the specification but not in your materials list. You can read the same principle in making quotes with AI: the prices come from your own price list, not from the language model.

Where does AI not work well yet?

AI does not work well for work where precision, expertise or responsibility is at the core. Specifically:

• Measurements and quantities. See above: let them come from your own software or measurement.
• Calculating. A language model is not a calculator. Let ordinary software with fixed rules do the calculations.
• Estimating. How difficult is the job, how accessible is the location, which crew suits it? A tradesperson knows that better than software.
• Explaining rules and standards. AI can summarise a regulation, but it can be wrong about what applies to your project. If in doubt, read the source yourself.
• Deciding. What you order, who you schedule and what you agree with the customer is for you to decide.

AI can also make up something that looks convincing. This is called a hallucination. Makers of AI software, such as Microsoft, also write that answers from generative AI can contain errors and that you should assess them before sending them to others. In work preparation that means: AI does the groundwork, the work planner checks and decides. This way of working is called human-in-the-loop.

There is one more limit. If your project documents are messy, the answer will be too. If specifications, drawings and changes are spread across email, folders and phones, AI will not find them either. So the first step is often: all of a project's documents in one place.

How do you connect AI to your cost estimating, scheduling and administration?

Through integrations between the software you already use. AI works best when it does not stand on its own, but can fetch data from your systems and put a draft back. A good setup has three parts:

1. AI reads and suggests. It reads a specification, an email or a delivery note and makes a draft.
2. Ordinary software calculates and saves. With fixed rules, in your own cost estimating, scheduling or administration system.
3. A person approves. Only then does an order go out or are the hours entered in the administration.

Such an integration usually works through an API: a fixed entry point of a software package. Many business packages have one. Whether that also applies to your cost estimating and scheduling software differs per package. If there is no API, an export and import is sometimes enough. You can read how an integration works in what is an API integration.

In construction, parties also work together on agreements about digital information. The digiGO network promotes the use of (open) information standards throughout the chain, such as NL-SfB and the BIM basis ILS. If you already work with such agreements, your information is easier to search and connect. You do not need them to get started with AI. You can read more about connecting systems on integrations and automation.

What does the Wkb mean for your project file, and does AI help with it?

Under the Wkb (Wet kwaliteitsborging voor het bouwen, the Dutch Quality Assurance for Building Act), the contractor gives the client a file at handover. AI can help to keep that file up to date during the project.

The Wkb has applied since 1 January 2024. According to the Dutch government (Rijksoverheid), the new system of supervision initially applies only to buildings in the lowest consequence class, such as homes and simple commercial buildings. An independent quality assurer then checks whether the building meets the statutory technical requirements.

The Wkb also amended the Dutch Civil Code (Burgerlijk Wetboek). For work preparation, two rules from it matter most:

• A file at handover. In a building contract, the contractor gives the client a file when the work is ready to be handed over. It contains, among other things, drawings and calculations, a description of the materials and installations used, and the information for use and maintenance (Article 7:757a of the Civil Code). This applies to building contracts concluded from 1 January 2024.
• Liability after handover. The contractor is liable for defects that were not discovered at handover, unless those defects cannot be attributed to the contractor (Article 7:758(4) of the Civil Code).

For work preparation, that means: what you record during the project, you will need at handover. So have photos, job sheets, changes and product sheets go straight to the right project. At the end, AI turns them into a draft file and flags which documents are still missing. A person checks that the file is correct and complete before it goes to the client.

This is an explanation in plain language, not legal advice. Which rules apply to your project depends on the type of building and your contract. If in doubt, check the Dutch government's website or ask your trade association.

What agreements do you make about AI and project data?

Make agreements before your team uses AI for projects. In work preparation, you often work with other people's documents: specifications and drawings from the client, and names and addresses of residents. Pay attention to these points:

• An AI policy. Record which AI tools are allowed, which documents may go into them and who checks the result. There is an example in an AI policy on one page.
• Agreements with your client. Does your contract say that project documents are confidential? Then check whether you may put them into an AI tool, and which one.
• Personal data. Names, addresses and phone numbers of residents are personal data. The GDPR applies to them. The Dutch Data Protection Authority (Autoriteit Persoonsgegevens) recommends first checking whether you can use generative AI without personal data. If that is not possible, take privacy into account from the start.
• A business account. Use a business version of an AI tool with a data processing agreement, not a free account. Also read whether ChatGPT is safe for businesses.
• Explanation for your team. Make sure your team knows what AI can and cannot do. This is called AI literacy. You can read what the EU AI Act asks of businesses on this in AI literacy is mandatory.

How do you start small with AI in work preparation?

Start with one task that comes up often, on one project. That way you learn what AI can do, without much at stake.

1. Choose one task. For example: summarising every new specification into a list of points of attention and open questions.
2. Put the documents in one place. Specification, drawings, changes and emails of one project in one folder.
3. Try it on a completed project. There you already know the answers. That way you see straight away where AI does well and where it gets things wrong.
4. Agree who checks. And what happens if an answer is wrong.
5. Only connect it afterwards. If it works, you connect it to your cost estimating, scheduling or administration. Then you have to retype less.

Not sure yet which task should come first? Then read you don't know where to start with AI. Do you also want to tackle the scheduling of crews and fitters? Then read automating scheduling with AI.

How do I help you with AI in work preparation?

I first look at how work preparation currently goes at your company: which documents come in, which software you use and where the retyping is. Then you get a proposal with a fixed price. Then I build it myself, connected to the systems you already have. AI does the groundwork, you decide.

An example of that principle: for De Rijplaten Specialist and De Slowfeeder Specialist, two online shops from Ede, I made an automatic quote generator. AI recognises an enquiry in the email, puts the details into the system and suggests a price. After a click on approve, the quote goes to the customer. They are not construction companies, but the setup is the same as with a specification or an order: AI reads and suggests, a person approves. View the case.

More examples for construction, with videos, are on AI op de bouw (in Dutch). Also read how your price is determined. Or schedule a conversation: the first conversation is free.

Frequently asked questions

Sources: Rijksoverheid (Dutch government): Supervision in construction through the Quality Assurance for Building Act (Wkb) (start on 1 January 2024, lowest consequence class, quality assurer), IPLO: Consumer file (handover file) and the Quality Assurance for Building Act (file for contracts from 1 January 2024), Dutch Civil Code Book 7, Articles 7:757a and 7:758 (file and liability after handover),Dutch Data Protection Authority: Guidance on generative AI and the GDPR (July 2026), Dutch Data Protection Authority: Data processing agreement, Microsoft Learn: Data, privacy and security for Microsoft Copilot (assessing answers from generative AI yourself) and digiGO: What is digiGO (information standards in construction). Consulted on 22 October 2026.