Fixed Price, Time and Materials or Proven IP: How to Buy IoT Engineering After AI

Why paying for hours fits connected products less every year

Most engineering firms still sell time. The more hours a project takes, the more the vendor earns, and the buyer pays for effort rather than for a working product. That arrangement was always a compromise. AI coding tools now make it harder to defend: a prototype that once took a team months can be stood up in days, and buyers have started to ask the obvious question. If your engineers use AI, why am I paying for the same hours?

I discussed this with Taras Vasylyshyn on his podcast in October 2026 (the conversation is in Ukrainian). My short answer was that the hourly model is living out its last years. If all a vendor sold you was hours, a request for a discount is fair and cannot be stopped. If the vendor sold you something that is hard to reproduce, such as tested electronics, supplier relationships and the scars of earlier projects, the conversation is different.

Watch the full interview on YouTube (in Ukrainian with English auto-dubbing, 1 hour 17 minutes). If the English audio does not start on its own, choose it under Settings, then Audio track.

Large firms talk about outcome-based contracts, where the vendor takes a share of the revenue a feature brings. That can work between two enterprises with research teams on both sides. For most connected products it is hard to agree what the vendor's work actually caused, so you need a model you can measure today.

What AI tools still cannot see in a hardware product

Code is the cheapest part of a connected product to change. The expensive mistakes happen where software meets copper, components and factories, and that is exactly where an AI assistant has no feedback until the parts arrive.

For a small side project I wanted to stream encrypted video from my own drone. An AI assistant assured me the setup would work and the stream would be encrypted, so I ordered the parts. When they arrived, the module did not support RTMPS at all, so there was no encrypted channel. An engineer who had built that kind of link before would have caught it before the order.

The same pattern repeats on real products:

• Boards. A trace that is too thin burns on the first power-up, and the fix is a respin with a new lead time, not a commit.
• Components. Two parts with the same datasheet values can behave differently, and one of them may be hard to buy next year.
• Factories. Knowing which manufacturer delivers on time, and where a factory is likely to let you down, comes only from having shipped with them.
• Materials. Surface finish, copper oxidation and assembly tolerances decide whether a design survives production, not just the lab.
• Regulated sectors. Defence, healthcare and safety products must follow required architectures and certification rules that a general-purpose model does not enforce.

AI helps an experienced engineer move faster and check more options. It does not replace the experience that tells you which option will fail on the bench.

Fixed price vs time and materials vs proven IP compared

Each model puts the risk somewhere different. Pick the one that matches what you actually know about your product today.

Fixed price looks safest, but on hardware it often turns into a negotiation over change requests, because the first prototype usually teaches you something the specification missed. Time and materials is honest about uncertainty, but it rewards the vendor for taking longer. Adapting proven IP removes the riskiest part of the journey, the first board that has never been built before. Many projects mix them: a time-and-materials discovery, then a fixed price for each well-defined phase.

What adapting proven IP looks like in practice

At Indeema we started selling this way in 2026. When a client needs electronics close to something our hardware engineering team has already designed, built and tested, we show them our own design and offer a choice: adapt it, or build from zero. If they choose to adapt, we sell two things, the engineering hours to modify the design and the rights to use the modified IP. We signed three contracts on this model in 2026, and we plan to move more of our work to it.

For the client, the gain is risk they no longer carry. The base design has already been through the thin-trace, wrong-supplier and failed-test stage, so the timeline is easier to predict and only the modification is new. For us, every adaptation adds to what we know for the next client, so the model gets stronger with use instead of resetting with every project.

It is not the right answer for every product. If your differentiation is the electronics themselves, or you need full exclusive ownership of every line of the design, a clean-sheet build may still be the better choice. The point is to make that a deliberate decision, not the default.

What to agree about ownership before you sign

Licensed or adapted IP changes the contract more than the engineering. Before you sign, get clear written answers to these points:

1. Rights. Do you own the modified design, or hold a right to use it? For which products, markets and period?
2. Exclusivity. Can the vendor reuse the same base design for others, including in your market?
3. Your changes. Who owns the modifications made for you, and the improvements you pay for later?
4. Deliverables. Will you receive schematics, layout files, the bill of materials, firmware source and manufacturing documents, or only finished units?
5. Continuity. What happens if you bring development in-house, or if the vendor stops supporting the design?

None of these questions is unusual. A vendor that has sold its own IP before will have answers ready; one that hesitates may be selling hours under a new name.

A decision rule for your next connected product

Before you ask for quotes, answer these questions with your own team:

If the scope is still open, start with a short paid discovery on time and materials. If it is stable and mostly software, a fixed price is reasonable. If a vendor can show you tested electronics close to your need, ask what adapting them would cost and what you would own. Either way, choose a partner by what they can prove has worked, not by how many hours they can supply.