Connext: a year in indoor radio coverage

In this story: How I ended up here · What it does · The work · Lessons learned

How I ended up here

As an IoT software engineer I knew software development, protocols and hardware inside out. Radio, and how radio waves propagate, was a black hole. For antenna designs, link budget calculations or where to put the LoRa gateways, I had to trust other engineers. It is a science of its own.

So while I was CTO of two IoT companies, I joined the local radio club, to get at least some pointers on how these systems work in depth. The radio amateur virus stuck, and I am still a member.

Through my network I was then referred to Connext. I liked how the founders saw a collaboration with me. They had an open mindset and were open to discussing any matter, no matter how hard. I agreed to the adventure because, with their presence and leadership, I was sure that if business got harder, we would find a way out.

What it does

Connext installs distributed antenna systems: it brings radio coverage inside buildings. Modern construction is very good at keeping radio out: insulation, coated glass, concrete and steel. A distributed antenna system picks up the signal outside, carries it in, and radiates it again where people actually are. That gives the building cellular coverage, or coverage for ASTRID. ASTRID is the Belgian radio network of the emergency services. For many buildings, ASTRID coverage is a requirement rather than a preference.

The technical design of these systems is not that complex. The difficulty is compliance: you must never disturb the other users of the same frequencies. That is what makes indoor coverage a challenging business.

The work

I served one year as project manager at Connext. Most of the job is following up the construction sites where these installations have to go. Getting the cables and the active equipment in place takes a kind of experience that is not my cup of tea. A large part of the telecom industry is, in practice, a construction business.

After a couple of months, I told the founders in one of our one-to-one sessions: “Let’s consider this contracting project management job a mistake, and let’s part ways.” They made an honest effort to keep me for a while longer, and trusted me with a few hard nuts in their business that needed cracking.

Private tenders

Many construction companies issue tenders and ask for frame agreements to get their projects started. Their price requests run to hundreds of cost items. I built simulation tools to make the price lists competitive enough, and to get all those items in line my AI experience came in very handy. The price of a frame agreement came out of a simulation over all those items, not out of a gut feeling about a few of them.

Big customers

Telecom has two kinds of customers: construction companies and industrial sites. Construction companies want the most cost-effective solution. Industrial customers want a system that works and is future-proof.

I never got to manage the construction companies. They work on trust and only open up to the salesperson they trust. Industrial customers, on the other hand, want to see business plans, risk assessments and project plans before they decide. Those relationships are very demanding and need in-depth knowledge of the radio systems installed. I still thank the people at Connext for giving me that chance.

Public tenders

For a public tender, giving your price is not enough. You have to sell your competence and your trustworthiness as a company to even be considered. I wrote several tender applications for different SEVESO sites, the industrial sites that fall under Europe’s rules for major-accident hazards. Most of the people evaluating a tender are not radio engineers. The answer has to make radio compliance readable for them: what will be installed, why it will not disturb the other users of the frequencies, and how that is proven at acceptance.

Price calculation

They asked me to price several projects. As an IT person I did not do that by hand. I built a Claude agent that reads and interprets the request, downloads the building plans and interprets those too. In the end I wrote a program that draws and fills in the antenna plans semi-automatically, so the quote is ready to send.

Lessons learned

Building trust

I arrived at Connext as an IoT and software person in a business of radio engineers and installers. The people in installation companies are different from the R&D people I used to work with. They are no-nonsense, and translate the problem straight into a solution. Nobody there was going to trust my opinion on an installation because of my CV. Credibility had to be earned in their terms.

The first step was doing the homework. I had joined the radio club precisely because radio was my blind spot, and I kept learning until I understood the radio side of the business in depth. That made me a complementary personality to the installers. It was a tough marriage, but in the end we gained a deep respect for each other. With industrial customers, understanding the radio side was the price of admission: they only trust a business plan or a risk assessment from someone who clearly knows the radio systems behind it.

The second step was delivering on the hard problems. The founders trusted me with some of their toughest nuts, even after I had told them I would leave, and the results were what counted: tenders priced, quotes out, and discrepancies found that an experienced reviewer had missed.

Trust also has limits you cannot shortcut. Construction companies only open up to the salesperson they have known for years, and I never got that far with them. Some credibility is simply a matter of time.

Aligning agendas

With industrial customers, you need to read the different people at the table, and that is hard. What is everybody’s agenda? What are the strong and weak points of each actor at the table? For multiple customers, I managed to get that profound human insight. I ended up leading these meetings, because I was able to align all these contradicting agendas with the least friction and reach the best solution on a human and a technical level. The technical answer is rarely the hardest part. A solution only holds when every party at the table can live with it.

Changing course

A job that does not suit you does not get better by waiting. After a couple of months I said it openly to the founders, instead of letting it drag on. That honesty is what made the rest of the year work: rather than parting ways straight away, they trusted me with the tenders, the big customers and the pricing, where my background added value. Naming a mismatch early gives everyone room to change course.

AI

I came with extensive experience in signal analysis and the AI around it. For SMEs, AI through language models like Claude will make or break their future way of doing business. These models are so fast and reason so well that, as a human, you can no longer keep up. In two of my quotes the architect’s documents contained an inconsistency. It came up during the quotation phase, I asked the construction company to decide, and they chose the most cost-effective solution for themselves. In a quote for a government organisation, publicly available meeting minutes had not been carried over properly into the specifications. Without the AI tools, none of these discrepancies would have been noticed. The experienced salesperson who reviewed my work admitted he had not spotted them either.

Catching these in the quotation phase is what matters. An inconsistency found on paper costs an e-mail; found on the construction site, it costs a change and a delay. And when the architect’s documents contradicted each other, the choice was not mine to make. Putting it in front of the construction company kept the decision with the party that has to live with it.

For small SMEs, using these tools or not will make the difference. Maybe not now, but in three years the companies that did not adopt them will have fallen behind.

Change management

Bringing AI agents and simulation tools into a company where quotes were built by hand is a change, and the people in installation companies are not won over by a presentation. They are practical: they want to see that it works on their own projects.

So I did not argue for the tools; I used them on real work and let the results speak. Price lists for private tenders, antenna plans and quotes came out of the tooling and went through the same review as any other quote. The turning point was not a demo but a review: the experienced salesperson checking my quotes found that the tools had caught discrepancies he had missed himself. Adoption starts there, when the most experienced person in the room sees the tool do something he could not.

Network

Together with the radio amateur club, Connext pulled me hard into the telecom world. The synergies between the telecom sector and the IT sector are still in their infancy. I got in touch with many telecom companies that will make that shift soon. I went looking this way out of a hunger for an in-depth understanding of the market and the technology.