Construction

Why one solved task beats a full platform in legacy industries

Annija Eizenarma on why the AI companies that sell into construction solve one task first: one design partner can open 15 doors, then 100.

The AI products that get bought in construction and other legacy industries do one narrow thing, plug into the systems already in place, and prove a result before anything larger is proposed. Annija Eizenarma, a founder, investor and adviser with seven years building technology businesses in legacy industries, describes the compounding: one design partner can open 15 doors, and those 15 can open 100.

The platform pitch fails in the same market: the buyer cannot evaluate ten claims at once.

What is a legacy industry?

A legacy industry is an established, profitable, physical-world business sector — construction, manufacturing, logistics, agriculture, hospitality, trades and similar — that has operated for decades on minimal software, and where the core work happens in the physical world rather than in a system. Eizenarma, asked for the term constantly, defines it this way:

"Are businesses that are rooted in the physical world, that are real-world businesses like construction, equipment rental, parking, manufacturing, agriculture, logistics, trucking, hospitality, plumbing — all of the businesses that are everywhere. They're profitable, they have been running basically for over a decade and decades with very simple tech solutions, and are very physical."

Three properties there set the commercial rules: profitable, so they can buy; decades old, so they have habits and vendors already; physical, so software describes the job rather than being the job.

Why does the physical workflow matter more than the technology?

Because the strongest predictor of success in Eizenarma's account is whether the builder learned the day-to-day work first, not the model or the stack.

"The companies that do manage to succeed are really good in understanding the physical world and the work that's done by every single entrepreneur in legacy industries. Everything starts with relationships, and to the detail understanding their workflows, their day-to-days."

And the other side of it:

"The ones that are trying to reinvent the process of how the legacy industries are working are often less successful, because we do have to find a common language with the people that have been in a business for decades to gain that trust."

The distinction is between automating the process that exists and replacing it with a better one. The second is the more interesting engineering problem and the worse business decision. A workflow that has survived thirty years is not arbitrary: it encodes constraints the builder has not discovered yet.

What does a site visit produce that market research does not?

Requirements. Galina Fendikevich, host of Artificial Reality, the AI buyer-evidence podcast, described a walnut shell factory in California where a planned one-hour visit became a six-hour conversation with a member of the family that grew the walnuts.

Her read: the AI opportunities became obvious in the specific — "little bits here and there" — rather than as a platform. And the visit produced what a market-sizing exercise never does: "you get the best product requirements you could ever ask for from them."

Why does one task beat ten features?

Because trust here is earned sequentially, and each solved task buys the next conversation. Eizenarma's most repeated point:

"I think the most successful ones are the ones who can really narrow down their focus and focus on one thing only to begin with... by starting with one specific task that you can solve, you can show the results and you can build that trust afterwards, and then heading to the next task and next task."

The platform alternative asks the buyer to commit to everything at once:

"It's overwhelming for the legacy industry buyer to even learn about all these features and all of the use cases and all of the tools you have built. So taking it one step at a time, and not exposing all of the ideas, all of the vision you have right away, is often a very smart move."

Fendikevich describes how that lands from the buyer's side: like Disneyland — dizzying, and impossible to answer the question of how you transform your entire business around a platform like that. Breadth reads as risk.

How does the first customer become distribution?

By vouching for you inside rooms you cannot buy into. Referral is the go-to-market motion here:

"The industry players really listen to each other, and if you find somebody that can truly recommend your product from their industry and vouch for you, that is the absolute golden card in the industry. That's how the legacy industries operate. They don't trust anybody besides their trusted advisory board."

That makes the first customer a distribution decision, not a revenue one. Her selection criterion is growth intent: "There are many legacy businesses that really don't want to grow... I'm exactly where I want to be, never call me again, I don't want to grow. And that's fine." A growth-minded first partner becomes an ambassador; a satisfied one is a dead end however well the product performs.

What should a builder change this quarter?

Cut the roadmap to one task that produces a visible result on its own.
Build on top of what is installed rather than asking the operator to move off it.
Spend the visit, not the survey. Six hours in one facility beats a market-sizing deck.
Qualify the first partner on growth intent, not deal size: that customer is your distribution.
Strip the vision from the first meeting. Show one result, then earn the second conversation.

The test to run on your own pitch this week: could your first customer describe what you do in one sentence, to a peer, without the word platform? If not, you are asking a profitable decades-old business to evaluate ten things at once, and it will decline by not answering.