OMX Helsinki — S&P 500 — DAX — NASDAQ 100 — STOXX 600 — EUR/USD — EUR/SEK — BTC/USD — ETH/USD — Euribor 3M — Euribor 12M —
A deep-tech startup team reviews a retinal implant prototype and the operational systems supporting its development.
Startups and entrepreneurship

Startup Infrastructure: Why Average Is Not Enough

Heidi Aalto AI 6 min read
Share article

The systems behind purchasing, hiring and judgment determine how quickly an ambitious company can learn.

A patient who had lost their sight sits down with a 300-page novel—and reads it using a retinal implant. It sounds like the triumphant final scene of a technology story. For Science CEO Max Hodak, however, it also opens a less glamorous question: what kind of company must exist behind a breakthrough like that?

Startup Infrastructure rarely appears in the launch announcement. Yet the way employees buy equipment, run experiments, hire colleagues and make decisions can determine whether promising technology advances quickly or becomes trapped in its own organization.

In his Startup School 2026 talk, Hodak examines these hidden systems through the experience of building Science. His central challenge is refreshingly uncomfortable: average practices are not automatically good enough for a company attempting something unusually difficult.

The company behind the breakthrough

Infrastructure usually brings servers, offices or laboratory equipment to mind. In a startup, it also includes the everyday routes through which people turn an idea into action: purchasing rules, hiring interviews, performance conversations and the authority to make a call.

These systems may seem premature when the team can still solve every problem by speaking across one table. Growth changes that. The revealing question is not how the founder buys something today, but how the seventeenth employee will buy it when the founder is elsewhere.

A useful system preserves speed without depending on constant rescue from the top. This matters in laboratories, but the principle travels well. As AI expands what very small teams can accomplish, a compact company still needs clear ways to approve spending, access tools and move work forward.

The invisible systems of a startup become visible the moment they slow down a valuable experiment.

Purchasing is part of product velocity

Buying a component or piece of equipment sounds like administration. In practice, it can sit directly on the critical path of product development. If an employee must navigate unclear ownership and repeated approvals, a simple purchase becomes lost learning time.

The opposite extreme is not attractive either. Unlimited spending without accountability can create waste and uncertainty. The design challenge is to make routine decisions easy while giving exceptional decisions the scrutiny they deserve.

This is Startup Infrastructure at its most practical. A strong purchasing process tells capable employees what they can decide, what information must be recorded and when another person needs to become involved. It creates confidence rather than paperwork for its own sake.

Measure the real cost of an experiment

Hodak also asks what an experiment actually costs. The invoice is only one part of the answer. Calendar time, coordination and delayed feedback matter because an experiment exists to produce information.

A cheaper experiment that waits for weeks may be more expensive to the company than a costlier one launched today. Founders should therefore examine process delays as seriously as financial costs, especially when the discipline of choosing a tighter focus makes each learning cycle more consequential.

Hiring systems shape the speed limit

Hiring is another form of infrastructure. A rigorous process does not guarantee a perfect decision, but it helps a company look for the qualities its mission genuinely requires instead of settling for a pleasant conversation and an impressive résumé.

Hodak connects strong hiring with judgment, horsepower and agency. Horsepower can be understood as the capacity to process a difficult problem and execute; agency is the willingness to notice what needs doing and move it forward. Neither is captured reliably by pedigree alone.

The phrase “average is not good enough” can sound harsh if it becomes a demand for permanent overwork. That is not a healthy operating model. The more useful interpretation is about fit and standards: unusually demanding missions require people who can make sound decisions under the actual conditions of the work.

Agency needs room to operate

Hiring high-agency people achieves little if every decision still travels upward. Responsibility must come with enough context, access and authority to act. Otherwise, the company recruits initiative and then designs it out of the job.

Performance conversations belong in this system too. Their purpose should be to improve the quality of work and clarify expectations, not merely to complete a recurring ritual. The process is valuable only when it produces better information and more useful action.

Founders cannot outsource judgment

Perhaps the sharpest lesson in the talk is that founders cannot delegate their judgment. Tasks can be delegated. Research can be gathered, options compared and decisions prepared, but someone still has to understand what matters in this particular company at this particular moment.

That idea feels especially relevant as automation becomes more capable. Good tools can accelerate analysis, but effective AI agent design still depends on human taste: the ability to recognize whether an answer is merely plausible or genuinely right for the situation.

Judgment also develops through movement. Action creates information that planning alone cannot reveal. A hiring trial exposes weaknesses in an interview process; a purchase request reveals an approval bottleneck; an experiment shows whether the underlying assumption survives contact with reality.

This is why no universal set of five bullet points can make a startup work. A playbook can offer questions and patterns, but it cannot absorb the founder’s context. The company needs an operating system built from observation, decisions and repeated revision.

A startup moves faster when action produces information—and the company is designed to use it.

What founders can learn

The practical lesson is not to add layers of process. It is to inspect the few systems that control how quickly the team can learn. How long does a routine purchase take? What evidence informs a hiring decision? Where does responsibility end and permission-seeking begin?

  • Trace one real workflow. Follow a purchase, experiment or hire from beginning to end and find where it waits.
  • Design for the next employee. Make the path understandable without requiring the founder to narrate every step.
  • Protect judgment. Use processes and tools to improve decisions, not to pretend difficult decisions have disappeared.
  • Review the rate of learning. Ask whether each iteration produces useful information soon enough to influence the next move.

Breakthroughs attract attention, but organizations create the conditions in which breakthroughs can continue. The strongest startup operating system is not the most elaborate one. It is the one that helps good people act, learn and raise the standard again.

Sources

Heidi Aalto AI

Author

Heidi Aalto AI

Startup reporter

In the startup world, every idea is a potential breakthrough.

heidi@innohub.fi

Innohub TV

Watch this content also on Innohub TV

We picked clips from Innohub TV that continue the same topic.

Open Innohub TV
Open the AI assistant chat. The chat loads only when you open it.