Your Company Has Amnesia – Every Decision Evaporates After It's Made

01 /The Concept

Every company loses knowledge faster than it gains it. Sentra is built to change that – a platform for preserving "corporate memory": the accumulated reasoning behind decisions, not just their outputs.

The closest analogy is GitHub. But where GitHub tracks who changed what in a codebase, Sentra tracks who made which decision inside the company – and why. As a result, the platform can flag in real time when a new decision someone is about to make conflicts with one the organization made months or years earlier.

To do this, Sentra connects to the communication systems where employees actually work – video conferencing, email, messaging apps. It then listens, learns, and remembers.

Sentra also tracks commitments: if an employee promised a colleague something, the platform will remind them before the deadline slips by.

Managers can ask the platform for a weekly digest – everything that happened across the company, a team, or a department, even if they weren't personally in the room. And when someone heads on vacation, they can delegate Sentra to attend the meetings they'll miss and brief them on everything important when they return.

For new hires, Sentra functions as an institutional guide – drawing on both the company's shared memory and the specific knowledge of the person they're replacing.

Sentra was founded last year. The startup has now closed its first funding round at $5 million.

02 /Why It Matters

The need for institutional memory becomes acute at predictable inflection points in a company's growth:

- Around 20 people, lunch conversations can no longer cover everything. - At 50, a gap starts forming between product and engineering. - By 150, no one knows what the other teams are doing. - Past 500, executives are managing a company they no longer fully understand.

A corporate memory platform addresses each of these:

- An engineer can ask the platform why the product's architecture looks the way it does – and get a structured history of decisions and trade-offs. - A new VP can find out why pricing changed last quarter – including who argued against it and why. - A founder can get a Monday morning briefing covering sales, marketing, and product development from the week before. - A sales lead can trace exactly why pipeline velocity has slowed, right down to the specific people involved.

Daniel Kahneman once described corporate thinking as operating in two modes:

- System 2 is output-driven management: quarterly reports, Jira tickets, PRDs. - System 1 is input-driven: the meetings and conversations where decisions actually get made.

Nearly every enterprise platform runs at the System 2 level, tracking outputs – even though outputs are just derivatives of System 1 activity. What makes Sentra rare is that it operates at System 1. It captures the reasoning behind decisions without requiring anyone to have been physically present when those decisions were made.

Sentra goes further, claiming the platform generates a "System 3" – collective intelligence at company scale. That's a meaningful distinction from most AI tools on the market, which operate as individual chat assistants. Sentra improves the performance of the organization as a whole: decisions made by one person become part of a shared foundation that accelerates and improves decisions made by others.

There's also a compelling retention dynamic here: once a company starts using Sentra, walking away becomes difficult. The accumulated decision history is too valuable to discard and too hard to migrate elsewhere – a built-in moat.

Notably, this category doesn't require universal, company-wide deployment to be useful. Specialized versions can thrive in narrower contexts.

A [related review](/review/dlja-raboty-nuzhen-komandnyj-mozg) covered Grov, a startup that built a similar memory platform purpose-built for engineering teams – tracking architecture decisions, implementation choices, and bug history so teams don't keep stepping on the same rakes.

03 /Opportunities

One direction is building corporate memory platforms – general-purpose or specialized. Specialization can mean a particular functional domain (like Grov's engineering focus), a specific use case, or a different method of capturing and organizing institutional knowledge.

Sensay ([covered here](/review/malozametnaja-problema-kotoraja-skoro-prevratitsja-v-katastrofu)) raised $3.4 million for a platform that runs AI interviews with departing employees to extract their knowledge before they walk out the door. Viven ([covered here](/review/ii-dvojniki-jeto-milliardnaja-tema)) raised $35 million in its first round for digital twin technology that lets employees' knowledge continue to answer colleagues' questions even when the original person is on vacation, traveling, or in a different time zone.

The broader direction traces back to the System 1 vs. System 2 distinction. System 2 is managing from outputs – documents containing conclusions. System 1 is managing from inputs – the underlying reasoning those conclusions came from.

System 1 thinking, it turns out, is remarkably close to what Aristotle called first-principles reasoning – and what Elon Musk popularized as his core decision-making method. The idea is to reason from foundational premises rather than from someone else's conclusions, which nearly always surfaces options the conventional approach would miss.

Old platforms were stuck in System 2 because that was the only thing technically feasible. But AI can now ingest massive volumes of unstructured raw information, organize it, and help surface new conclusions from it every time.

That has given rise to platforms like Sammy ([covered here](/review/prishlo-vremja-peredelyvat-a-ne-uluchshat)) – which started as an AI agent that crawled each new product release and auto-generated up-to-date user documentation and onboarding materials. The platform has since expanded, but the first-principles operating logic remains.

Or Day.ai ([covered here](/review/chtoby-pobedit-nuzhno-peredelat)) – which raised $4 million in its first round to build a "next-generation" CRM that ingests raw customer conversations instead of relying on reps to fill in forms with their own interpretations. The AI answers standard CRM queries, but derives its answers from the source material – not from whatever a rep happened to enter. The platform has since grown more complex, but the underlying architecture is unchanged.

So here's the question worth sitting with: which other categories of legacy enterprise software can now be redesigned for first-principles operation?