The Org Chart Is a Lie. This Tool Maps How Work Actually Flows.

01 /The Concept

Most performance reviews are built on memory and guesswork. Windmill replaces that with a continuously updated "context graph" – a live map of how frequently, how much, and how well employees actually interact with each other.

The problem it addresses: a company's formal org chart and team structure reflects reporting relationships on paper, not what's actually happening on the ground. The map of official hierarchy rarely matches the map of real collaboration.

If leadership wants accurate performance reviews, meaningful feedback loops, a real-time sense of what's happening across the organization, and a basis for improving overall effectiveness – they're better off grounding that work in the actual interaction structure among employees, not in the formal org chart.

Windmill's context graph surfaces that structure across four levels of depth.

Level one is a pure interaction map. Who gives direction to whom, who reports to whom on a practical basis, and who communicates with whom most frequently.

Level two breaks that interaction map into layers by topic – which employees interact on which subjects. These layers are built from the record in company platforms: GitHub for engineering, Salesforce for sales, calendars for meetings. But beyond those separate worlds, the system also tracks where they intersect – in shared Slack threads, in cross-functional docs, in meetings that bring multiple teams together.

Level three adds content analysis. The AI examines what's actually happening in those interactions: what principles guide collaboration, what employees are exchanging, what each party expects from the other, what each person's priorities are.

The AI can even detect which behaviors receive positive reinforcement in these interactions – what gets praised, what compromises get made, which trade-offs between competing priorities actually land.

Level four is the synthesis layer: structured findings based on observed interaction patterns, grounded in real interaction history and direct quotes. These findings are surfaced only to those with appropriate access – typically a given employee's manager.

The result is that performance reviews stop being fabricated from memory. Managers and employees spend dramatically less time preparing them – because the AI's observations typically account for 90% of the finished report, leaving only a light editing pass.

A review that used to take three hours now takes six minutes. Weekly check-ins become viable where quarterly reviews were the best anyone could manage. Coverage of actual activity rises from roughly 35% (what people can recall manually) to 80%.

The first 10 employees can be added to Windmill for free; each additional seat costs $10 per month.

Windmill raised an undisclosed seed round early last year, and just closed a $12 million round.

02 /Why It Matters

Windmill's tagline is "built for companies that treat their people as a competitive advantage." At first glance the platform seems only loosely connected to that claim.

But look closer. The era of individual genius is over. The real competitive strength of a team lies in how effectively its members function together – which requires understanding who actually collaborates with whom, on what, and how. Only then can you start improving it.

Confirm ([related review](/review/rezultat-prinosit-ne-sotrudnik-a-komanda)) was [covered previously](/review/rezultat-prinosit-ne-sotrudnik-a-komanda) back in 2023, when it had raised $18.5 million on an essentially identical premise – "end the nightmare of performance review cycles" by analyzing real employee interaction patterns rather than formal org structure.

Confirm has since evolved into a coaching platform: the AI now gives individual employees feedback on how they can improve, drawing on the same real-interaction data it was already collecting. A new funding round of undisclosed size followed in early 2025.

Wethos ([related review](/review/bez-jetogo-ne-budet-rezultatov)) raised $7.5 million on a related concept in the summer of 2024. Its platform analyzed team member interactions and built behavioral profiles for each person – "wind roses" of working style. These profiles could be used both to improve existing team dynamics and to assemble new teams for specific projects.

Wethos then took an unexpected step: it offered to simulate meetings. If the platform already understood each team member's behavioral patterns, it could model how they'd behave in a given meeting – what they'd say, what questions they'd raise, what they'd propose.

Naturally this can't be taken to an absurd extreme. So the platform analyzes scheduled meetings and clearly flags which ones it can reliably simulate versus which ones genuinely require participants to be in the room – because the AI can't confidently predict their reactions to the specific agenda items at hand.

03 /Opportunities

The central takeaway: to succeed, a company must operate as a unified team. Performance analysis should therefore focus on collaborative effectiveness – which requires knowing who's actually working with whom.

Or, to borrow an analogy that Elon Musk has used: a company's output is the sum of the vectors of where its people are heading. To maximize that sum, you first need to articulate a clear direction for the company – then continuously check whether each employee's vector aligns with it.

Brev ([related review](/review/vse-vkalyvajut-a-rezultata-net)) raised $3.3 million a couple of weeks ago on a platform where a manager can describe the company's goals and the AI breaks that down into work plans at the individual employee level.

You can't build those plans sensibly without understanding how employees actually interact – you need to know whose vectors need to align with whose.

Notably, Brev has already updated its tagline since then – now explicitly stating that vectors need to align not just between human employees, but between humans and AI agents. The concept didn't change fundamentally, but the framing became even more relevant.

The broad direction: platforms that map and analyze how employees actually collaborate – unconstrained by formal org structures. Which is exactly what AI now makes possible at scale: quickly, accurately, continuously, and in real time. Something that was practically impossible before.

One obvious extension: these systems should soon track not just human employees but AI agents too – which will shortly make up a large share of most companies' effective workforce. Brev beat this observation to print.