---
title: "Becoming a frontier company, in the open - Atlan Frontier Labs"
url: "https://atlan.com/frontier/essays/becoming-a-frontier-company-in-the-open/"
description: "Getting AI tooling does not make you a frontier company. So what does? We don't have the playbook - the primitives are being invented while we use them. Notes on what we're learning, experiments and failures included."
---

> Atlan is hosting Context Conference, bringing together the leaders and builders at the frontier of giving AI the context it needs to understand their business. It runs online on October 28, 2026, from 11:00 AM to 2:00 PM ET. Atlan co-founder Prukalpa Sankar opens and closes the day. Leaders from AstraZeneca, BNY and Verizon share why they invest in context and what they get from it. Registrants get early access to The AI Context Gap, a new study from MIT Technology Review Insights. Register: https://atlan.com/context-conference/

The first essay from Atlan Frontier Labs (Operating Notes, essay 001) by Prukalpa Sankar, Co-founder, Atlan, dated 2026.07.23, 1,653 words. Getting AI tooling does not make you a frontier company. So what does? Notes on what Atlan is learning, experiments and failures included. Part of [Becoming Frontier](https://atlan.com/frontier/).

## I sent that to our team a year ago

The essay opens on a Slack message the author sent to the team: "Phase 3 (using AI to do things better) is the next phase. Doing things better and not just faster. AI slop is a yellow flag at Atlan. Shipping something just because AI generated it isn't the bar. The bar is using AI really well to accelerate a real business outcome. If the output feels generic, lazy, or low-craft, we need to raise the bar for AI-native builders."

It followed a year of watching more work arrive as polished AI-generated artifacts: enablement documents became interactive websites, ideas became prototypes in hours, tabs upon tabs of things that looked finished. AI usage was rising and the work often looked better. But much of it went unused or was discarded. The cost of making something fell without raising the odds it would matter, and low-value output became context for the next person, or model, to build on. Slop produced more slop.

The word builders is the point: the model generated the artifact, but a human chose the problem, accepted the quality, and decided it deserved other people's attention.

"Not using AI might make one person slower. Using AI without judgment can make the whole organization slower."

## Activity is easy to measure. Capability is not

Atlan tracks licenses provisioned, weekly active users, tokens consumed, agents created, hours supposedly saved, artifacts shipped, and investors routinely say it is in the top 1% of AI adopters. Those numbers measure whether people use AI, not whether the company is getting more capable. An enablement site matters only if it changes how the field sells; an agent only if it improves a consequential decision; a campaign only if it earns attention and trust.

The value ladder (interactive figure):

| Rung | Who reaches it | What it means |
|---|---|---|
| 01 Consumption | about everyone | Tokens burned, tools opened, licenses provisioned. Only proves people are encountering the technology. |
| 02 Production | many | Artifacts get made. Volume and polish arrive, and are routinely mistaken for value because they used to be expensive. |
| 03 Adoption | few (most stop here) | The work is actually used and behavior begins to change. Most companies stall here, or one rung below, mistaking a shipped artifact for a result. |
| 04 Outcome | the point | A consequential decision, sale, or metric actually moves. The only rung that means the company became more capable. |

When creation was expensive, a polished artifact signaled commitment; effort was a natural brake. AI removed most of that brake and instincts haven't caught up. We still read volume and polish as evidence of value. Left unchecked, the result is "AI output theatre."

## The durable advantage is organizational metabolism

AI is not just a tool to adopt; it is a foundational capability to redesign work around. The question is whether the organization can turn a new capability into a new behavior, then a new outcome.

Building slides for an AI Engineer talk, the author noted that Atlan's Customer Success AI architecture changed four times in twelve months, each version making the last obsolete:

1. Notebook LLM: generic, basic customer context; summarised transcripts from documents you pasted in yourself.
2. Intelligence Engine: reads call notes, fetches account data from Vitally, Gong and Zendesk, drafts the follow-up email.
3. Agent org: named agents under a human manager, each judging one dimension (product and health signals).
4. Relay (current): one shared memory layer with full customer context; its queue showed 78 open tasks across 16 accounts, each tagged with a risk or opportunity signal.

The models were not the variable; what changed was how much of the customer the system could see. The last version is the first that can raise a risk or opportunity nobody asked about. The author does not regret rebuilding four times: "For a company operating at the frontier, that might be the point." Models and tools will change, and today's frontier workflow will be tomorrow's default. "The durable advantage is organizational metabolism: how quickly you can notice a new possibility, test it against reality, absorb what you learn, and change how you work." Change the outcome, not just the output.

## Frontier companies get more ambitious, not just more efficient

Productivity talk focuses on multipliers: more code, campaigns, content, work per person. If the result is ten times as many emails, slides or lines of code, production has been confused with capability.

- The old question: how can AI help us do this faster?
- The new question: if today's capabilities had always existed, would we have designed this task, role, or team at all?

Live experiment, one marketer: a personalized web experience that uses intent signals to infer a visitor's company, pulls together the context Atlan already has about that account, and generates the experience as the visitor arrives, giving each visitor the information most relevant to who they are. It is not pre-generating a hundred landing-page variants; it changes the unit of work from producing pages to assembling an experience around a person's context, where what helps the visitor most turns out to be what converts fastest. Still to learn: it is an experiment, not a proven growth engine; where personalization becomes inaccurate, invasive or untrustworthy, and what review and stopping rules it demands. It changed the question from making more variants to whether every visitor can arrive to an experience built around who they are. Imagining a new unit instead of producing more of the old one is what greater ambition looks like.

## The organization has to absorb the change

Knowledge work split outcomes across specialist functions connected by handoffs, which created scale and protected expertise but fragmented ownership and made moving context a job of its own. AI gives one person capabilities that once lived across functions: a marketer can build, a product manager can prototype, an operator can create internal software. Expertise still matters; where it sits changes. Some boundaries stay: high-stakes claims need independent review, deep technical work needs deep expertise, privacy, security and brand risk need accountable owners.

The decision rule being tested: remove a handoff when it exists mainly to transport context; preserve independence when it exists to challenge judgment, control risk, or contribute scarce expertise.

In February, Atlan's marketing team reorganized into pods built around outcomes instead of functions. Still an experiment, with early signs it's working: at the marketing retreat (Atlan marketing meetup, 2026) the author expected to hear how the team was coping without traditional management and mostly heard gratitude, and that they couldn't believe they'd ever worked the old way.

## Is the real frontier human?

The more powerful AI becomes, the more the human operating it matters. Given the same model, one person produces ten acceptable versions of an existing asset; another asks whether the asset should exist, reframes the problem, and attempts an outcome once beyond an individual's reach. Judgment decides what's worth doing, taste recognizes what excellent looks like, ambition sets the altitude. An expert has to keep consenting to become a beginner. Atlan has started rethinking human skills directly, including a simple grid with two axes, human skills and AI leverage, to think about where human impact comes from.

The next few years may feel like fight, flight, or freeze as roles change and familiar skills become abundant, but fear cannot be the operating principle. The organization must create room for learning, make changing expectations explicit, and help people develop the skills the new system requires. "AI supplies the leverage; the human chooses the direction." The human frontier is whether a company can increase machine leverage without diminishing human agency.

## We don't have the playbook

Nobody has a finished playbook; the primitives are being invented while we use them. Open questions: what happens to the role of the manager? How should AI-native work be measured? What does a 100x outcome look like without becoming a 100x productivity treadmill? Which expertise matters more, and which handoffs disappear?

Many companies writing openly about AI-native work were born in the AI era with no pre-AI organization to unwind. Most companies have existing customers, roles, systems and identities and must transform without losing trust, expertise and business. Atlan sits between those worlds: rebuilding itself around AI while building the technology that helps some of the world's largest enterprises do the same, seeing both Bay Area AI-native startups and institutions with decades of history.

Writing in the open because social media turned AI transformation into a highlight reel that hides failed experiments, abandoned architectures and quiet fear about jobs. "Atlan will be our laboratory, not the hero of every story." Views will change as the evidence changes, and Frontier Labs will keep returning to the question missing from much of the AI conversation: what happens to the human?

"The frontier company will not be the one that uses the most AI. It will be the one that can responsibly shorten the distance between a new capability and a better outcome - and keep changing its form without losing its purpose." Maybe the ultimate frontier isn't technological at all; that's the question Frontier Labs will keep testing.

## Author

Prukalpa Sankar, Co-founder, Atlan / Frontier Labs ([LinkedIn](https://www.linkedin.com/in/prukalpa/), [X](https://x.com/prukalpa)). Prukalpa co-founded Atlan, the context layer for AI. She writes from inside the attempt to build an AI-native company: organizational metabolism, the context layer, and what happens to the human as intelligence becomes abundant.

More essays: [Becoming Frontier](https://atlan.com/frontier/).