Our thesis for the new era of work

Every company is about to be rewritten. Both in their product and in their operating model.

The way software ate the world twenty years ago, AI agents are starting to eat software. The companies that defined categories by digitising manual processes will be replaced by companies that eliminate the processes entirely. CRMs that just store records. Support tools that just route tickets. HR platforms that just hold data. All of it, rebuilt from the ground up.

For two decades, the cost of running people operations has tracked headcount almost perfectly. Roughly $3,000 per employee per year, barely moving. Company doubles, ops team doubles. Every new entity, every new jurisdiction adds another workflow, another spreadsheet, another person holding institutional knowledge in their head.

That model was never designed. It happened. And now it is breaking.

The chatbot isn’t the answer

When AI arrived in HR software, the first instinct was to bolt a chatbot onto existing systems. Answer employee questions. Summarise policies. Surface help articles.

But answering questions was never the hard part.

The hard part is the operational work behind the many questions that can’t be resolved with documentation. The work that fires across multiple systems every time someone joins, transfers, or leaves. Contract generation by jurisdiction. Payroll coordination by entity. Compliance documentation by location. System provisioning. Manager follow-ups. The multi-step, cross-system, policy-dependent execution that consumes 80% of a People team’s capacity.

The industry saw a problem of information access. We saw a problem of execution.

From systems of record to systems of action

The last decade was about systems of record. Storing what happened. The next belongs to systems of action. Making things happen.

That’s what Athena is. It connects to your existing stack — HRIS, payroll, ATS, IT provisioning, compliance tools, Slack, Teams — and executes the work autonomously. A single trigger and Athena generates a contract by jurisdiction, notifies payroll with the correct effective dates, updates the HRIS, queues compliance documents, provisions system access, and informs the employee. Not a ticket. Not a notification routed to a human. The actual work, done.

Building this required a deliberate set of bets.

We built Athena on the most capable frontier models from Anthropic and Google. Trillion-parameter systems designed for complex reasoning, not the small, cost-optimised models most HR AI runs on. HR is ambiguous, policy-dependent, exception-heavy, and cross-system. Scripted FAQs don’t survive first contact with a real organisation. Agentic behaviour requires frontier intelligence. There is no shortcut.

We built it as a standalone platform, not a feature inside an HRIS. One that works with any system of record — because intelligence deserves its own foundation. And because we’d rather build something genuinely useful than something that locks you in.

We built it to run natively where your team already works. Slack. Teams. No new login. No new tab. No adoption curve.

Every one of these decisions made Athena harder to build. Every one made our AI agents easier to adopt.

A model that compounds

Every workflow you configure, every jurisdiction-specific rule you encode — it stays. When a new entity opens, Athena already knows the contract template, the statutory benefits, the compliance deadlines, the payroll provider.

After twelve months, Athena isn’t a tool you use. It’s the operating logic of your entire organisation — encoded, automated, and improving.

That knowledge has always lived in someone’s head, or a wiki nobody reads, or a process that breaks when its owner goes on leave. Now it lives in infrastructure that acts on it autonomously.

The headcount equation

The conventional model says a business doubling headcount needs to roughly double its People team. That model is now redundant.

Athena absorbs the transactional volume that would otherwise require additional operational hires. AI is collapsing the link between company growth and headcount growth across every function. Organisations can run leaner, move faster, and compound advantage.

The operators we talk to didn’t get into this work to chase signatures or reconcile payroll. They want to build cultures, develop leaders, attract extraordinary talent. When the operational work runs itself, they finally can.

Engineers don’t write code by hand anymore. The same is becoming true for operational work. The companies that move first will run a large organisation with the precision that used to require many times the overhead.

This is not a software subscription. It is infrastructure investment with a direct return against growth targets.

Guardrails by design

People teams make decisions that affect real humans. Athena operates strictly within defined boundaries. At the model level, constitutional AI safeguards refuse unauthorised or ambiguous requests. At the platform level, it can only act through explicitly enabled tools and workflows. If an action isn’t permitted, it won’t execute. For sensitive or irreversible actions, it pauses and hands off to a human with full context and a clear decision point.

The human is always in the loop where it matters. The human is not in the loop for work that should never require them.