Agents are breaking your software
The old architecture was never intended for them
GTM needs a new architecture
The story
For thirty years, software has come in two shapes.
Introduction
Application software was built for people. It was safe and forgiving, with the aim of enabling low volumes of high-context, non-deterministic work. It didn’t matter that it was slow, because people could only move so fast, and there was a limit to the work that could be done.
Infrastructure was built to be driven by machines. It was fast, complex, and designed for deterministic work at scale. You could query millions of rows and get answers in seconds, but only for predetermined questions about highly structured data.
Neither shape was built for agents.
An agent can do the high context work of a person at the speed and scale of a machine.
Agents unlock previously impossible customer experiences at unprecedented scale, and provide a radically personalized customer experience, freeing your team to focus on the highest leverage work. An agent can assist every customer relationship, hold more context than any single person could, and find growth opportunities that you couldn’t see before.
In the history of software, there’s never been a better time to be unreasonably ambitious about what you can achieve, and there is so much more potential.
Agents + automations reach far beyond the manual work a person can finish: rings grow out from a person to an agent at the outer edge.
But we’re only scratching the surface.
To realize this potential and build context at scale, agents need more than any system of record was built to provide.
People build context by consuming an extraordinary amount of information. We read emails, listen to calls, and message each other on Slack. That information lets us provide great customer experiences, and the system of record was the artifact we created with that context.
For agents to deliver these experiences, they need to centralize and index all of that disparate, unstructured data, within the system of record.
The problem is that this rich context exceeds what those systems can store, and forces agents to use infrastructure instead. Infrastructure wasn’t built for this. Infrastructure doesn’t offer the guardrails, observability, or validation that you need. Agents are constrained by the limitations of our tools.
There are five reasons why:
Scale
For every one record a human might check, an agent can read thousands. Software built for human traffic can’t carry that load.
Context
Agents are only as good as the context you give them. They’ll read and act on whatever they can find, even if it’s only half the picture.
Interface
Agents need an interface that represents the rules and validations of the system without the overhead of UI made for people.
Language
Agents are trained to use common coding techniques. Proprietary languages and complex UI don’t make sense to them.
Permissions
Agents require a different kind of identity to people. They need to see deeper context than a person would, but for a much narrower slice.
Every CRM is built as application software, and inherits these problems. All were built for a person , and all are about to be driven by something that is not one.
Software has never mattered more, yet it’s never been less prepared.
Power at this scale needs something that has never existed before.
- A new architecture with the guardrails of application software and the throughput of infrastructure.
- A single system, equally built for people and agents.
- GTM needs a new shape of software.
One System, People + Agents: rings around a person and rings around an agent overlap in a single shared system.
The architecture
This is the architecture for agentic revenue.
The next phase of Attio.
Attio is a hybrid system, architected for people and agents in equal measure, with the control of application software and the scale of infrastructure.
GTM teams have always had to choose between speed, scale, and quality. We removed the choice. Attio works at the speed of an agent, across your entire market, to the standard of your best rep. That is an order of magnitude more value from your time, because each task an agent completes leaves context behind for the next one, so that context compounds. The longer it runs, the sharper it gets.
The same ring drawn at rising resolution as an agent completes 1, 10, 100 and then 1,000 tasks, until it resolves into a true circle.
This is what we’ve always been building.
Three stacked layers. Context: Universal context, Apps, Permissions. Orchestration: Agents, Workflows. Ecosystem: App SDK, MCP, API.
Context
Every customer signal is captured in one shared context layer for your team and agents. It’s high-resolution, always current, continuously compounding, and rich enough to act on.
Orchestration
Agents cover every motion your revenue team needs to run, supported by the context layer. They act on the right signals the moment they appear.
Ecosystem
Attio reaches wherever you and your agents operate. It’s extensible by default, with first-class agentic access on every surface, held to the same standard everywhere.
The launch
Introducing the next phase.
SQL Engine
A team of data scientists on every question.
Attio now speaks SQL, the data language of agents. They use it to reason across millions of records, objects, and relationships at once. Complex joins, calculations, and aggregations now run natively in Attio.
Ask questions no predefined endpoint can answer. Calculate weighted forecasts, compare win rates, or measure your progress against targets in one query. Attio will perform the computation itself and return an exact result, keeping every query visible for you to inspect.
SQL at the centre, reached from Ask, MCP, API and Workflows.
Data Connectors
Skip the sync and go straight to the source.
Attio now exposes a live, Postgres-compatible endpoint, connecting your CRM directly to your agents and your data stack in realtime.
Build dashboards, run complex analyses, and let agents query live customer data without exports, data warehouses, or nightly syncs. All your data stays live, consistent and readable straight from the source.
A globe of tools connected to Attio data: Hex, Omni, Looker, Metabase and dbt.
Agent Scale
Thousands of agents. Zero traffic jams.
Attio is engineered to support hundreds of millions of records and retrieve them at agent scale.
Research every account in your market at once and re-qualify entire pipelines overnight. As your business changes, keep millions of records current with thousands of agents reading, reasoning, and acting with the same permissions as the people they work with.
A curve climbing steeply through 0, 5, 10, 750, 5,000 and 1,000,000 as the work scales.
Beta
Activities
If it moves the deal, it’s on the record.
Your customers show up in more places than just your inbox. With Attio, pull in support tickets, social DMs, product events, and whatever else moves your deals. Custom schemas let you capture the details unique to every interaction.
Each activity becomes structured, first-class data on the relevant record, giving your team a more complete understanding of every relationship, and your agents the context they need to act on it.