Attio is the most interesting CRM to appear in years and its central idea is worth understanding whether or not you use it. Traditional CRMs give you objects, being companies, contacts, deals, and force your business into that shape. Attio lets you define the objects and relationships that match how your business actually works, which sounds like a minor flexibility feature and turns out to be the whole product. If you sell through partners who introduce you to portfolio companies who employ multiple decision makers, modelling that in a conventional CRM produces a mess of custom fields and workarounds that nobody trusts within six months.
Modelling it correctly from the start is the difference between a CRM people use and one they route around. Layered on top is the agent story, which is where they have placed their bet. Research agents that enrich and qualify leads without being asked. Ask Attio for natural language search and record creation.
Call intelligence that records and analyses meetings. Workflow automation that routes and follows up. The framing is a CRM that works continuously rather than one that waits for a human to type into it, and if the agents are accurate that is a substantial change to how the tool feels to own. The Academy covers this reasonably.
Lessons are short, a few minutes each, and phrased as essential features explained rather than as a course, which is at least accurate labelling. The AI features get concrete treatment, showing what an agent actually does to a record rather than talking about intelligence in the abstract. Combined with the help centre and the developer docs at their docs subdomain, a competent person can get productive without much friction. Where it falls short is the data modelling, which is precisely where it should be strongest.
The flexible object model is Attio's best idea and its biggest risk, because flexibility means you can get it wrong, and getting your CRM schema wrong is expensive to unwind once there is data and automation depending on it. What I would want is a proper treatment: here are three worked examples of business shapes, here is how to model each, here is how to decide when something should be an object versus an attribute, here are the migrations that hurt later. That content does not exist, and new users are left to work it out from feature documentation and then live with their early guesses. This is the single highest-value thing they could build and its absence is the main reason the rating is not higher.
The other gap is the one every AI-native product currently shares. Research agents produce claims about companies and people. Some of those claims will be wrong, out of date, or conflated with a similarly named organisation. A CRM whose records are partly agent-generated needs a story about verification, provenance and what happens when a sales person acts on bad information in front of a customer.
The training does not have that story, and nor does any competitor I have looked at, and it will eventually be a requirement rather than a nice addition. There is no certification, which I do not particularly mind. CRM certifications are mostly credentialism and nobody is hiring on an Attio badge. An admin path would be genuinely useful though, since whoever owns the workspace configuration is making decisions everybody else lives with.
Three point six. Clear, free, and much thinner than the product it documents. Attio is growing fast, with search interest up substantially year on year, and the training has not kept pace.