Introduction
Agents can now research accounts, draft emails, update the CRM and watch a pipeline around the clock. Almost every sales tool has added one. What most teams still lack is an answer to a simpler question: which jobs should an agent actually do for us?
You can't pick a tool before you know the jobs. So we wrote them down: every job in sales, RevOps and customer success that an agent can take on today, placed by how much of the job the agent owns and how complex the job is.
Where teams are today
As of 2026, most teams use agents to give their reps leverage. Traditional CRMs like HubSpot, Salesforce and Pipedrive have bolted AI on: conversation intelligence logs what was said into fields, and the AI summarizes calls, drafts follow-ups and preps reps before meetings. The rep reads the output, decides and sends.
We call this the Leverage Lab: low autonomy and low complexity. It is a sensible place to start. The agent saves the rep time, and a mistake costs little because a person checks every output. It is also where most teams stop. The jobs that change how a team sells sit further up and to the right.
The Agent Jobs Map
To see where agents can go next, we placed each job on two axes. Autonomy is how much of the job the agent owns, from answering a question to owning a goal. Complexity is what the job touches: how much context it needs, what a mistake costs and how many people it reaches.
That gives four groups. The Leverage Lab is the bottom left, where an agent helps one rep do the job faster. Bottom right, an agent analyses every account and the team decides. Top left, background agents run on their own and flag what a person should act on. Top right, agents act on their own toward buyers and the CRM, which only works with full context and clear rules.
Which GTM jobs can agents own?
These are the jobs agents can take on today. The list will grow as agents take on more of the work, and we will add jobs as they become real.
Build or buy
Once you know the jobs, the next question is how to get them done. Teams have two paths.
Build around what you have. Most teams already run a CRM like HubSpot or Salesforce, and those CRMs now ship AI that summarizes calls, suggests tasks and drafts simple follow-ups. That covers much of the Leverage Lab. They stop where a job needs the full customer story: what was said on calls and in email, what was promised and how the relationship has changed. A CRM stores the current state of a deal, not that history. Building means you keep your CRM and add the missing context and the agents around it.
Buy a complete solution. AI-native CRMs like Lightfield, Day AI and Dreamteam rebuild the CRM around agents, with the context and the agents inside one product. You get a lot working quickly. The cost is switching: you move your data, your team and your processes into a new system of record, and you work with the agents and limits that product chooses.
So the real question in GTM right now is whether you build or buy, and for which job.
The database
We built the database to answer that job by job. For each job it shows what you can buy today and where those tools stop, and what it takes to build it yourself: the tools you combine, the steps and the skill an agent runs. Where a tool already does the job well, we say so.
Our build path runs on Northstar, our open-source context graph. It connects your calls, email and CRM into one record per customer, so any agent can read the full story before it acts. You keep your CRM, pick the jobs you want agents to take on and build around your own needs.
