Every customer-facing team should work with the same team of AI agents. SDRs, AEs, CSMs and renewals chase different numbers, but from the same customers and the same events. Share the agents that find the facts and the people. Give each team its own account plans, its own orchestration and its own dashboard. Watch critical risks all the time, research everything else by priority, and let reps decide when outreach goes out.
Most revenue teams buy AI one department at a time. SDRs get an outbound tool. AEs get deal intelligence on top of their call recorder. Customer success gets a platform with its own AI. Renewals get a spreadsheet and a calendar reminder.
Every one of those tools is watching the same customers, separately, with its own idea of what matters. So a key exec leaves a customer and the CSM hears about it a fortnight later. The AE working a deal in another part of the same business doesn't hear at all. The SDR keeps reaching out with messaging written for a strategy that no longer exists.
I spend most of my week setting up agents for exactly these teams, and I've come to a firm view: customer-facing teams should share one team of agents, with each team orchestrating it differently. Here's why, and how I think it should be set up.
Your customer doesn't care about your org chart
Every customer-facing team is measured on something different. SDRs on engagement, AEs on conversion, CSMs and account managers on expansion, renewals on retention. It's easy to treat those as four separate problems that need four separate tools.
In practice they share almost everything. A customer expanding into a new region is an expansion lead for the CSM, a new opportunity for the AE who owns a different business unit, a reason for the SDR to call the new regional team, and either good news or an early warning for renewals. One fact, four moves. If each team has to find that fact on its own, three of them find it late or not at all.
Strip the job titles away and every rep who talks to customers needs the same four things from their agents: to stay ahead of risk, to stay well informed, to be talking to the right people, and to walk into every conversation with the right strategy, context and messaging. None of that is role-specific. A budget freeze matters to the AE and the CSM on the same account. A new CFO matters to all four teams.
That's the case for sharing. When the facts live in four places, the customer gets four companies talking to them, and the gap between net new revenue and net revenue retention is often just information that never got passed on.
Share the facts, not the conclusions
Sharing agents doesn't mean every team sees the same thing. That's the quickest way to bury a CSM in sales noise. The line I'd draw is between facts and strategy.
Facts should be shared. Account research, contact research and persona cards describe the customer, and the customer is the same whoever is looking. Building them separately for each team just gives you four slightly different versions of the truth.
Strategy should be built per team. The same restructure is a risk to the AE's deal, an expansion conversation for the CSM and a reason for renewals to check who signs off this year. Each team needs its own account plan reading the shared facts through its own lens, then its own cockpit, then Plays that prepare the move.
Keep those layers apart and dozens of agents add up to clear next moves. Mix them up and you get noise, however good each agent is on its own.
Every signal should end in a rep's move. In the example above, one finding about a new region becomes a meeting plan for the AE, an expansion flag for the CSM, a stronger case for renewals and openers for the SDR. Nobody had to ask the other teams what they knew.
Watch what could cost you, all the time. Everything else can wait its turn
Most teams get research frequency wrong in one of two ways. Either they research everything, constantly, and drown reps in updates nobody reads. Or they research when a rep remembers to ask, which means the things that matter most get missed exactly when nobody is looking.
I think the right answer is three tiers:
- Critical risks, always, on every account. For AEs, that's anything that could waste months of deal work, like an exec exit or a restructure. For CSMs, account managers and renewals, it's churn signals. Nobody should be blindsided by a customer they thought was safe.
- Everything else, by priority, on a schedule. Key contacts on quiet accounts don't need checking every week. Every few months is plenty. New leaders and initiatives deserve a closer eye on the accounts that matter most.
- Deeper research, on demand. Switched on for a campaign or a project, then switched off again.
Two more rules make the difference between signals reps trust and signals they ignore. Every signal should be dated and sourced, so a rep can check it before quoting it. And every signal should be dismissable, so once it's handled it stops nagging until something new comes up.
Automate protecting revenue. Let reps lead on winning it
The question I get asked most is how much should be automated. My answer: anything that needs keeping on top of should run on its own, and reps should decide where agents go further, like finding new contacts, preparing outreach or building a meeting plan.
Where that line sits depends on the team. The narrower the book, the more you can automate. Renewals and CS work a defined set of customers, so most of their monitoring can run quietly in the background and land in their view. SDRs work the widest patch, where judgement about which account deserves attention this week is the job. Protecting revenue is mostly automated. Winning it mostly starts with a rep.
And outreach should never be generated in the background and sent on autopilot, for any team. Agents can prepare the email, the LinkedIn message, the call script and the meeting plan. Nothing should go out until a rep says go. I've written more about why in The Anti-AI SDR Manifesto.
If it isn't on screen at login, it doesn't exist
This is where most agent rollouts quietly fail. The research is good, the signals are right, and reps never see them, because seeing them means filtering a list down to their own accounts or asking an agent and waiting for an answer. They won't do either.
So I'd build one cockpit per role, around that role's job, and give every rep their own version of it: their accounts only, grouped by what needs doing (risks to act on, accounts nobody has touched, champions to engage), with the columns that matter for that job. Most of the agents behind it are the same for everyone. The front door is different for each role.
The cockpit is also where the rest of the business plugs in. When marketing launches a campaign, it should land in every team's cockpit at once, so SDRs can use it as an outreach topic and every Play tells the same story. When sales enablement rolls out a methodology, it should be built into the account plans and Plays every role already uses, so new starters get the same foundation as reps who've been there for years. (More on that in How GTM agents make MEDDIC, MEDDPICC, and SPICED stick.)
Give the agents one owner
If every team builds its own agents, you're back to four stacks watching the same customer. If every rep builds their own, you get a sprawl of near-duplicates nobody can maintain.
I'd give the whole setup one owner, usually in RevOps or enablement. They build for every team, so reps never have to build their own. They review it with each team regularly: what reps act on, what's missing and what's gone stale. And they move the line between automated and rep-launched as each team's needs become clear.
Start with two reps, not four teams
The temptation is to roll this out to every customer-facing team at once. I wouldn't. Prove it with a couple of reps in one team, an expert and a newer starter, so you learn what helps both. Once they're acting on what they see, copy it across that team. Then add the next team on the same agents, with a new orchestration and cockpit rather than a new agent stack.
If you want a structured way to check readiness before any of this, The 6 Dimensions of Agentic GTM Teams covers the audit worth running first.
The question to ask before you buy another AI tool
Before your next AI purchase for a customer-facing team, ask one thing: will every team that touches this customer see what it finds? If the answer is only one team, you're building another silo, however good the tool is.
The teams that get this right will be the ones where an SDR, an AE, a CSM and a renewals manager all know the same thing about the same customer on the same morning, and each knows exactly what to do about it.
Frequently asked questions
Can customer success and sales teams use the same AI agents?
Yes. The research behind a good sales conversation and a good customer success conversation is largely the same: who the account is, what's changing, who matters and what could go wrong. Sharing those research agents means every team works from the same facts. Each team then gets its own account plans, orchestration and dashboard on top.
Which AI agents should be shared, and which should be team-specific?
Share the agents that find facts and people: account research, contact research and persona cards. Build the agents that interpret those facts per team, mainly account plans, since an SDR, an AE, a CSM and a renewals manager each read the same signal differently. Plays and dashboards are set up per role too.
How often should AI agents research your accounts?
In three tiers. Critical risks are researched always, on every account: anything that could waste months of deal work for AEs, and churn signals for CSMs and renewals. Everything else runs on a schedule by priority, such as key contacts on quiet accounts every few months. Deeper agents are switched on for a specific campaign or project, then switched off.
How much of a revenue team's work should AI agents automate?
The narrower the book, the more you can automate. Renewals and customer success work a defined set of customers, so most of their monitoring can run on its own. SDRs and AEs work wider patches, so more of their work starts with a rep choosing where agents should go further. In every team, outreach should wait for a rep's go.
What should AI agents do for customer success and renewals teams?
Mostly watch. Churn signals such as budget cuts, competitor moves and restructures should be monitored on every account, champion job changes checked on a schedule, and expansion opportunities kept current in the account plan. Renewals should see the same risks as CS, alongside which renewals are coming up and who the new champions are.
Who should own AI agents across a revenue team?
One person or team, usually in RevOps or enablement. They build the setup for every customer-facing team, so reps never have to build their own agents, and review it with each team regularly to decide what moves from rep-launched to automated.
Different numbers, same customer. SDRs, AEs, CSMs and renewals chase engagement, conversion, expansion and retention, all from the same accounts and events.
Share the facts, not the conclusions. Research, contact research and persona cards are shared. Account plans, orchestration and dashboards are built per team.
Critical risks always, everything else by priority. Watch what could cost a deal or a customer all the time, refresh the rest on a schedule, and switch deeper agents on for campaigns.
Automate protecting revenue, let reps lead on winning it. The narrower the book, the more you can automate. Outreach waits for a rep's go.
One owner, two reps first. Give the agents one owner, prove it with a couple of reps, then add teams on the same agents.