The 2026 Revenue operations Playbook

Why RevOps changed

Revenue operations was always a plumbing job, and the strategy part of the role got whatever hours were left on Friday afternoon. The team wrote routing rules, merged duplicate accounts, chased reps for close dates, and rebuilt the forecast spreadsheet every Monday. Territory design, pricing policy, and the comp plan waited until a crisis forced them onto the calendar.

Agents change the plumbing first. A routing agent reads every inbound lead, finds the account, checks the territory map, and assigns an owner in minutes. A hygiene agent sweeps the CRM every night. A deal desk agent builds the approval packet before a human opens the request. None of this is new work. It is the same work that used to eat 60% of a RevOps team’s week, now done in the background and checked by exception.

That shift creates a new risk. An agent that writes to the CRM at 3 a.m. can fix 9,000 records before breakfast, or break them. A deal desk agent that approves discounts can set a price precedent with a customer that your CFO never saw. So the question in this playbook is not whether to use agents in revenue operations. You will. The question is where you draw the line between the work agents do alone, the work they prepare for a person, and the calls that stay with people no matter how good the model gets.

A rule for the split

Sort the work by cost and visibility, not by difficulty. Most teams start from the wrong question. They ask which tasks are easy enough for an agent. Enrichment looks easy, so they automate it; the forecast looks hard, so they keep it. But difficulty is the wrong axis. An agent can do hard analysis well. What matters is what a wrong answer costs and how quickly you see it.

1. 01List every recurring RevOps task for one month. Include the small ones: reassigning a lead, fixing a close date, answering a rep’s “who owns this account?” message.
2. 02For each task, write down what a wrong answer costs in dollars or in trust, and how many days pass before someone notices it.
3. 03Give agents full control of the tasks where a mistake is cheap and shows up within a day. Routing, enrichment, and most hygiene fall here.
4. 04For tasks where a mistake is expensive, let the agent prepare the work and recommend an answer. A named person makes the decision. Deal approvals outside guardrails and renewal risk calls fall here.
5. 05Keep with people every call that sets a precedent with a customer, a rep, or the board: pricing exceptions, the forecast commit, territory design, and comp disputes.
6. 06Write the split into the system, not into a wiki. Use field-level permissions and approval steps so an agent cannot cross the line even when its instructions are wrong.

The last step is the one teams skip. A rule that lives in a prompt is a suggestion. At Northpass, the agents use an integration user in HubSpot that has no write access to deal amount, stage, or owner on open opportunities. If the agent decides a stage is wrong, it opens a task for the rep. It cannot change the stage itself, and that is on purpose.

• Agents act alone: lead routing, enrichment, duplicate merges with high match confidence, activity logging, stale-field reminders.
• Agents prepare, people decide: discounts outside guardrails, non-standard payment terms, renewal risk flags, forecast adjustments.
• People only: list price and discount policy, the forecast commit, territory boundaries, quota, and comp plan exceptions.

Routing and enrichment

Lead routing is the clearest win, because speed matters and every mistake is visible the same day. A lead assigned to the wrong rep gets a reply from the wrong person, and that rep forwards it within hours. Compare that with the old failure: a lead that sat in a round-robin queue over a weekend while three rules disagreed about who it belonged to.

A routing agent does four things in order. It matches the lead to an existing account by email domain, then by company name and address when the domain is a personal one. It enriches the record with headcount, industry, and region. It checks the territory map and any open opportunity on the account. Then it assigns the owner and writes a one-line reason on the record, such as “Matched to Acme Freight by domain; open opportunity owned by J. Ortiz.” The reason line is the part that saves you during an escalation.

Brightwater routes about 1,450 inbound leads a month this way. The median time from form submit to assigned owner fell from 2 hours 50 minutes to under 5 minutes. Reassignments, the count of leads that a rep sends back as “not mine,” dropped from 11% to 2.4%. That second number is the one to watch. Speed without accuracy just moves the mess faster.

Keep the CRM clean

CRM hygiene is a nightly job now, not a quarterly project. Every RevOps leader has run the cleanup sprint: two weeks before the board meeting, the team exports accounts to a spreadsheet, finds duplicates by hand, and merges them in batches. Three months later the CRM is as dirty as before, because the causes never stopped.

A hygiene agent works the other way. It runs every night on the records that changed that day, plus a rolling sample of older ones. It looks for duplicate accounts and contacts, opportunities with a close date in the past, open deals with no activity in 21 days, and contacts who left the company. It fixes what it can fix safely and sends the rest to the right person as a task.

At Northpass the first full sweep found 11,200 duplicate accounts. The agent merged 9,870 of them automatically, because the domain and the legal name both matched and neither record had an open opportunity. It sent the other 1,330 to a review queue. In the first month, people reversed 41 of the automatic merges, a 0.4% error rate. Every one of the 41 was a parent company and a subsidiary on the same domain, so the team added a rule: no automatic merge when either record has a parent account.

• Merge automatically only when two independent signals match, such as domain and legal name, and neither record has an open deal.
• Never let the agent edit amount, stage, or close date on an open opportunity. It asks the rep with a task and a link.
• Log every agent change with the old value, the new value, and the reason, so you can reverse a bad batch in one step.
• Audit a random sample of 50 agent changes every week. Raise the sample when the error rate goes above 1%.

The deal desk

The deal desk is where agents save the most time and where the line matters most. A deal review used to mean a rep filling in a form, a deal desk analyst pulling margin data, a finance partner checking payment terms, and a legal reviewer reading redlines, each in a different tool and on a different day. Most of that time was assembly, not judgment.

Now the agent assembles the packet the moment a rep submits a quote. It compares the discount with the guardrails for that segment and term length. It pulls the five most similar closed deals and their final prices. It computes gross margin with the support and hosting cost for the plan. It flags non-standard payment terms and summarizes any legal redlines against the approved clause library. The approver reads one page instead of opening six tabs.

Fennick Health reviews about 212 deals a quarter. Requests inside the guardrails — up to 15% off list on an annual contract, net-30 terms, standard paper — now get approved by the agent with no human step, and that covers 64% of volume. The rest go to Priyanka’s team with the packet already built. Median turnaround fell from 31 hours to 3.5 hours across all deals. The team did not get smaller. It spends the saved time on the 36% of deals that need a real decision.

Watch for one quiet failure. If your guardrails are loose, an agent that approves inside them will approve a lot of bad deals very fast. Before you turn on automatic approval, run the agent in recommend-only mode for one full quarter and compare its calls with your approvers’ calls. Fennick found that 9% of deals the agent would have approved were deals a human had rejected, almost all for multi-year terms with a price lock. They tightened the guardrails before they went live.

Renewals and the forecast

Renewals reward early evidence, and agents are good at collecting it. At 120 days before each renewal date, the Northpass agent builds a renewal file: seat usage against contract, the trend over the last two quarters, open support tickets, changes in the customer’s champion or economic buyer, and the tone of recent calls from Gong. It scores the risk and sends the file to the account manager with a recommended next step.

The agent does not decide the renewal strategy. It does not offer a discount to save an at-risk account, and it does not mark an account as churned. Those calls go to the account manager and their director. What changed is that every account manager now starts the renewal conversation with the same file, four months out, instead of finding the problem in the last two weeks.

This is why the forecast commit stays human. The commit is not a prediction. It is a promise the CRO makes to the CEO and the board, and the CRO is accountable for it. An agent roll-up is a second opinion with no ego in it, which is very useful. But the person who signs the number must understand every deal that moves it, and must be able to say why they disagree with the agent when they do.

Use the same pattern for territory planning. Let the agent model the options: account counts, total addressable revenue, and historical win rates for each proposed split. Let the sales leader and RevOps pick one and own the conversation with every rep it affects.

The stack, ranked

Choose your tools for what agents can read and write, not for what the dashboards look like. We asked the four contributors to score the CRM and RevOps tools they had used with agents in the last eighteen months. The scores reward a complete API, field-level permissions, a clean change history, and approval steps that an agent cannot skip.

The CRM matters less than the discipline around it. Every contributor ran agents well on a different system. What they shared was a separate integration user for each agent, a written list of fields each agent may change, and a change log that someone reads every week.

• HubSpotCRM, marketing, and sales software for growing companies
• SalesforceCRM and business applications for sales and service teams
• AttioThe CRM for go-to-market teams

So what do you do?

Start with the work nobody wants, because that is where the hours are and where mistakes are cheap. Do not start with the forecast. A forecast agent that is wrong in its first quarter will cost you the trust of every sales manager, and you will not get that trust back for a year. Routing and hygiene earn the trust that the harder work needs later.

1. 01Week 1: Run a two-week time study of your RevOps team. Log every task and its hours.
2. 02Week 3: Sort the tasks by cost of a mistake and days to notice it. Agree the split with your CRO and CFO in writing.
3. 03Week 4: Create one integration user for each agent with only the fields it may change. Turn on a change log.
4. 04Weeks 5–8: Launch the routing and enrichment agent. Track time to owner and the reassignment rate.
5. 05Weeks 6–10: Launch the hygiene agent in report-only mode, then let it merge with two-signal matches. Audit 50 changes a week.
6. 06Quarter 2: Run the deal desk agent in recommend-only mode for a full quarter. Compare its calls with your approvers before you allow any automatic approval.
7. 07Quarter 3: Add the agent forecast roll-up next to the rep commit. Never in place of it.

Then use the time you get back. Adaeze’s team used it to rebuild their territory model for the first time in three years, and to write a discount policy that the deal desk agent could enforce. That is the real return. Agents do not make revenue operations smaller. They give the team the hours to do the strategy part of the job, the part that was always on the calendar for Friday afternoon.