GM. Jumping in.
Most AI in RevOps fails because teams start with the wrong object.
They start with a dashboard.
A bigger forecast view. A smarter score. A cleaner activity summary. A chat interface that explains the same messy data in complete sentences.
That may be interesting.
It usually does not change the work.
The problem is not that RevOps lacks visibility. Most teams already have more views than operating discipline.
The problem is that the same avoidable workflow decisions get made late, inconsistently, or not at all.
A rep works an account with no buying signal. A manager reviews a deal after it has already gone stale. Marketing sends leads before sales has a clean acceptance rule. Customer success flags expansion interest, but no one creates the handoff.
AI can help with those moments.
But only if you point it at workflow automation instead of dashboard theatre.
Use a simple rule:
AI should route work, prepare work, inspect work, or trigger work.
If it only describes work, be sceptical.
Start with decisions, not tools
Before buying or building anything, list the recurring RevOps decisions your team makes every week.
Not metrics.
Decisions.
For example:
Should this inbound request become an MQL, sales lead, or support redirect?
Should this account be routed to an AE, SDR, partner manager, or CSM?
Should this opportunity be inspected because it is stale, single-threaded, or missing a next step?
Should this expansion signal create a sales task or stay with customer success?
Should this forecast change require manager review?
These are strong automation candidates because they have inputs, rules, and a downstream action.
A dashboard can show that lead response time is weak.
A workflow can identify the lead, enrich the context, assign the owner, create the task, and escalate when nothing happens.
That is the difference.
AI should not be judged by how impressive the interface feels.
It should be judged by whether the decision happens faster, cleaner, and with less manual coordination.
Use the four-job automation filter
Every RevOps AI idea should pass through four possible jobs:
Route: Send the record to the right owner or motion.
Prepare: Summarise context before a human acts.
Inspect: Find missing evidence, risk, or policy breaks.
Trigger: Create the next action when a condition is met.
If an AI use case does not fit one of those jobs, it may still be useful.
But it is probably not workflow automation.
Routing is where many teams should begin.
AI can classify messy inbound requests, match accounts to territories, detect partner influence, or identify whether a customer signal belongs to support, renewal, or expansion.
Preparation is the second practical layer.
Before a manager review, AI can summarise deal history, recent emails, missing fields, next-step language, stakeholder coverage, and prior objections.
The manager still judges the deal.
The system just removes the preparation tax.
Inspection is where RevOps gets leverage.
AI can scan for opportunities with no buyer-confirmed next step, late-stage deals with no economic buyer, renewal accounts with usage risk, or leads accepted without the required fit evidence.
Triggering is the final mile.
When the system finds the condition, it should create the task, notify the owner, update the field, or place the record in the right queue.
Description alone is not enough.
Keep humans on judgement, not assembly
The best RevOps AI workflows do not remove human judgement.
They remove the manual assembly around it.
Bad automation says:
The AI decided this deal is commit.
Better automation says:
This deal moved to commit, but the system found no buyer-confirmed next step, no close-plan update, and no new stakeholder activity in two weeks. Manager review required.
Bad automation says:
The AI scored this lead 91.
Better automation says:
This inbound account matches ICP, mentions an active migration project, has employee growth, and belongs to Territory East. Route it to the SDR queue with these three context bullets.
Bad automation says:
The AI predicts churn.
Better automation says:
Usage dropped, the executive sponsor left, renewal is inside the inspection window, and no success-plan update exists. Create a CSM risk review.
Humans should decide qualification, forecast, account strategy, and commercial trade-offs.
AI should gather context, detect rule breaks, and push work to the right place before the human is late.
Build around clean inputs and visible rules
AI does not excuse poor RevOps design.
If your stages are vague, your ownership model is unclear, or your CRM fields are optional decorations, AI will mostly accelerate confusion.
Before automating a workflow, define four things:
What input data is trusted?
What rule or pattern should the system evaluate?
What action should happen when the condition is met?
Who owns the exception when the system is wrong?
This is where many teams skip the hard work.
They ask AI to “surface pipeline risk” before defining what risk means operationally.
They ask it to “prioritise accounts” before agreeing which signals matter.
They ask it to “improve forecasting” before stage exit criteria are enforced.
Start narrower.
Pick one workflow with a clear owner and a visible failure mode.
For example: stale late-stage opportunities.
Inputs: Stage, close date, last buyer activity, next-step field, stakeholder count, and manager notes.
Rule: Flag late-stage opportunities with no future buyer meeting, no next-step update, or no activity inside the inspection window.
Action: Create a manager-review task and add the deal to the weekly inspection queue.
Owner: The front-line manager reviews the deal and either confirms the risk, updates the record, or overrides it with a reason.
That is not glamorous.
It is useful.
Measure whether the workflow changed
Do not measure AI adoption by logins, prompts, or dashboard views.
Measure whether the workflow improved.
For each automation, ask:
Did the right owner receive the work faster?
Did fewer records sit in the wrong queue?
Did managers inspect risk earlier?
Did reps spend less time assembling context?
Did fewer exceptions require manual follow-up from RevOps?
You do not need invented precision.
You need operational evidence.
If the workflow still depends on someone noticing a dashboard, copying context, sending a Slack message, and reminding three people to act, the AI layer has not changed much.
It may have made the dashboard smarter.
It has not made the system better.
Weekly action
Pick one recurring RevOps workflow this week.
Use this filter:
What decision is being made?
What inputs are needed?
Who owns the decision?
What happens when the decision is late or wrong?
Could AI route, prepare, inspect, or trigger the work?
What action should the system create?
What human judgement should remain human?
Then write the workflow in one sentence:
When [condition] happens, the system should [route/prepare/inspect/trigger] so [owner] can [decision/action].
If you cannot complete that sentence, do not automate yet.
You do not need more AI theatre.
You need one workflow that moves the right work to the right owner before the revenue system drifts.
— Pipeline Playbook