Most RevOps problems are not exotic. The same six mistakes account for most of the pain in most go-to-market teams, and they’re worth naming precisely because AI changes the stakes: some of these get dramatically cheaper to fix, and some get dramatically more expensive to ignore, because automation built on a broken foundation produces wrong answers at scale.
1. Buying tools before defining the process
The classic sequence: a conversion problem appears, a vendor demo looks great, the tool gets bought, and six months later it’s shelfware with a renewal date — because nobody defined what the process should be before automating it. The tell is a stack audit that finds two enrichment vendors, three scheduling tools, and an ABM platform nobody logs into.
The AI era makes this worse before it makes it better, because the category is currently flooded with tools that demo beautifully. The discipline is unchanged: write down the workflow on one page — trigger, steps, owner, output — and only then decide what executes it. Increasingly the answer is not a new subscription but a pipeline you already have the pieces for: an automation platform, an LLM API, and a scraper. That’s most of what we reach for first; the specific actors we use are on our tools page.
2. Treating the CRM as a junk drawer
Duplicate accounts, freeform fields, leads imported from a 2023 conference with no source, seventeen deal stages nobody can define. Everyone knows this is bad; teams live with it anyway because cleanup never makes the quarter’s priority list.
Here’s why it can’t wait anymore: dirty data used to cost you reporting accuracy, but now it poisons everything downstream. Score a lead with an LLM against a CRM full of duplicates and the model confidently scores the duplicate. AI raises the return on clean data and the penalty for dirty data at the same time. The good news is that models are also decent at the cleanup itself — deduplication, field normalization, and note summarization are exactly the kind of tedious classification work they do well. Run the cleanup pipeline first, then the clever stuff.
3. Routing leads slowly
Speed-to-lead is the most measurable number in RevOps and still one of the most neglected. Leads that sit overnight because the routing rules didn’t cover a territory, or because the one person who reassigns them was on PTO, are deals lost to whoever answered first.
This is the mistake AI fixes most cleanly. A model can read the form fill and the company’s website, qualify against your rubric, route, and book — in seconds, at 2 a.m., with its reasoning logged. If you automate one thing this quarter, automate this. The catch is monitoring: an unwatched routing agent that starts misclassifying doesn’t back up visibly like a human queue does. It fails silently, which is why the job shifts to auditing — more on that in what a RevOps engineer actually does.
4. Reporting nobody trusts
If sales and marketing bring different numbers to the same meeting, the meeting is about the numbers instead of the business. Usually the cause is definitional: two dashboards counting different dates, or an attribution model chosen to flatter a channel.
AI does not fix this one, and pretending it does is how the mistake compounds. A model asked to explain pipeline from inconsistent data will produce a fluent, confident, wrong narrative — which is more dangerous than an obviously broken dashboard, because people believe it. The fix is old-fashioned: one set of definitions, written down, owned by RevOps, agreed to by the people who will be measured by them. Automate the reporting only after the definitions hold.
5. Treating RevOps as a ticket queue
When RevOps exists only to answer field requests and pull lists, the strategic work — process design, data modeling, the definitions from mistake #4 — never happens, and the team burns out doing service work that never shrinks.
The AI angle is straightforwardly good here: the ticket-shaped work (list pulls, field changes, “can you check why this lead didn’t route”) is increasingly self-serviceable or automatable, which frees the team for the work that compounds. But that only happens if leadership resets expectations along with the tooling. If the queue stays the job description, the pipelines just make the queue move faster.
6. Automating a broken process
The most expensive mistake on this list, and the most tempting one right now. Automation is an amplifier: point it at a good process and you scale the good; point it at a broken one and you industrialize the damage. An AI sequence enthusiastically emailing a dirty list doesn’t fix your targeting — it burns your domain reputation at machine speed.
The test before automating anything: does the manual version work? Not “could it work” — does it, today, when a human does it? If yes, automate and measure. If no, fix the process first, run it manually until the numbers behave, then hand it to the machine.
The pattern across all six: AI moves the bottleneck, it doesn’t remove it. The teams getting real results — the kind that show up in where RevOps is heading — are the ones that fixed foundations first and automated second. It’s less exciting than the demo. It’s also the only version that survives contact with the quarter.
If you’re still mapping the basics, start with what AI RevOps is; if you’re weighing a first ops hire against a leaner setup, see when to hire RevOps.