Blog · ai
How RevOps teams use AI, and why few say it added pipeline
RevOps teams use AI for research, call notes and CRM updates. In surveys, 70% report returns but under 9% report more pipeline. Judge it against a baseline.
RevOps teams use AI as a layer around the CRM
Day to day, RevOps teams use AI as a layer around the CRM. That is the picture in all 5 AI answers we captured on 4 October 2026, from ChatGPT, Google’s AI Overview and AI Mode, Gemini and Perplexity.↗ ChatGPT and Perplexity both use the phrase “operational layer around the CRM”. Sorted by where they sit in the sales process, the jobs they list are:
- Before a call: account research and a brief built from CRM records, emails and past calls
- Routing: inbound leads scored against the ideal customer profile and sent to the right rep
- After a call: transcripts, summaries, next steps and suggested CRM updates
- Follow-up: drafted recap emails and tasks for the rep to check
- Pipeline: stalled deals, missing next steps and slipped close dates flagged for the manager
- Hygiene: duplicates, empty fields and enrichment gaps found and fixed
- Forecast: a second opinion on the rep’s number, built from activity and call data
What no answer says is whether any of this works. The answers describe use cases, not results, and they draw them mostly from people who sell the tools or the services around them.
Of the 45 citation links in those answers, 24 go to software vendors’ own pages, 13 to consultancies or agencies and 8 to videos, social posts or community pages.↗ None goes to independent research.↗ Gemini takes 6 of its 10 citations from a single consultancy’s blog post.↗ That post’s claim that automated account research “saves 15–30 minutes per account” comes with no source.↗
Why do surveys put AI use anywhere from 25% to 99%?
Because each survey asks a different question of different people. The answers lean on surveys from 6sense, HockeyStack, Default and Revenue Wizards, all from 2026, and each publisher sells to the people it surveyed.
6sense finds that 99% of business development reps (BDRs) report using AI, up from 62% in 2025.↗ Its release gives no sample size. In Revenue Wizards’ survey of 26 CROs and senior revenue leaders, 73% “operate past experimentation”.↗
In HockeyStack’s survey of 315 GTM staff, half have rolled AI out widely and 41% are running pilots.↗ Default asked a stricter question: 25% of its 300+ RevOps leaders are integrating AI across all go-to-market functions.↗ Only 4% call their organisations “highly AI-driven”.↗
Survey figures for AI use run from 25% to 99%, depending on what the question counts.
Data behind this chart
| Item | Value |
|---|---|
| Report using AI | 99% |
| Past experimentation | 73% |
| Rolled out widely | 50% |
| Integrating across all GTM functions | 25% |
Using AI at all, rolling it out widely and running it across every function are different claims. A BDR who drafts the odd email with a chatbot would count in the first. Only a team with AI at work in every go-to-market function counts in the last.
The samples differ too. HockeyStack’s respondents are 87% from software companies, and 70% work at firms with 1,000+ employees.↗ Default surveyed leaders at fast-growing companies, “from seed-stage startups to public software companies”.↗ Neither report splits its results by company size or deal size.
| Survey | Publisher sells | Who answered | Headline figures |
|---|---|---|---|
| 6sense, 2026 State of BDR Report (April 2026) | A revenue intelligence platform | BDRs; the release gives no sample size | Use: 99% of BDRs report using AI · Results: none on AI in the release |
| HockeyStack, State of AI in RevOps (fielded January–February 2026) | A revenue data and attribution platform | 315 GTM staff; 70% at firms with 1,000+ employees, 87% in software | Use: half have rolled AI out widely · Results: 70% report positive or strong returns |
| Default, The State of AI in Revenue Operations (H1 2026) | Workflow, enrichment and lead-routing software | 300+ RevOps leaders at fast-growing companies, from seed stage to public software companies | Use: 25% integrate AI across all GTM functions · Results: under 9% say AI helped generate more pipeline |
| Revenue Wizards, CRO AI Adoption Survey (2026) | RevOps consulting | 26 CROs and senior revenue leaders | Use: 73% past experimentation · Results: 46% name revenue gains their primary benefit |
70% report returns, but fewer than 9% report more pipeline
Both figures can be true, because they answer different questions. HockeyStack asked about returns on AI investment: 70% of respondents report positive or strong returns, and most of them (65%) put the return at 1–2×.↗ Yet average satisfaction with current AI tools is 3.9 out of 7.↗
Default asked what had measurably improved. Fewer than 9% say AI has helped generate more pipeline, and 7% have seen improved conversion.↗ More of them, 11%, report faster lead routing and follow-up.↗
Default’s report heads one section “AI saves a few hours a week for teams”, and most of its respondents save under 5 hours a week.↗ Another section is headed “AI is helping teams work faster, not grow faster”.↗
The survey that puts revenue first is the smallest. Revenue Wizards reports that 46% of its respondents name revenue gains as their primary benefit.↗ With 26 respondents, that is about 12 people.↗
Asked about results, 70% report returns on AI but fewer than 9% report more pipeline.
Data behind this chart
| Item | Value |
|---|---|
| Positive or strong returns | 70% |
| Revenue gains the primary benefit | 46% |
| Faster lead routing and follow-up | 11% |
| More pipeline | 9% |
| Improved conversion | 7% |
Read together, the larger surveys describe a modest payback that shows up as hours saved more often than as pipeline. They also suggest why the pipeline effect is hard to show.
Data quality and integration is the top AI challenge for 61% of HockeyStack’s respondents.↗ In Default’s survey, 19% name poor data quality as the biggest blocker, and nearly 1 in 4 organisations say AI adoption has no clear owner.↗
Every figure here is also a survey answer. None is a before-and-after measurement. If nobody owns the rollout and the fields are unreliable, there is nothing solid to measure against. We asked whether CRM data is clean enough for AI recommendations in an earlier post.
Where do RevOps teams still keep a person in charge?
RevOps teams keep people in charge of pricing, forecasts and strategy. In HockeyStack’s open answers, respondents consistently flagged pricing and deal desk decisions, forecast calls, context-dependent analysis and strategic decisions as off-limits for AI.↗
The AI answers disagree on this point. Google’s AI Overview says autonomous agents “trigger outreach sequences, and update pipeline stages without constant human oversight”, citing 2 software vendors’ blogs.↗ Google’s AI Mode has agents reviewing discount requests against policy, close to the deal desk decisions HockeyStack’s respondents keep for people.
ChatGPT leaves the rep to validate what AI pulls from a call and to review and send the follow-up. Perplexity says CRM updates often wait for a person’s approval before they sync.
The surveys draw the line more clearly than the answers do: pricing, forecasts and strategy stay with people. We found the same split when we compared AI tools for account research: AI drafts, a person checks.
Let AI do the associate work, and judge it against a baseline
We think AI earns its place in RevOps doing the associate work, with a person approving every first touch. We also think its effect on pipeline can only be judged against a baseline taken before rollout.
Associate work is what AI does well and a person can check fast. Account research, call notes, CRM updates and first drafts fit that description. It is also how Panelhop runs: agents do the research, drafts, analysis and reports, and a person approves everything before it goes out.
What to say to a specific buyer, and accountability for it, stays with a person. That is why we don’t build or run anything that sends a first touch without someone reviewing it. Review can be quick, as long as a person makes the call. That line also keeps the work defensible under outreach law and platform terms.
Without a baseline, almost any number can be framed as a win. “Positive returns” in a survey is an impression. Pipeline created per quarter, measured the same way before and after rollout, is a comparison. A baseline can’t separate AI from everything else that changed in the business, but we think it is the minimum bar for a credible claim.
The baseline is only as good as the fields it reads. In Validity’s 2025 survey, 76% of CRM users say less than half of their CRM data is accurate and complete.↗ If stage dates, sources and owners are patchy, before and after can’t be compared, and AI writing into those fields at scale spreads the errors faster. In our view the fields don’t need to be perfect, only the ones the rollout and its measurement depend on, the same rule as for cleaning up a CRM before lead scoring.
In practice it comes down to 4 steps, 3 of them before anything goes live.
- 01Step 1
Name the after-numbers
Pick the few numbers the rollout should move, such as pipeline created per quarter, stage conversion and speed-to-lead. Count hours saved too, as a separate result.
- 02Step 2
Fix the fields they read
Check the stage dates, sources, owners and amounts behind those numbers. Clean those fields first; the rest of the CRM can wait.
- 03Step 3
Freeze the baseline
Before go-live, record each number with its definition, date range, data source and sample size. Log any later change to a definition.
- 04Step 4
Report against it
After rollout, report the same numbers the same way, next to the baseline, every month or quarter.
In practice
How we do it at Panelhop
Every Panel Check (GTM audit · 2–3 weeks) ends with a baseline sheet: 7 metrics, among them qualified pipeline from target accounts, speed-to-lead, CRM hygiene and forecast accuracy, each recorded with its definition, date, data source and sample size. The client signs it off and we freeze it, so everything built afterwards, AI included, is reported against the same numbers. The metrics and their definitions are on the method page.
We run on the split this post argues for. In Signal Desk (in-market accounts, weekly), agents score the accounts, map the buying group and draft a brief for each. A person checks the batch before it reaches your CRM, and your rep owns the first touch.
Questions buyers ask about this
Is AI replacing the CRM?
Not in how teams describe their work today. ChatGPT and Perplexity both call AI an operational layer around the CRM, and the jobs the answers list, from account research to deal-risk flags, read from it or write to it. In our view, that makes the CRM’s data quality the limit on what the AI can do.
Does AI in RevOps pay back in time or in pipeline?
So far, mostly in time. In Default’s survey of 300+ RevOps leaders, most respondents save under 5 hours a week, while fewer than 9% say AI has helped generate more pipeline. We’d count the hours, but report them separately from pipeline.
How do we show whether AI added pipeline?
Take a baseline before rollout: pipeline created per quarter, stage conversion and speed-to-lead, each with a fixed definition and data source. Check that the fields behind those numbers are filled and reliable before you switch anything on. Then report the same numbers the same way after rollout, next to the baseline.
Why do surveys on AI in sales disagree so much?
They ask different people different questions. 6sense counts BDRs who report using AI at all (99%), while Default counts teams integrating it across every go-to-market function (25%). Each publisher also sells to the people it surveys, so read the question and the sample before the headline.
Should AI send outreach without a person reviewing it?
We don’t build or run that. We think every first touch to a prospect needs a person’s decision, however good the draft. The review can be quick, as long as a person makes the call.
