How B2B sales teams actually use AI for lead generation (and whether it’s working)

Every AI enthusiast or vendor pitch deck says the same thing: AI is transforming… literally anything. B2B outreach and lead generation aren’t an exception. Outreach writes itself. Contact lists build themselves. SDRs become obsolete. The promise is seductive, and the spending follows it.

But we wanted to know what was actually happening inside revenue teams, not what was being sold to them. So we surveyed 120 B2B sales professionals across industries, company sizes, and deal sizes, and asked them how they really use AI for lead generation.

Ninety-six of those respondents had moved past pure manual prospecting and were actively using AI in some form. Their answers are the foundation of this report. What emerged is a picture that is more interesting, and considerably more useful, than either the hype cycle or the backlash against it: a market in the early innings, using AI as a capable assistant rather than an autonomous operator, getting real but modest results, and still largely unable to say whether any of it is paying off.

Who we talked to about using AI for lead generation

Before the findings, a word on who answered. The sample skews toward smaller, tech-adjacent organizations: 57.5% of respondents work at companies with 10 or fewer employees, and three industries — Marketing & Advertising, IT & Services, and SaaS — account for 58.3% of the total sample. Enterprise respondents (500+ employees) number only five.

Why are we flagging this at the very beginning? Because small, digitally native teams tend to adopt new tools faster and with fewer layers of approval than large enterprises. The findings below should be read as a strong signal from the leading edge of adoption — lean, tech-forward sales organizations — rather than a definitive statement about how a Fortune 500 sales floor uses AI. We flag every place this skew is likely to matter.

AI adoption in B2B sales is still at day one for most companies

The single most important fact to hold onto while reading the rest of this report: most teams using AI for lead generation have been doing it for less than six months.

Among the 96 respondents who passed our screening question (meaning they use AI for lead gen in some capacity), 53% adopted it within the last six months, and another 30% have been at it for six to twelve months. Only 16.7% have more than a year of experience, and just five respondents (5.2% of the sample) have used AI for lead generation for two years or longer.

That timing context should color every other statistic in this report. When 53% of respondents say their results have “improved slightly” rather than “improved significantly,” that is not necessarily a verdict on what AI can do for lead generation. It may simply be a snapshot of teams still climbing the learning curve, still refining prompts and workflows, and still working out where AI earns its keep.

There is a tempting counter-read here: the small group of long-tenured users (2+ years, n=5) reported meaningfully stronger engagement gains than everyone else — an average improvement score of 2.40 on our 0-to-3 scale, versus 1.07 to 1.15 for every shorter tenure bracket. That is a striking gap. It is also a five-person group, far too small to treat as proof that results compound with time. We mention it because it is suggestive, not because it is settled.

Practical takeaway: if your team is six months into AI-assisted prospecting and the results feel underwhelming, you are not behind — you are typical. The market has not yet had enough time to separate the tools and habits that compound from the ones that plateau.

AI is a co-pilot, not an autopilot

The dominant story in our data is augmentation. Across all thirteen execution tasks we asked about, from list building to discovery call prep, “AI-assisted with human review” was the most common answer for ten of them. Fully autonomous, no-human-review AI use never exceeded 20% adoption on any single task.

Data enrichment and contact finding came closest to full automation, with 20% of teams letting AI run that task without review — understandable, since enrichment is a structured, low-stakes lookup task with little room for the AI to embarrass anyone. Lead list building followed close behind at 19% fully automated.

The picture flips entirely for synchronous, human-facing work. Cold calling and phone prospecting remain human-only for 55% of teams, with just 8% trusting AI to run that task without a human present. Objection handling guidance is even more human-anchored: 57% human-only, and only 5% fully automated. Meeting scheduling coordination sits similarly, at 53% human-only.

The pattern makes intuitive sense once you see it laid out: the tasks where teams trust AI to act alone are the structured, asynchronous, low-stakes ones. The tasks where humans stay firmly in control are the ones involving live conversation, persuasion, and judgment calls — exactly the skills that are hardest to automate and riskiest to get wrong in front of a prospect.

This complicates the narrative, popular in some corners of the AI conversation, that SDRs are being systematically replaced. What the data shows instead is a division of labor: AI handles volume and structure, humans handle conversation and nuance. Whether that division holds as the technology matures is an open question — our data describes today, not five years from now.

Practical takeaway: if you are evaluating AI tools for lead generation, the realistic near-term goal is not to eliminate headcount for synchronous tasks. It is freeing capacity on structured, repeatable tasks so your team has more time for the calls and conversations that still need a human.

Beyond execution: AI is quietly becoming a strategy tool

Lead generation conversations tend to focus on execution — writing emails, scoring leads, finding contacts. Our data suggests teams are pushing AI further upstream than that framing implies.

When we asked which strategic use cases teams apply AI to, beyond the execution tasks already covered, defining or refining the ideal customer profile (ICP) topped the list at 46.9% of respondents — ahead of account prioritization (24.0%), buying-intent detection (20.8%), and even real-time sales coaching (21.9%). Analyzing performance and recommending strategy changes came in second at 38.5%.

This is a meaningful finding because ICP definition sits upstream of essentially everything else a revenue team does. Get the ICP wrong, and better lead scoring, better copywriting, and better outreach timing are all optimizing against the wrong target. The fact that nearly half of respondents are using AI here — rather than only at the execution layer — suggests teams increasingly see AI as a thinking partner for strategy, not just a production tool for tasks.

A caveat worth naming: our question did not distinguish between a quick ChatGPT session to brainstorm an ICP and a dedicated account-intelligence platform running continuous analysis. “Using AI for ICP work” covers a wide range of sophistication, and this finding should be read as evidence of intent and direction, not depth.

Practical takeaway: if your AI strategy so far is purely about execution — writing better emails, scoring leads faster — you may be underusing the technology relative to where the market is already heading. Many teams in our sample are already using AI to help answer “Who should we be targeting,” not just “How do we reach them faster.”

📚 Relevant reading: Who are you actually selling to in B2B? The 2026 buying committee study

AI’s impact on sales performance: Real gains, but mostly modest, with volume beats quality

Ask teams whether AI has improved their lead generation results, and most say yes — but “yes, slightly” is the answer that shows up over and over again. Across five performance metrics — database quality, engagement rate, lead-to-opportunity rate, deal-closing rate, and SDR daily outreach volume — “improved slightly” was the single most common response for every one of them. Significant improvement, by contrast, was reported by only 10% to 17% of respondents depending on the metric.

Two things stand out in that table. First, database quality shows the strongest improvement signal in the entire dataset — 74% of teams report some level of improvement, with 17% calling it significant. Of every outcome we measured, this is the one AI appears to deliver most reliably, which tracks with how the technology is actually being used: enrichment and contact-finding are exactly the tasks teams trust AI to run with the least supervision.

Second, and more telling: deal-closing rate is the weakest performer on the list, with 29% of teams reporting no change at all. Compare that to SDR outreach volume, which improved for 66% of teams — the strongest volume-side metric we measured.

Put together, this is the clearest paradox in the dataset. AI is very good at making sales teams busier — more contacts found, more lists built, more outreach sent. It has not yet shown the same strength at making that activity convert into closed revenue. That is not necessarily a flaw in the technology; it may simply reflect that closing a deal depends on dozens of variables AI cannot influence, from pricing to product fit to a prospect’s internal budget cycle. But it is a gap between the marketing promise of AI (“more revenue”) and the lived reality (“more activity”) that deserves honest acknowledgment.

Practical takeaway: if you are adopting AI primarily to move the needle on closed revenue, set expectations accordingly. The strongest, most reliable wins in this data are upstream — list quality and outreach volume — not downstream at the close.

More mature AI programs perform better (but they are not happier)

One of the more counterintuitive findings in this research: teams with more advanced AI adoption get measurably better results, but they are no more satisfied than teams just getting started.

We grouped respondents by their self-rated AI maturity, from “testing, minimal adoption” (level 2) through “fully AI-native” (level 5), and compared their average improvement in engagement rate against their overall satisfaction score.

The trend appears meaningful: more mature teams report close to double the engagement-rate improvement than the teams who are still in the testing phase. But satisfaction barely moves — it hovers in a narrow band between 3.2 and 3.4 out of 5 regardless of maturity level. The statistical relationship between maturity and engagement gains is moderate and meaningful; the relationship between maturity and satisfaction is essentially nonexistent.

Why would better results not translate into higher satisfaction? A few explanations can take place, and our data cannot fully adjudicate between them. Teams that go further with AI may also run into more friction along the way — more integration headaches, more cost overruns, more time spent training the system — that offsets the performance win in how they feel about the overall experience. It is also possible that more mature teams set higher internal benchmarks, so a real improvement still feels like falling short of what they expected. Either way, the lesson is the same: results and satisfaction are not the same thing, and a vendor or internal champion who only tracks the former is missing half the picture.

Practical takeaway: if you are scaling up your AI footprint, budget for change management and expectation-setting alongside the technology itself. Performance gains alone will not guarantee a happier team.

Does AI have a positive or negative impact on ROI when used for lead generation?

Here is a number that should give every sales and RevOps leader pause: 77% of teams using AI for lead generation cannot say whether it has delivered a positive return on investment (ROI). Forty-five percent simply have not measured it. Another 32% say it is too early to tell. Only 18.8% have confirmed positive return on investment, and a small 4.2% have confirmed it’s negative.

Some of this is defensible — with 53% of the sample under six months into adoption, “too early to tell” is a rational answer, not an evasive one. But “haven’t measured” is a different category entirely, and at 45% of the sample, it represents the single largest group in this question. That is a measurement gap, not a performance gap: the absence of an ROI figure tells you nothing about whether the investment is working. It tells you the organization has not built the instrumentation to find out.

This gap matters because of what we found next. Among the variables we tested — AI maturity, adoption duration, performance improvement on any individual metric — the single strongest predictor of overall satisfaction was simply having confirmed positive ROI. Respondents who could say “yes, this is paying off” reported average satisfaction of 4.11 out of 5. Everyone else — unmeasured, too-early, or negative — clustered around 3.0.

Taken together, these two findings suggest the highest-leverage move available to most teams in this market is not a better AI tool — it is a better measurement framework. Teams that can actually see and confirm ROI are dramatically more satisfied than teams operating on faith. Given that satisfaction did not track with AI maturity in the previous section, this may be the more reliable lever: it is easier to build a tracking dashboard than to wait years for a maturity level to compound into a feeling of success.

Practical takeaway: before adding more AI tools or expanding usage, invest in the ability to measure what the tools already in place are doing. The data suggests that confirmed proof of value, not adoption depth, drives genuine satisfaction.

What vendors promise vs. what teams actually experience when adopting AI

Nine out of ten respondents (90.6%) said at least one common AI vendor claim has not matched their real-world experience. Only 9.4% said every claim they encountered held up.

What is more interesting is which specific claim disappointed people most. It was not the dramatic, headline-grabbing promises — “Will eliminate the need for the SDR team” (35.4% called this unmet) or “Will write outreach indistinguishable from humans” (27.1%). The single most-cited unmet claim, at 43.8%, was something far more mundane: “requires no training or setup time.”

That ordering tells a story. The disappointment in this market is less about whether AI can perform and more about how hard it is to get it performing. Setup friction — the unglamorous work of integration, configuration, and training — is the gap between the demo and the daily reality more often than any failure of the AI’s actual output.

This reading is reinforced by what teams cited as their biggest implementation challenges: AI output quality being “Generic, obvious, or robotic” topped the list at 37.5%, followed by poor data quality (31.3%), higher-than-expected costs (26.0%), and longer setup time than anticipated (24.0%). Output quality and setup friction together account for the lion’s share of frustration — not a fundamental failure of the technology, but a mismatch between how it was sold and how it actually has to be implemented.

Practical takeaway: when evaluating an AI vendor, discount promises of zero setup time on sight. Ask instead about training requirements, integration timelines, and what “good output” looks like after the honeymoon period — those are the areas where real teams report the real gap.

How AI adoption in lead gen varies by company size and industry

Small companies (1 to 10 employees) dominate this sample, and within that group, AI maturity skews slightly earlier-stage than among companies with 1 to 50 employees — 73% sit at the “testing” or “specific tasks” maturity levels, versus 57% for the next size bracket up. Beyond that, small companies do not look meaningfully different from the rest of the sample: their share of confirmed positive ROI respondents roughly matches their share of the overall sample, and satisfaction is statistically indistinguishable from larger companies.

We cannot say the same with confidence for larger organizations. Companies with 200 or more employees number only 18 in our screened sample, and those with 500 or more number just five — too small a group to support reliable conclusions about how enterprise sales organizations specifically experience AI adoption. The same limitation applies to industry-level analysis: outside of Marketing & Advertising, which had enough respondents (32) for a basic check and showed no meaningful deviation from the overall pattern, every other industry in the sample has fewer than ten respondents.

We also found no meaningful relationship between company size and AI maturity (the correlation was effectively zero) — larger companies in this sample are not further along in their AI adoption than smaller ones, which runs against the common assumption that bigger budgets translate into faster adoption. Given how few large companies are represented, this is best read as an open question rather than a settled finding.

Will AI replace SDRs? What B2B sales teams say about their plans

Roughly a third of respondents say they are planning to eventually replace their SDR team with AI. That statistic, taken alone, would make a dramatic headline. Examined alongside the rest of the data, it looks more like ambition running ahead of evidence.

Of the 30 respondents who selected “definitely expanding — planning to replace our SDRs with AI,” only 17% describe themselves as very satisfied with their current AI implementation. Twelve are neutral, and four are outright dissatisfied. Sixty percent of this same group have been using AI for lead generation for less than six months.

In other words, the people most bullish on full SDR replacement are disproportionately the newest, least-proven users in the sample — not the seasoned, highly satisfied power users you might expect to be making that call. We should also flag a known data quality issue here: the survey question this comes from behaved as a multi-select when it was designed as single-select, meaning some respondents who chose “replace our SDRs” may have simultaneously selected contradictory options elsewhere in the same question. This figure should be read as a signal of ambition and cost pressure circulating in the market, not as a reliable forecast that a third of B2B sales teams are on a path to all-AI prospecting.

Practical takeaway: treat “we’re replacing our SDRs with AI” as a strategic aspiration under active testing, not a proven playbook. The evidence so far suggests it is being driven more by cost pressure and early enthusiasm than by demonstrated results.

What this all adds up to

Step back from the individual findings, and a coherent picture emerges: B2B sales teams are not living through the transformation the loudest vendor pitches describe, nor are they experiencing AI as a complete disappointment. They are in the early, uneven middle of adopting a genuinely capable tool.

A few conclusions are strong enough to state with confidence. This is an early-stage market, with most teams under a year into adoption. The dominant mode is augmentation — AI handles structured, repeatable work while humans keep control of conversation and judgment. Results are real but modest, concentrated in volume and data quality rather than conversion. And the biggest open problem isn’t the technology’s capability; it’s that most teams haven’t built the measurement discipline to know if their investment is paying off, even though doing so is the strongest observed relationship.

Other patterns — the maturity-performance link, the SDR-replacement ambitions, the small high-tenure cohort’s stronger results — are worth watching but too thin to build a strategy on. The teams most likely to be satisfied with their AI investment a year from now probably aren’t the ones chasing the boldest automation claims. They’re the ones building unglamorous infrastructure: clear use cases, realistic timelines, and a real way to measure whether it’s working.

Methodology note: This report is based on a survey of 120 B2B sales professionals conducted between May 2–June 22, 2026. Ninety-six respondents passed an initial screening question confirming active use of AI for lead generation and answered the full questionnaire. All findings are descriptive and observational; no causal claims are made. Subgroup analyses based on fewer than 20 respondents are explicitly flagged as directional throughout this report.