Executive summary

B2B go-to-market has been quietly restructured over the last 24 months, and the 2026 data is now clear enough to call it. The markets that sell tools for outbound pipeline are growing in single digits. The markets that sell intelligence, research and signal are growing in double digits, some above 20% annually. Meanwhile, the category that defined 2023 and 2024, AI SDRs, has visibly collapsed as a reliable growth strategy.

At the same time, the ground beneath outbound has shifted. Buyers spend three-quarters of their journey off-platform, invisible to attribution tools. In 2026 that invisibility got deeper because nearly all B2B buyers now interrogate large language models like ChatGPT, Perplexity and Claude during their purchase journey. Cold email reply rates have halved in three years. LinkedIn DMs are ignored by four in five decision-makers. The old engine has not just slowed down. It is running at a loss.

What has replaced it is harder to see because it does not come in a neat SaaS category. It is the investment shift into research, intent data, competitive intelligence and buyer enablement. It is the rise of a new role, the GTM engineer, who looks less like an SDR and more like a data operator. It is a quiet pivot back toward fewer, better conversations, triggered by real signal and informed by real research.

This report synthesises the 2026 public data, from 360iResearch, Fortune Business Insights, Roots Analysis, 6sense, Instantly and others, with our own observations across 50+ client engagements at ORRJO. The goal is not to pile on with more predictions. The goal is to explain what has actually happened, why the shift is permanent, and what it means for how B2B companies should build pipeline in the next 12 months.

Section 01The market for B2B intelligence is growing faster than the market for pipeline tools.

If you want a single number to understand what is happening to GTM in 2026, start with this. The global B2B market research industry is sized at $43.9 billion in 2026 and is projected to grow to $78.5 billion by 2032, at an 8.61% compound annual growth rate [360iResearch]. That is the floor, not the ceiling, because the tools layer that sits on top is growing much faster.

The fastest-growing sub-category in GTM is competitive intelligence tools. The market sits at $0.87 billion in 2026 and is forecast to reach $4.03 billion by 2034, a 21.17% CAGR [Fortune Business Insights]. Buyer intent data tools, the layer that tries to tell you which accounts are researching your category right now, is growing from $4.5 billion today to a projected $20.9 billion by 2035, at 16.62% CAGR [Roots Analysis]. And the broader B2B information services market sits at around $156 billion, growing at nearly 15% annually.

Compare that to the growth rate of sales engagement platforms (the tools category) which is running at closer to 10% CAGR, and you start to see where the investment is going. The market is voting with its wallet. The research layer is being repriced upward. The execution tool layer is being compressed, because execution without research has stopped working.

This is not a rotation into a fashionable category. It is a structural response to outbound's declining economics. When reply rates halve and meeting attendance drops, the cost to book a meeting goes up. The only way to pull that cost back down is to improve the quality of the upstream targeting. That is a research problem, not a dialling problem.

Section 02The AI SDR collapse and what it means.

Two years ago the AI SDR was the category that was supposed to eat outbound. Platforms like 11x, AiSDR, Artisan and dozens of others raised substantial rounds on the promise of fully autonomous pipeline. By 2026 the category had crossed $5.8 billion in combined spend, but the performance numbers have quietly diverged from the marketing.

The cleanest public data point: only 2% of companies report successful implementation of AI SDRs [Gartner 2026]. Churn on the platforms is running between 50% and 70% inside the first three months, multiples of what you would see in healthy SaaS, where 5 to 10% is typical. 11x, the category's most visible brand, was criticised in public forums as "incredibly underwhelming" in eleven separate high-engagement threads on Hacker News and Reddit in the first quarter of 2026 alone [HN discussions]. One founder documented receiving zero replies on 1,400 emails across three weeks of an 11x campaign.

"We gave it a quarter. Fourteen hundred emails. Zero replies. Zero meetings. The promise was pipeline. What we got was a very expensive way to damage our domain reputation."

This is not about one vendor. The category has a structural problem. AI SDRs were sold as a way to do more of what worked in 2022, faster. The issue is that what worked in 2022 stopped working in 2023, and it has been getting steadily worse since. The AI layer did not fix the upstream problem, it amplified it. More volume into a broken motion is more broken motion.

There are three root causes worth naming. First, no research upstream. Most AI SDR tools scrape LinkedIn or Apollo and write a generic opener based on a job title. That was possible to get away with when reply rates were 6%. At 3.43% it is not. Second, quality degrades at volume. The models optimise for sounding fluent, not for being relevant. Fluency without relevance is the worst possible combination because it triggers pattern recognition in the reader. Third, buyers pattern-match AI instantly. After eighteen months of exposure to GPT-generated outreach, senior buyers can identify AI-written messages in less than two seconds. The cost is not just a lost reply. It is brand damage and domain reputation damage.

The counter-trend is already visible. Research-led outbound is coming back. Teams are shrinking volume by 10x, investing the saved time into primary research, and sending fewer but dramatically more relevant messages. The math on that trade is the defining commercial question of 2026.

Section 03The dark funnel is getting darker.

The dark funnel is not new. Forrester has been writing about it since 2020, and the rough consensus for the last four years has been that 73% of the B2B buying journey is anonymous, happening off-platform and invisible to traditional attribution. What is new in 2026 is that the dark funnel got measurably darker, because buyers now route a meaningful portion of their research through large language models.

The data: 94% of B2B buyers now use LLMs like ChatGPT, Perplexity and Claude somewhere in their purchase journey [6sense 2025 report]. A separate SEMrush study counted 89,000 unique LinkedIn URLs cited in AI-generated responses across business questions, suggesting that LLMs are actively indexing and recommending specific people and companies during purchase research. The conversations themselves are invisible. Intent data tools, which rely on scraping observable activity across IP ranges, cannot see into ChatGPT conversations.

This is a strategic shift. In the observable funnel, outbound could at least trigger an account. In the dark funnel, buyers evaluate you without you knowing they exist. In the AI dark funnel, a model summarises who you are and who your competitors are based on what it has indexed. You do not get a seat at that table unless you are already visible in it.

What gets you visible inside the dark funnel is not ICP-targeted ads. It is brand, thought leadership and original research. The research reports Perplexity and ChatGPT cite are not white papers written to capture email addresses. They are data-backed pieces of primary analysis that both humans and machines treat as authoritative. That is why our strong recommendation to B2B founders right now is to invest in dark funnel visibility at least as seriously as you invest in paid performance.

The corollary is harsh for anyone still running pure ICP-targeted ad budgets. The effectiveness of that motion is decaying faster than ever because the buyer is not on the platform where the ad runs. The budget line that was quietly generating 20% of your pipeline two years ago is now generating 8 or 9%. The accounting rarely catches up to the reality until the next planning cycle.

Section 04Cold outbound is halving in effectiveness every three years.

There is a chart that explains more about the current state of GTM than any single datapoint: the cold email reply rate over time. In 2023, Instantly's benchmarking dataset put the average B2B cold email reply rate at 6.8%. In 2026, the same dataset puts it at 3.43% [Instantly benchmarks 2026]. A halving inside three years. The LinkedIn picture is similar. 79% of B2B decision-makers report ignoring cold DMs outright, according to a 2025 LinkedIn survey [LinkedIn B2B Insights].

There is a meaningful counter-datapoint though. Inbound-led outbound, where the outreach is triggered by a first-party signal (content download, pricing page visit, product activity), converts at 14.6%, versus pure outbound's 1.7% [Salesloft 2026]. That is an 8x delta. Same tech stack, same SDRs, same scripts. The only variable that changed is that the target raised their hand first.

Do the math. If your cold reply rate is 3.4% and your inbound-led outbound reply rate is 14.6%, every hour of SDR time is four times more productive in the second motion. You do not need more outbound capacity. You need more signal. You need more research. You need enough intelligence to know which accounts are in a buying moment, and enough depth to say something relevant when you arrive.

This is why 2026 is going to be remembered as the year the industry stopped arguing about outbound versus inbound and started talking about signal-triggered, research-led motion. It is not a new channel. It is the combination of first-party signal (from your site, your product, your content), third-party signal (intent data, hiring data, funding data) and manual research into the account and buying committee.

Section 05ICP activation is the biggest operational gap in B2B.

If you ask a B2B founder whether they know their ICP, almost all of them will say yes. If you then ask them to show you the document, fewer than half can find it. And if you ask them how the ICP is used operationally, day to day, by marketing, sales and product, the number drops again.

The public benchmark: 42% of B2B companies have a formally documented ICP [Gartner ICP research]. Our experience across 50+ engagements suggests that fewer than half of those 42% actually activate the ICP. Activation means the sales team qualifies against it, the marketing team writes for it, the paid team targets to it, and the product team prioritises features for it. In most companies at least one of those is broken.

The most common failure pattern is what we call the "workshop ICP". The company runs a three-hour offsite, produces a Notion page, congratulates itself, and never refreshes the document. Six months later the ICP on paper does not match the customers that are actually closing, nor the prospects the team is actually pursuing. The document becomes decorative.

The cost of a fuzzy ICP is not subtle. It shows up as diluted messaging, because you are trying to speak to three personas at once. As wasted pipeline, because 60% of your outbound is going to accounts that will never buy. As wrong-fit customers, who close but churn inside a year. And as slow GTM experimentation, because without a clear ICP you cannot isolate what is working.

The remedy is boring and it works. You pull the data on your last 100 closes, you look at where they cluster by size, sector, tech stack, trigger event and buying committee composition, and you rebuild the ICP from the closes, not from the workshop. You refresh it quarterly. You put it on a dashboard. You reject leads that do not match. Most teams know this. Almost none of them do it consistently. That is the gap ORRJO's ICP research engagements are built to close.

Section 06The rise of the GTM engineer.

The role of "GTM engineer" barely existed in 2022. By spring 2026 there were 400+ open postings carrying the title across LinkedIn and specialist boards, with postings up around 205% year on year and a median base salary of $160K, roughly 20% above comparable ops roles [GTM Now salary report]. At AI-native companies like Clay, Common Room and 11x itself, GTM engineers manage up to ten client accounts each, layering research, enrichment and signal workflows that previously required a full SDR team.

What the GTM engineer actually does, in practice, is the technical implementation of research into automation. They live inside tools like Clay, n8n and Zapier. They wire together intent signals, firmographics, hiring data and AI enrichment into a single pipeline that scores accounts in real time. They write the prompts that the AI uses. They build the dashboards that tell the SDR team which 20 accounts to call today.

But there is an important limit worth naming. The GTM engineer is not a replacement for the research itself. They are the plumber that moves the research output around the organisation. If the research is shallow (bought list, no trigger signal, no depth on the buying committee), the engineer's automation just delivers shallow output faster. The stack only works when there is a genuine research capability feeding it.

This is the most overlooked point in the GTM engineer conversation. Companies are hiring the role expecting it to solve the pipeline problem, and some of them will be disappointed because they have hired the pipe but not the water. The correct sequencing is: research first, ICP first, intelligence first, and then the engineer who operationalises all of that. Hiring backwards is a 2025 mistake that will show up in 2026 Q3 revenue reviews.

Section 07Why research is the new pipeline.

The thesis of this entire report, if you want it in one sentence, is this. Pipeline used to be a volume problem. In 2026 it is a relevance problem. And relevance is a research problem.

Volume worked in a world where inboxes were less saturated, LLMs did not exist, and buyers had not yet learned to pattern-match cold outreach. All three of those conditions have changed. Saturation keeps rising. LLMs industrialise generic outreach faster than any human can. Buyers have eighteen months of training. Volume stopped being a moat and started being an insult.

Relevance, by contrast, works better than it ever has. A well-researched, precisely-timed, specifically-relevant outbound message in 2026 gets a response rate close to what mid-range outbound got in 2019. The premium on relevance has gone up, not down, because the contrast with the AI-generated noise is sharper.

Producing relevance requires three things: a sharp ICP, a genuine research capability and a system for converting research into outreach without losing the signal. Those are the three layers that are being repriced upward in 2026. Everything else (the tooling, the automation, the scoring) is infrastructure on top of them.

The implications ripple through every part of a B2B organisation. Marketing org design is moving from "content factory plus paid media plus events" toward a research-led shape with a small content team, a small paid team, and a disproportionately sized research and intelligence function. Budget allocation is rotating from tool spend toward research spend. Agency selection is changing. Volume agencies and dialling-for-dollars shops are losing business to research-first consultancies and intelligence firms.

This does not mean pipeline becomes slower or less scalable. It means the unit economics invert. Fewer touches, each worth more. The same quarterly number, built on a smaller and better-targeted top of funnel. The operators who already get this are quietly running outbound that looks nothing like a 2023 playbook, and their numbers are going up while the market's are going down.

Section 08Where GTM research is heading by end of 2026.

Five predictions, based on the data above and on what we are seeing across active engagements.

1. Research-as-a-service grows 25% or more in 2026.

The underlying markets are already growing at 15 to 21%. The services layer that packages research into productised engagements (not consulting, not research reports, something in between) is growing faster than the underlying tools market, because the tools need an operator to turn them into output. Expect the sub-$50K research-as-a-service category to become crowded by Q4.

2. AI-augmented research becomes the default.

Nobody in 2026 wants to read 300 pages of LinkedIn profiles to find a buying committee. The workflow will standardise: AI enrichment and synthesis, with a human analyst validating the top 10% of accounts. The human is the judgement layer, not the scraping layer. Firms that do not adopt this split will be priced out.

3. Competitive intelligence moves downmarket.

The CI category historically sold to Fortune 500 accounts at $100K+ price points. By end of 2026, we expect at least three new entrants offering credible CI at sub-$20K, driven by cheaper LLM compute and improved web-crawling stacks. This is going to put a lot of pressure on the incumbents (Klue, Crayon, Kompyte) to either move upmarket into full-fat platforms or compete on price.

4. ICP activation becomes a named discipline.

Today, "ICP" is a document. By end of 2026 it will be a function. Someone on the revenue team (possibly the GTM engineer, possibly a dedicated head of ICP operations) will own the quarterly refresh, the activation across marketing and sales, and the hard filter on unfit leads. Expect a handful of thought leaders and a few frameworks to coalesce around this over the next 18 months.

5. Original research reports become the #1 SEO and AI-search play for B2B.

This is already happening, and it will accelerate. 73% of AI business answers cite original research according to emerging data from Perplexity and SEMrush. That is three quarters of the AI-generated answers your buyers read. The implication: a well-distributed original research report has a longer half-life and a bigger search moat in 2026 than ten middle-of-the-road how-to blogs. The economics favour depth over volume in content, just like they favour depth over volume in outbound.

Section 09What this means for your GTM in 2026.

Recommendations by stage, because the right move is different at $2M ARR than it is at $50M.

Seed and Series A ($0 to $5M ARR)

Research before you scale outbound. The worst mistake at this stage is hiring two SDRs and buying a dialler before you have a validated ICP and a validated message. Do 30 founder-led conversations, build the ICP from the closes, nail the pitch, then add capacity. If you skip this, you will spend 18 months finding out your ICP was wrong, and the SDRs you hired will churn with the wrong data in their Calendly headers.

Series B ($5M to $25M ARR)

Rebuild your ICP with data from your first 100 closes. The ICP you shipped with at Seed is almost certainly wrong for the shape of your customer base at $15M ARR. Pull the data, cluster the wins, identify the trigger events and buying committees, rewrite the playbook. This is the single highest-ROI exercise most Series B companies can run this year.

Series C and beyond ($25M+ ARR)

Operationalise continuous ICP refresh. Own a research capability in-house or through a trusted partner. At this stage the problem is not that you do not have data, it is that the data is fragmented across CRM, product telemetry, marketing automation and finance. Pull it into one view, refresh quarterly, and feed the findings back into marketing and sales as named workflows.

Enterprise

Dedicated research team or dedicated research partner. Either works, and the right choice depends on whether your revenue team has the capacity to manage an in-house function. At enterprise scale the cost of a wrong ICP is denominated in tens of millions of dollars of wasted GTM investment, so the cost of the research function is trivial by comparison.

Every stage: audit your dark funnel exposure.

Ask ChatGPT and Perplexity the five questions your buyers would ask before purchasing your category. If you do not show up, or if you show up with worse positioning than your competitors, that is the top-priority problem to fix in the next 90 days. It is unlikely your marketing team is tracking this. It is more important than almost anything else on their roadmap.

Section 10How ORRJO is responding.

A note on where we sit inside this. ORRJO is a B2B growth agency. We have been running demand generation and outbound for clients for four years. The reason we built ORRJO Intelligence as a separate product line, and why this report exists, is that we got tired of running campaigns on assumptions. Too many engagements were starting with a guessed ICP, a generic message and a hopeful dashboard. We wanted to flip the order.

ORRJO Intelligence is a research-first engagement. A named senior analyst. A 14-day turnaround. Primary source verification. Fixed fee. The output is not a slide deck to be ignored. It is a working document the GTM team refers to every week: who the ICP actually is, what the trigger signals look like, who is on the buying committee, what the competitive landscape really looks like, where the dark funnel exposure sits.

We run four tiers. Intelligence Sprint for teams that need a focused ICP or market sizing piece. Intelligence for the full engagement, rebuilding the ICP and the supporting research. Intelligence Pro for ongoing refresh and competitive monitoring. Intelligence Pulse for enterprise teams that need continuous signal and a named embedded analyst.

If any of this report sounds like a problem you are running into right now, the 30-minute conversation is free. We will tell you whether an Intelligence engagement is the right fix or whether the problem is somewhere else. Most of the time it is the ICP and the research layer. Sometimes it is not, and we will say so.

Frequently asked questions.