Buyer Intent Signal Synthesis With AI

Fit tells you who could buy. Intent tells you who is buying.

Your ideal customer profile (ICP) model identifies 4,000 accounts that match your ideal customer profile on firmographic and technographic dimensions. Revenue range, headcount, tech stack, industry vertical. Those accounts are all plausible buyers. But at any given moment, maybe 200 of them are actively evaluating a purchase in your category. Those 200 accounts are worth your outbound effort this week. The other 3,800 can wait.

The problem is figuring out which 200. Intent data is the answer, but raw intent signals are overwhelming. A single third-party intent provider might flag 600 accounts per week for your category. G2 reviews, pricing page visits, competitor research activity, content downloads, conference attendance, and job posting changes each add signals. Without synthesis, your team drowns in noise.

AI synthesizes intent signals from multiple sources into a unified in-market score with a rationale. That's what makes the signal actionable. This article covers the signal taxonomy, how synthesis works technically, the vendor landscape, and the false positive problem that every intent-driven program eventually encounters. The Forrester Wave for Intent Data Providers for B2B, Q1 2025 evaluates the major vendors and finds that signal synthesis across multiple sources remains the key differentiator between leaders and laggards. For the ACE pattern powering this, see Generative Research: Compressing Hours of Reading.

First-party, second-party, and third-party intent signals

Key Facts: Buyer Intent Data Impact on B2B Sales

• Companies that integrate intent data into their pipeline process see 37% higher lead conversion rates while reducing acquisition costs by 25%. (The Insight Collective, 2025)
• Teams that act on intent signals within 48 hours see 4x higher conversion rates than teams that respond after the 48-hour window. (Landbase, 2025)
• 93% of B2B marketers report conversion rate improvements when implementing intent-based targeting strategies. (Shortlister, 2025)

Not all intent signals come from the same place. Understanding the taxonomy is the prerequisite for understanding why synthesis matters.

First-party signals

These are signals from your own properties. The account is engaging with you directly.

• Pricing page visits (high intent, especially multiple visits in a short window)
• Free trial starts or product-qualified lead actions
• Documentation or integration pages (signals evaluation in progress)
• Demo request form starts or abandons
• Webinar registrations in a specific product area
• Email open and click patterns on sales sequences

First-party signals are the highest-confidence signals you have because you own the data and the context is specific to your product. The limitation is coverage: first-party signals only tell you about accounts that have already found you. They don't identify accounts evaluating your category who haven't visited your site yet.

Second-party signals

These are signals from partners or co-ops where companies share intent data directly.

• G2 Buyer Intent: accounts viewing your listing or competitor listings on G2
• LinkedIn activity: accounts where multiple employees have viewed your company page or engaged with content
• Category-specific data co-ops where companies share behavioral data with each other

Second-party signals extend your coverage to accounts that are researching the category but haven't directly engaged with your site. G2 Buyer Intent is the most widely used source in B2B SaaS because category research on G2 is a reliable signal of active evaluation. An account where 3 employees viewed your competitors' G2 listings in a 10-day window is clearly doing comparative research.

Third-party signals

These are signals from external data providers that monitor behavior across the broader web.

• Bombora: tracks content consumption across a co-op of 5,000+ B2B sites, flags accounts showing "topic surges" in relevant categories
• 6sense: predictive intent using AI to model in-market accounts from anonymous buying signals
• TechTarget Priority Engine: monitors research activity on TechTarget properties
• DemandBase: account identification plus intent signal aggregation
• ZoomInfo Intent: web research signals plus the ZoomInfo data graph

Third-party signals have the broadest coverage and the lowest signal-to-noise ratio. An account "surging" on a Bombora intent topic might be doing research for reasons unrelated to buying: a competitor analysis, an internal training project, a journalist writing an article. The signal is probabilistic, not deterministic.

The signal synthesis problem

Here's the core technical challenge: each source gives you a fragment of the picture.

An account might show moderate Bombora intent (they're reading category content), zero first-party signal (they haven't visited your site), a G2 competitor listing view (they're doing comparative research), and a recent job posting for a "Head of Revenue Operations" (they're building the function that would buy your product). None of those signals alone crosses a threshold. Together they tell a coherent story: this company is evaluating whether to formalize their RevOps stack, probably in the next 90 days.

AI synthesis combines signals across sources to surface accounts where the aggregate picture is compelling, even when no single source provides a clear signal.

The synthesis pipeline in the ACE Framework:

Ingest collects signals from all connected sources. Intent feeds via API (Bombora, 6sense), CRM first-party event tracking, G2 integration, and any custom signals (job posting scrapers, LinkedIn monitoring). Each signal arrives with a timestamp, source, account identifier, and signal type. The Ingest capability covers how multi-source data collection works at this foundation level.

Analyze normalizes, weights, and deduplicates. The same company might appear as "Acme Corp" in one feed and "Acme Corporation" in another. Account matching is the first job. Then weighting: not all signals carry equal information. A pricing page visit yesterday is weighted more heavily than a whitepaper download three months ago. And recency decay: signals older than 90 days typically get discounted significantly in the model.

Generate produces an in-market score (a number representing probability of active evaluation) and a rationale brief. The rationale is what separates AI synthesis from a raw score: "This account is showing intent because: pricing page visited 3x in 7 days, 2 employees viewed competitor G2 listings, Bombora topic surge on 'sales analytics software.' Combined score: 84. Recommended action: prioritize for direct outreach by account executive (AE) within 48 hours."

The rationale is what the rep actually reads. A number alone doesn't tell a rep why to call. The rationale gives them the opening.

How recency and signal strength weighting works

Intent signals decay. An account that downloaded a whitepaper 6 months ago was interested then. They may have already bought a competitor. They may have shelved the initiative. They may have forgotten they ever downloaded it.

Recency decay in intent synthesis works like this: signals are weighted by a decay function based on age. A common model uses exponential decay, where a signal's weight halves every 30 days. A pricing page visit yesterday has full weight. The same visit 30 days ago has half weight. 90 days ago has an eighth of the weight.

Signal strength weighting is separate from recency. Some signals are inherently stronger than others, regardless of when they occurred:

• Demo request form start: very strong (explicit purchase intent)
• Pricing page visit: strong (evaluating cost)
• G2 competitor comparison: strong (comparative shopping)
• Blog post read: weak (awareness level content consumption)
• Bombora topic surge: moderate (category interest, not product-specific)

The synthesis model combines recency decay with signal strength. An account with a pricing page visit today plus 2 G2 competitor views this week scores higher than an account with 10 blog post reads across the last 60 days. That distinction matters for prioritization.

Most dedicated intent platforms (6sense, Bombora) build these models internally. When you're connecting signals yourself through a tool like Clay or a custom data pipeline, you need to define the weighting logic explicitly. The default of treating all signals equally produces noisy prioritization.

Connecting intent scores to routing and action

A synthesized intent score sitting in a database doesn't do anything. The Execute step is what turns the signal into sales motion.

When an account crosses a defined intent threshold (say, a combined score above 75), the system should:

1. Flag the account in the CRM with an intent alert and the rationale brief
2. Check whether the account is already in an active deal or sequence
3. If not in motion, trigger an alert to the owning SDR or AE with the recommended action
4. If account routing rules apply (the account belongs to a specific territory or is an existing customer), route to the appropriate owner

AI Lead Scoring Beyond Rules-Based Models covers the scoring mechanics in detail. The routing step here is more specific: intent signals often arrive for accounts that don't fit the normal inbound lead flow, accounts that were cold before the signal appeared. Your routing rules need to handle that case.

Real-Time Account Tier Reassignment With AI describes how intent signals can trigger account tier changes dynamically, moving an account from cold outreach to priority AE coverage without waiting for a quarterly planning cycle.

Intent vendor comparison

Six vendors dominate the B2B intent data space. Each has different signal coverage, use cases, and cost structures.

Bombora is the foundation of most intent stacks. Their "Company Surge" data monitors content consumption across a co-op of 5,000+ business content sites. Strong for broad category research signals. Integrates with Salesforce, HubSpot, and most customer data platforms (CDPs). Pricing is enterprise contract, typically $2,000 to $6,000/month depending on topics and account volume.

6sense goes beyond raw signals to build predictive in-market accounts. Their AI model attempts to identify which accounts are in each stage of the buying journey, not just which ones are consuming content. Strong for accounts that have anonymized their web behavior. Higher cost and more implementation complexity. Most appropriate for teams with dedicated RevOps capacity.

DemandBase combines account identification (connecting anonymous web visitors to companies) with intent data. Strong for first-party intent enrichment: knowing which company is on your website even before they fill out a form. Also provides third-party intent via their own data graph.

G2 Buyer Intent surfaces companies researching your G2 listing, competitor listings, and category pages. Uniquely valuable because G2 research is specific to software evaluation. High signal quality, limited to G2 platform behavior. A natural fit for SaaS companies. Moderate cost; integrates directly with major CRMs.

TechTarget Priority Engine is domain-specific to technology buying. Strong coverage of enterprise IT evaluations. Most useful for technology vendors selling to IT and engineering buyers.

ZoomInfo Intent combines their firmographic data graph with intent signals from monitored web sources. Convenient if your team already uses ZoomInfo for prospecting. Intent data quality is generally considered below Bombora and 6sense by practitioners, but the data consolidation is attractive.

B2B buyers conduct an average of 12 online searches before visiting a specific brand's website, and 81% of sales reps observe that buyers increasingly research vendors before initiating contact. (Gartner, 2025) By the time a first-party signal fires (a pricing page visit), the buyer has usually already done extensive competitive research through channels you can't see.

The Fit-Times-Intent Quadrant

The Fit-Times-Intent Quadrant is a two-axis prioritization model that plots accounts by ICP fit (high vs. low) on one axis and intent signal strength (high vs. low) on the other. The four resulting quadrants produce distinct recommended actions: high fit plus high intent means prioritize immediately; high fit plus low intent means nurture systematically; low fit plus high intent means qualify before committing AE time; and low fit plus low intent means do not prioritize. Teams that apply fit filtering before acting on intent signals convert 2-3x more intent-triggered opportunities than teams that treat all high-intent signals as equally actionable.

---

The fit-times-intent quadrant

Prioritization becomes clear when you plot it on two axes: ICP fit (high vs. low) and intent signal (high vs. low).

High fit, high intent: Prioritize immediately. These accounts match your ICP and are actively evaluating. Every rep should know these accounts by name this week. First-touch should be personalized and direct.

High fit, low intent: Nurture systematically. They're the right company, but they're not shopping yet. Stay visible with relevant content and account touches. Gartner's research on account-based marketing (ABM) best practices with intent data recommends a tiered cadence approach for nurturing high-fit, low-intent accounts. AI-generated personalized outreach at scale covers automated nurture at this tier.

Low fit, high intent: Proceed carefully. They're shopping, but probably not for your solution. Worth a quick qualification call to understand if there's a use case that bridges the fit gap. Don't commit AE time until qualification confirms fit.

Low fit, low intent: Do not prioritize. Outbound to these accounts is usually cost without return.

The most common mistake in intent-driven sales programs is treating "high intent" as sufficient for prioritization without applying the fit filter. An account with massive intent signals that doesn't match your ICP is a waste of SDR time.

The false positive problem

Intent data will send your team after accounts that aren't actually buying. Accept this as a design constraint, not a product failure.

A company's employees researching your category might be:

• Writing an industry analysis for internal strategy
• Doing competitor research on behalf of a company in your space
• A researcher or analyst preparing a market report
• Evaluating the category to not buy (to justify continuing with their current solution)

The signal is real. The purchase intent may not be.

How to manage the false positive rate:

Set conversion tracking on intent-triggered outreach. Track what percentage of accounts that triggered intent alerts actually converted to qualified opportunities. If the rate is below 10%, your threshold is too low, your signal weighting is off, or you're not filtering on fit.

Build in a light qualification step before AE time. An SDR email or phone call to qualify intent before routing to an AE conserves AE capacity on the signal-to-noise ratio problem.

Review dismissed intent alerts. When an SDR or AE dismisses an intent alert without action, capture the reason. Patterns in dismissal reasons reveal weaknesses in your synthesis logic.

Educate reps on probabilistic framing. High intent means higher probability of active evaluation, not certainty. Reps who treat intent signals as guaranteed pipeline are setting themselves up for frustration. Reps who treat them as prioritization inputs behave more effectively.

AI Account Research Before First Touch covers how to validate an intent signal with account research before reaching out, turning a probabilistic signal into a more confident decision.

Getting started with intent synthesis

For teams that don't have a full multi-source intent stack:

Start with first-party. Instrument your website properly. Know which accounts are visiting which pages. Tools like Clearbit Reveal, 6sense, or DemandBase identify the company behind anonymous web visits. This is lower cost than third-party intent and higher signal quality.

Add one third-party source. Either G2 Buyer Intent (if you're a SaaS company with G2 presence) or Bombora (if you want broader category coverage). Don't subscribe to four intent vendors simultaneously; you'll create more noise than signal before you know how to use one well.

Define your synthesis logic explicitly. Even if you're combining signals manually in a spreadsheet initially, document how you weight them. This becomes the specification for the automated system you'll build later.

Set a threshold and measure conversion. Pick a combined intent score that triggers outbound action, track what happens, and adjust the threshold quarterly based on conversion data.

The underlying pattern is Generative Research plus Scoring and Routing. Signal collection, synthesis, score generation, and action routing. The ACE Framework's Predict capability is central here: the synthesis model is essentially a prediction of in-market status from available signals.

Fit and intent together is the most actionable combination in AI-assisted sales prioritization. Neither alone tells you enough. Fit without intent is a static list of potential buyers. Intent without fit is noise. Combined and synthesized, they give you the 200 accounts worth calling this week. Forrester's framework for evaluating intent data providers offers a practical starting point for teams building a multi-source synthesis stack: it identifies which signal types produce actionable results versus which create noise without fit-qualification layered on top.

Rework Analysis: In Rework deployments, the most reliable intent stack for mid-market B2B SaaS combines three sources: first-party CRM activity (pricing page, docs, integration pages), G2 Buyer Intent (category comparison signals), and one broad third-party provider (Bombora or 6sense). Adding a fourth source before the first three are calibrated typically increases noise without improving the signal-to-conversion ratio. Calibrate the 48-hour response threshold first, then expand coverage.

---

---

What to read next

• Generative Research: Compressing Hours of Reading: the ACE pattern that powers multi-source intent synthesis
• AI Account Research Before First Touch: validating intent signals with account research before outreach
• AI Lead Scoring Beyond Rules-Based Models: how intent signals feed into the broader lead scoring model
• Competitor Battlecards Generated With AI: acting on competitive intent signals from G2 and third-party sources