How Account Intelligence finds buyers before they move

By the time a buyer emails you, the interesting part is over. 6sense's 2024 buyer research found that 85% of B2B buyers have largely fixed their requirements before they ever contact a seller. The 2025 follow-up went further: the vendor that eventually wins comes from the buyer's day-one shortlist 95% of the time, up from 85% the year before. The pitch you spend weeks preparing lands in the final third of a decision that was mostly made without you.

So the advantage isn't in the pitch. It sits upstream, in the quiet weeks before a buyer raises a hand, while the need is still forming and the shortlist is still soft. Account Intelligence is the engine that operates there. This piece opens the hood: how the target list gets built, how the need gets proven, how the system handles a signal that turns out to be false, and what your team ends up holding when they finally walk into the room.

The list is the product

Most target lists are firmographic guesses dressed up as strategy. Revenue band, headcount, an industry code, maybe a recent funding round. They tell you an account exists. They don't tell you it's in motion, and motion is the only thing that matters when your sales team has forty hours this week and no way to spend them everywhere.

Account Intelligence is built to answer a narrower question. Of all the accounts that could theoretically buy from you, which ones are forming the need right now, and who inside them will move first? The design goal is precision, not coverage: not the 2,000 accounts that look right on paper, but the 30 to 200 that actually sign.

Michael Lewis told this story in Moneyball. For a century, baseball scouts ranked prospects by how they looked: the build, the swing, the confident jaw, what the trade quietly called a "good face". Billy Beane, running a team that couldn't afford to be wrong, threw that out and ranked players by on-base percentage instead. An unglamorous statistic that actually predicted runs. He bought wins that everyone else's eyes had mispriced. A target list built on firmographics is scouting by good face. Account Intelligence scores by on-base percentage: the dated, checkable signal that a need is real, not the profile that merely looks the part.

Step one: two to five ICPs, each with its own playbook

Everything starts with the Ideal Customer Profile, and here the number matters. Account Intelligence builds two to five ICPs, drawn from three inputs: your market, your genuine product fit, and your historical wins. Not who you wish bought from you. Who actually did, and why.

Each ICP is a full description of a segment rather than a demographic sketch. It covers the market dynamics that segment lives inside, its real pain points, how it makes decisions, and how it moves budget. Then each one carries its own playbook: the specific objections you will hear from buyers in that segment, the counter-arguments that resolve them, and the proof points that move the conversation forward.

Why two to five and not one? Because a single ICP flattens a market that behaves in segments, and a flat map gets your best rep lost. April Dunford, who has built a career on B2B positioning, keeps returning to the same point: context is what makes a product make sense, and different buyers sit in different contexts. Why not twenty ICPs? Because past five, the playbooks stop being distinct. You are no longer segmenting a market. You are making noise with a template.

Step two: research agents that hunt for evidence, not names

Once the ICPs are codified, the framework goes to work. Account Intelligence deploys AI research agents across the open web, and they are not looking for contacts. They are looking for evidence of need.

That means concrete, dated, public signals: a regulatory filing that forces a change, a job posting for a specific role that reveals a gap, an earnings call that names a budget line, a supply-chain disclosure that exposes a dependency on a legacy system nobody wants to defend. This is the difference that runs through the whole engine. LinkedIn guesswork tells you a VP of Compliance exists. A job posting for a compliance-automation lead, filed last month, tells you the gap is real and the budget is moving.

Every account the agents surface is then scored against the ICP framework, so you know three things: who fits, how well, and why. The score itself is composed, not conjured. It weighs how cleanly the account maps to an ICP, how strong and how recent each signal is, and how many independent signals point the same way inside a short window. Because the inputs are public and dated, the score is auditable: you can click straight through to the filing that produced it.

Arthur Conan Doyle handed his detective a rule worth taping to the wall: it is a capital mistake to theorise before you have data, because you start twisting facts to suit theories instead of theories to suit facts. Generic ABM theorises. It decides an account is a good fit, then goes looking for reasons. Account Intelligence runs the other way. It gathers the evidence first and lets the score land where the evidence puts it.

When a signal turns out to be false

Here is the fair objection a sharp reader will raise: what if the agent misreads the job posting? A "Head of Compliance Automation" opening might be a routine backfill, not a new initiative. A budget line on an earnings call might be last year's, restated. Machines pattern-match, and pattern-matching produces false positives. Ignore that and the whole system is just a faster way to be confidently wrong.

Three mechanisms keep a false signal from becoming a false account.

The first is corroboration. One signal is a hypothesis, not a verdict. An account only climbs the ranking when independent signals stack in the same direction inside a tight window. A lone job posting nudges it a little. A posting, plus a fresh filing, plus a leadership change, all landing within a fortnight, move it a long way. A single signal with nothing behind it stays a maybe, and a maybe does not reach your reps as a certainty.

The second is auditability. Because every input is public and dated, a human can open the source and overrule the machine before the account goes anywhere. The score shows its working. This is deliberate: a number your reps cannot interrogate is a number they will quietly ignore, so the system is built to be interrogated rather than trusted on faith.

The third is the ICP acting as a filter. Evidence only counts when it maps to one of your codified profiles. An account throwing off signals that fit none of your two to five ICPs does not rank, however loud it gets. That is what stops the engine chasing noise that merely resembles intent.

The honest version is that the machine will occasionally be wrong about a single account. It is designed so that being wrong is cheap and visible rather than confident and buried. A false positive costs you one rejected card in a review. That is the trade every intelligence operation makes: better a handful of false alarms you can dismiss in seconds than one real signal you never saw at all.

Step three: the Evidence Pack, built for your side of the table

The Evidence Pack answers one question, and it is an internal one: should we spend the forty hours on this account? It is built for your side of the table, the sales director running the pipeline review and the rep deciding where to aim this week.

Each qualified account arrives with the raw proof attached: the URLs, the direct quotes, and the technical specifications that show the need exists today, not someday. "This account matches our ICP" is a claim, and claims get argued down in the pipeline review. "Here is the filing, dated six weeks ago, that says they must solve this before their next audit cycle, and here is the earnings call where their CFO named the budget for it" is a case. Cases survive the review and get a name on the board.

So the Evidence Pack is proof plus a verdict: the sources, the score, and the reason the score reads the way it does. It gets the account into your pipeline. It does not tell your rep what to say once they are in the room. That is the next artefact's job.

Step four: the Strategic Briefing, built for the other side of the table

If the Evidence Pack is for your side of the table, the Strategic Briefing is built for the other side. It does not ask whether the account is worth pursuing. It answers how to win the conversation once you are pursuing it.

The Briefing lays out the account's specific pain points, the executive sponsor most likely to champion the project internally, the external pressure that makes this quarter the natural buying window, what to lead with, and the playbook for the three objections the CFO is most likely to raise.

Take a scenario the system runs often: a regional bank facing tightening anti-money-laundering rules. Three signals stack up inside a fortnight. A fresh AML filing marks the regulatory trigger. A VP of Compliance changes role, which resets the internal politics. A disclosure surfaces a legacy monitoring system the bank can no longer stand behind. Each signal is public, dated, and independently checkable, which is what got the account through the Evidence Pack stage. The Briefing turns that proof into a plan: your rep walks in already knowing the sponsor to court, the audit deadline that sets the clock, and the two objections procurement will lead with. The Evidence Pack decided the account was real. The Briefing decides what happens in the meeting.

Why the list never goes stale

A static target list is a photograph. The market is weather. Photographs of weather are useless within the hour.

Account Intelligence rescores accounts daily as signals shift, and reranks the whole list by how likely each account is to close in the next 18 months. The moment something material changes inside an account, a funding event, a leadership move, a new filing, the ranking changes with it. Meteorologists do not forecast once a season and go home. They nowcast, folding in fresh readings as they arrive, because a forecast that ignores this morning's data is just yesterday's guess with confidence added. The list works the same way. It is never finished, which is exactly why it stays useful.

From there, the strongest accounts hand off to The Knowledge Graph, where each one becomes a living dossier: who decides, who blocks, who signs, and the account-specific play waiting to run the day a trigger fires. Account Intelligence answers which accounts. The Knowledge Graph answers everything that happens once you are inside one.

The work before the decision

The decision gets made before the pitch, so the work has to happen before the decision. That single fact is the whole design. Account Intelligence builds the ICPs, sends the agents, gathers the evidence, discards the false signals, scores what survives, and hands your team a briefing instead of a spreadsheet.

The old model spent its energy on persuasion at the end of the journey, in the validation phase, where the buyer had already chosen and was mostly looking to confirm. This one spends its energy on discovery at the start, where the need is still forming and the choice is still live. Whoever sees that need first, with evidence in hand, gets to help shape it. Everyone else arrives later to validate a decision that was made without them.

Related reading in Insights: From Spray-and-pray to Forensic Precision: Why Account Intelligence Wins for the case behind the method, and The 3 Signals That Reveal a Buying Window for more on trigger detection.