They Decided Before They Replied

A connection request gets accepted. Four days later a reply arrives, and it is either warm or it is nothing. Most teams read that reply as the start of a conversation. It is closer to the end of one.

The person on the other side has already looked at your profile, read whatever you posted in the last month, opened your company page and formed a view. The message they send back is a verdict on research they did without you.

What the buyer did before you heard from them

Research on B2B buying behaviour puts a number on how early this happens. The 2025 Buyer Experience Report from 6sense, built on nearly four thousand buyer responses, found that buyers reach the point of first contact at roughly 61% of their journey, down from 69% the year before. Buyers initiated 79% of engagements. And four out of five deals are won by the vendor the buyer already favoured before anybody spoke.

That study covers considered purchases with a median cost between $200,000 and $300,000, in technology, services and manufacturing. If you sell a twelve thousand dollar engagement to a contractor, the proportions will differ. The direction will not.

It means your outreach message is not making the first impression. It is being read by somebody who has already formed one.

Why outreach still behaves like first contact

Because the sequence is the part you control. You can write five messages, set the delays, pick the list and press start. Everything the buyer actually used to decide, your profile, your last six posts, the case study they half read, sits outside the sequence and outside the tool.

So the work goes where the lever is. Teams rewrite the opener for the ninth time, test a new hook, shorten the message, and the reply rate moves a point. Meanwhile the thing that settled the outcome was a page somebody looked at on a Tuesday.

There is a second reason. Running outreach properly means knowing, for each person, what they have already seen and what is happening at their company this month. At fifty contacts somebody can hold that in their head. At five hundred nobody can, so everyone gets the same message and the system treats every conversation as though it started that morning.

What AI changed about this

The expensive part was never sending. It was reading. Knowing that this particular company posted three operations roles last month, opened a second location in the spring and still lists a process you replace, used to take a person ten minutes per contact. Across five hundred contacts that costs more than the deals it wins.

That reading is now cheap. A system can take a list, pull what is publicly visible about each company, compare it against what you actually sell and tell you which twenty of the five hundred have a reason to care this quarter. Not a score out of a hundred. A sentence saying what the reason is.

The second change is that the message can be written to the research rather than to the segment. Not a merge field with a company name dropped into a template, which every buyer now recognises on sight, but a first line that refers to something true and specific about that business. What the reader notices is whether anybody looked.

The third is timing. A system can watch for the events that make an approach reasonable, a role posted, a site opened, a leadership change, and surface the contact in the week that matters rather than in the week your campaign happened to run.

What this does to your profile and your posts

If the decision is largely made before the reply, your profile and your public activity are part of the outreach system rather than a separate marketing exercise. The profile is the landing page for every message you send. The posts are what somebody reads to decide whether you know the work.

This is uncomfortable, because it means outreach volume is capped by something that is not in the outreach tool. Sending more messages against a profile that says nothing specific raises the number of people who look and leave.

What AI will not do for you

It will not know your business. The judgement about which signals predict a good client, which industries waste your time, which job title holds the budget and which one only influences it, comes out of your own history of won and lost deals. A system can apply that rule across five hundred companies in an hour. It cannot invent it.

It will not supply a point of view either. The posts that make somebody take your message seriously carry an opinion that could be wrong, and a model asked for safe professional content reliably produces something nobody disagrees with and nobody remembers.

And it will not rescue a list of people who cannot buy. Careful research applied to the wrong five hundred companies produces well informed messages to people with no reason to answer.

What to build instead

Before buying anything, write down what a good client looked like. Pull the last twenty deals you closed and the last twenty that went nowhere, and find the two or three facts that separate them. That is the rule the system applies.

Put the research step before the sending step. For each company on the list, the system should produce one specific, checkable reason this approach makes sense now, or mark the company as having none. A contact with no reason does not get a message.

Treat the profile as part of the sequence. Whatever claim the message makes, the profile should support it within ten seconds of somebody arriving.

Then measure the whole thing on conversations started, not on connections accepted. Acceptance has been cheap for years. A reply from somebody who already knew what you do is the only output worth counting.

We build this as a system that does the reading, writes to what it found and tells you which contacts to leave alone, rather than a sequence tool pointed at a longer list. See LinkedIn Outreach.

Sources

6sense, B2B Buyer Experience Report 2025, announced 12 November 2025. 6sense.com