One conversation is an anecdote; three are a finding. Primary sources decide, secondary sources direct the search.

Pharma CI draws on two kinds of source. Primary sources are people and first-hand observation. Secondary sources are the record. The profession's rule has not changed in decades: a secondary source is worth what a primary source confirms. What has changed is the cost of each, and that changes how a CI team should spend its time.

Primary intelligence is harder to get and easier to poison

Two forces are at work. Compliance and media training have made people in the field more guarded, and congress conversations in particular now deliver a filtered, often rehearsed version of what a physician or coordinator actually thinks. And competitive intelligence is played as chess: competitors seed misinformation, a timeline, a trial design, a capacity figure, knowing it will be collected. A single-source finding can be a planted one.

The counter is method, not cynicism. Put the same question to several independent sources and cross-check the answers against each other and against the record. One conversation is an anecdote. Two that agree are a signal. Three, from people who do not talk to each other, are a finding.

Two more habits make the ladder work: grade every source for reliability and every item for credibility, and record who said what to whom. Without that record, nobody can tell three confirmations from one rumour repeated three times.

The richest source is the one companies read least

The best primary source most companies own is their own field staff. Clinical research associates visit competitor-active sites every few weeks. Medical science liaisons and application specialists sit with the people who run competitor trials. Sales representatives hear which product a centre prefers, and why. Sites are overloaded, so what coordinators say is shorter and worth more per sentence.

Almost none of it reaches the people who need it. It stays in site-visit reports and in the heads of the people who heard it. Where clinical operations are outsourced the gap is wider still: the people who visit sites most often work for the CRO, with no trigger to look for competitive knowledge, no incentive to share it and a contract that does not ask them to. The sponsor-side study managers who oversee the CRO are the channel the company controls, and one standing, compliant question in every oversight meeting, what has changed at the sites?, opens it.

That channel carries an obligation. What field staff learn about another sponsor's trial is often confidential under the site's agreements. Written rules of engagement must say what may be recorded, what may not, and who decides, before the first debrief is logged.

AI makes the record fast. Use it to aim the conversations

The vendor floor at this year's Pharma CI USA conference showed what AI now does well: it combines secondary sources in minutes rather than weeks, keeps a landscape current as the field moves, and reads registry amendments, hiring, manufacturing announcements and agency calendars together to estimate when a competitor event will land. None of that is intelligence until a primary source confirms it, and a model trained on public text inherits the misinformation in it.

The right use of AI is to decide where the scarce primary effort goes: which competitor moves matter, which centres and people are worth a conversation before the next congress, and which secondary finding most needs a human to confirm it. AI sets the questions and the targets. People get the answers. The record keeps both honest.

Comphas keeps the record and the conversations in one place: the trial that moved, the people worth asking, what they said, how far to trust it, and whether it is confirmed enough to decide on. See how Comphas aims the primary effort.

Start with the record

The secondary record tells you where to point the conversations. Ask it a question about your own field, then decide who you need to talk to.