Walk into most CI functions and audit what actually ships. Weekly competitor newsletters. Quarterly battlecard refreshes. Alert digests forwarded with a sentence of commentary. Conference summaries. A monthly deck that lists what rivals announced, in the order they announced it.
Now ask a harder question: which decision changed because of any of it?
In most functions, the honest answer is none, or almost none. The output is accurate, current, professionally formatted, and inert. Strip away the branding and it is a news clipping service with a salary band. Companies pay six figures per analyst for a product that Google Alerts delivers for free, plus formatting.
This is not an insult to the analysts. Most of them are capable, curious people who know something is wrong. The drift into summarization is not a talent failure. It is a structural outcome, produced by five forces that push every CI function toward the same equilibrium unless someone actively pushes back. This piece is about those forces, because you cannot fix a drift you have misdiagnosed as laziness.
Force 1: Summarization is legible, judgment is not
Start with the cruelest asymmetry in the job. A summary can be evaluated instantly. Anyone can check whether the newsletter covered the week's announcements, whether the dates are right, whether the competitor's press release was accurately condensed. Completeness and accuracy are visible at a glance.
Judgment cannot be evaluated instantly. A genuine analytical call, this move means X, therefore we should do Y, takes months to be proven right or wrong, and even then the proof is contaminated by everything else that happened. A manager reviewing a summary knows within minutes whether the analyst did the work. A manager reviewing a judgment has to either share the judgment or wait a year.
Organizations manage what they can see. So the visible virtues, coverage, timeliness, accuracy, polish, get rewarded on every cycle, while the invisible virtue, being usefully right about what matters, gets rewarded rarely and noisily. Analysts are not stupid. They optimize for the scoreboard that exists. The scoreboard measures summarization.
Force 2: The risk asymmetry punishes exactly the work that matters
A summary cannot be wrong in any career-damaging way. If the newsletter says a competitor launched a product, and the competitor launched a product, the analyst is safe. The worst case for pure reporting is a factual error, which is embarrassing and correctable.
A judgment can be wrong in a way everyone remembers. Call a competitor's move a strategic pivot, recommend a response, have the pivot fizzle, and your next ten reports get read with a discount. The analyst who never concludes anything never takes that hit.
So the incentive gradient inside every report points downhill, away from the conclusion. Add one more caveat. Soften the recommendation into "worth monitoring." Present both interpretations and let the reader choose. Each individual retreat is defensible. Compound them across a hundred reports and the function has fully converted into a summarization shop, without any single person ever deciding it should.
Here is the uncomfortable part: the readers cooperate. Executives rarely punish an inconclusive report. They skim it, nod, and derive their own conclusions in the hallway afterward. The report that concludes nothing generates no friction, and no friction feels like success. Both sides of the transaction are locally satisfied while the function's actual reason for existing quietly evaporates.
Force 3: The tooling industry sells collection, so collection becomes the job
Look at the CI software market. Almost every product in it is a collection and monitoring engine: track competitor websites, scrape pricing pages, aggregate news, alert on changes, auto-populate battlecards. The pitch is always the same: never miss anything.
Never missing anything is a summarization goal. No tool on the market sells "reach a conclusion," because conclusions do not demo well. But the tools shape the workflow, and the workflow shapes the job. An analyst whose day is structured by an alert queue becomes a queue processor. The unit of work becomes the item, cleared, tagged, forwarded, and the backlog is infinite by design, because the internet produces more competitor-adjacent content every day than any team can process.
This is how capable analysts end up genuinely busy while producing nothing that moves a decision. The queue is always full. Clearing it feels like diligence. And the deep work, sitting with three weeks of accumulated signals and asking what they add up to, has no queue, no alert, and no tool. It has to be defended as unstructured time, and unstructured time is the first thing that dies when the queue is full.
Volume of input became the proxy for quality of function. That proxy is not just wrong; it is inverted. The functions drowning in the most feeds usually have the least time to think.
Force 4: CI is org-chartered as a service, not as a participant
Ask where CI reports in most companies. Usually into marketing, sometimes product, occasionally strategy. Almost always it is chartered as a support service: internal clients submit requests, CI fulfills them. Sales wants battlecards. Product wants a feature comparison. An executive wants a briefing before a board meeting.
A service function answers the questions it is asked. And the questions it is asked are overwhelmingly descriptive: what did they launch, what do they charge, who did they hire. Descriptive questions have summary-shaped answers. The requester was never going to ask "what should we conclude from the pattern of the last six months," because the requester does not know there is a pattern; noticing patterns nobody asked about is precisely the job that falls to no one when CI is a ticket queue.
The analysts closest to the evidence are structurally excluded from the conversations where the evidence would matter. They find out which decisions were on the table after the decisions are made, usually by reading the same announcements they summarize for everyone else. Intelligence produced outside the decision loop can only ever be news, because relevance to a decision is the only thing that separates intelligence from news, and you cannot aim at a decision you cannot see.
Force 5: Nobody ever closes the loop
The final force is the missing feedback cycle. In functions that make real calls, forecasting, trading, underwriting, predictions get scored. Someone writes down what was claimed, waits, and checks. The pain of being visibly wrong is the mechanism that improves judgment.
CI functions almost never do this. Reports ship, land, and vanish. No one records what the report claimed would happen, so no one ever checks, so no one ever learns whether the function's judgment is any good, so the judgment never improves, and, more corrosively, never earns trust. Executives discount CI conclusions partly because those conclusions have no track record, and they have no track record because nobody kept one.
Summarization thrives in this vacuum because it is the one product that needs no track record. Yesterday's newsletter is not falsified by tomorrow's events. A function that never scores itself will drift toward the only output that cannot lose.
What the drift actually costs
The obvious cost is the salary line, paying analyst rates for aggregation work. That is the small cost.
The real cost is the decisions made blind while the CI team was busy. Every company that got surprised by a competitor's move it "knew about", the signals were in the newsletter, in week 34, bullet four, paid this cost. The information was collected, summarized, distributed, and never converted into a conclusion anyone had to confront. Collection without conclusion produces the most dangerous state in competitive strategy: the illusion of awareness. Leadership believes it is informed because the briefings arrive on schedule. Informed is not the same as prepared, and the gap between them is exactly the analytical work the function stopped doing.
There is a third cost, quieter: the analysts themselves. The people drawn to CI wanted to be detectives and became clipping services. The good ones leave for strategy roles where conclusions are the job. The function keeps the ones who made peace with the queue, which locks the equilibrium in for another cycle.
The test, and the direction out
One diagnostic separates an intelligence function from a summarization function, and it takes thirty seconds. Take the last five deliverables the team shipped and ask of each one: does this document make a claim that could turn out to be wrong?
Not a fact that could be mistyped. A claim. A judgment about what something means or what will happen or what should be done, stated plainly enough that reality could contradict it. If none of the five contain one, the function is summarizing, whatever its charter says. Risk of being wrong is not a flaw of analysis; it is the definition of it. A deliverable that cannot be wrong contains no analysis, only arrangement.
The direction out follows from the diagnosis, force by force. Measure conclusions shipped and decisions touched, not items covered, so judgment gets a scoreboard. Make the analytical call a required section of every deliverable, so the retreat downhill has a floor. Cap collection deliberately, fewer feeds, more synthesis time, because the queue will never volunteer to shrink. Put analysts inside the decision conversations, or at minimum tell them which decisions are live, so the work has a target. And score the calls: keep a visible log of what the function claimed, revisit it on a schedule, and let the track record, including the misses, accumulate. The misses are not the risk. The absence of any record is.
None of that is complicated. All of it is uncomfortable, because every step trades safety and legibility for exposure. That is the actual price of doing intelligence rather than reporting on the news, and most functions, without ever deciding to, have chosen not to pay it.
The competitors' announcements will be summarized either way. The only question is whether anyone in the building is paid to say what they mean.