The battlecard is dead. Your reps opened Claude instead.

You bought the tool to watch competitors. It builds battlecards, tracks pricing pages, pings you on launches. You pay annually. Then a competitor comes up on a call, and your rep opens Claude. Wynter's 2026 State of Competitive Intelligence in B2B SaaS — 101 product marketers at mid-market and enterprise firms — asks where they get competitive intelligence. Twenty-one percent cite ChatGPT, Claude, or Gemini. Fourteen percent name a dedicated CI tool. Reddit beats both at 23%. One senior product marketer said: "I haven't seen any CI tools worth the investment. I can replicate most of what Crayon and Klue do with an agent in Claude." Read that against what these tools cost. A general-purpose chatbot with zero CI features outranks the products built for the job. Buyers report no loss. The category is being hollowed out in real time by a product that was never aimed at it.

None of this is because AI is clever. It's because of what a CI tool hands you: a battlecard, a one-page brief a rep pulls up mid-deal. Wynter found 47% of battlecards go stale within three months, 82% within six, and only 2% last more than a year. Competitors now ship features, change pricing, and reposition every couple of weeks. A document refreshed quarterly describes a company that stopped existing between refreshes. Reps stopped opening it. Only one in three sales teams consistently uses the competitive content product marketers produce. Thirty-seven percent freestyle or ignore it. One director of product marketing: "They mostly freestyle. I see very little traffic to my content." A model has no such problem — ask it about a competitor and it assembles an answer live. The battlecard was a photograph aging on a shelf. The chatbot writes a new one every time you ask.

The fix is narrow. Put the intelligence where the work happens, and put a person on it. Wynter's own read is that the answer is intelligence in the workflow — wired into the tools reps already live in, Slack and your CRM, backed by a model with access to current sources, with a human accountable for accuracy. The static PDF sitting in Highspot was failing before AI showed up; AI only made the alternative impossible to ignore. The governance problem solves itself the same way: a named human reviews what the model surfaces, checks the sources, and owns the claim before it reaches a rep. Pull up your next renewal. Find the CI tool line. Before you re-sign, ask the two people who are supposed to use it when they last opened it, and what they reached for instead. If the honest answer is a chatbot, you already know what that line item is: shelfware you're about to pay for again.