Agents at Work: A Week of APEX for Product, Marketing, and Sales Teams

Most of the work behind a product, marketing, or sales decision is assembly. Someone finds the battlecard, pulls the reviews, reads the win/loss notes, checks what the rival announced, and turns it all into something a colleague can use. Sales reps report spending 70% of their time on non-selling tasks, according to Salesforce's sixth State of Sales report, published in 2024.[1] In Crayon's 2026 State of Competitive Intelligence, 57.5% of respondents said more of their deals are competitive than a year ago. Asked how prepared their reps are for those deals, they gave an average score of 6.3 out of 10.[2]

Product and bid teams carry the same load. About 52 percent of product managers in ProductPlan's 2023 survey said their product strategy is primarily influenced by requests from senior executives or direct customer feedback.[3] In Loopio's 2024 benchmark, teams spent an average of 30 hours writing each RFP response.[4] The evidence is usually somewhere. Gathering it is what takes the week.

PYRAMYD puts agents to work on that evidence. APEX is a supervisor with 20 specialist agents behind it and 150+ tools. It works on every page, hands each part of a question to the right specialist, and lists the records and sources an answer used, rather than a citation for every sentence. Our earlier post explains how APEX is built. This one shows it at work.

Monday and Tuesday: a new battlecard, then the call

Monday: product marketing drafts a card for a new rival

A new name keeps coming up in deal reviews, and sales wants a card before the week is out. The product marketer opens Competitors › Battlecards, chooses New Battlecard, picks her product under Our Product and the rival's under Their Product, links the deal it is for, and chooses Generate.

The card streams in section by section, drafted from buyer reviews, news, the rival's website, the product features in the graph, and any context nodes she adds. Sections that come back empty are filled in the same pass. The card shows its numbered sources and a data confidence, so she can check a claim before a rep repeats it. One limit: generation does not read call recordings today.

Then she asks APEX to "refresh the win/loss on this card" and, once she has read it through, to "publish it." APEX hands the refresh to its battlecard specialist, which says when it is done, and publishes the card from the battlecard page.

Tuesday: a seller gets ready for a competitive call

An account executive has a call with a buyer who is also talking to that rival. In Accounts › Opportunities › Pipeline she chooses Ask APEX on the deal's row, and APEX opens with a ready-made briefing question for that deal and answers it. She follows up in the same chat:

• "Find our battlecard against them and give me the top three talking points." APEX answers from Monday's card and names the card and section it used.
• "Search win/loss interviews where buyers mention their pricing." APEX quotes the interviews, if your team runs them in PYRAMYD.
• "What does our CRM say about this account?" APEX pulls the account's context from your connected CRM. It is as current as the last sync, and a sync runs when you choose Sync now.

Her manager doesn't open the app. Working in Claude Desktop, the manager reads the same battlecard through PYRAMYD's MCP connection, with a scoped token made in Settings. Outside AI assistants, over MCP or ChatGPT Actions, can read battlecards, competitors, and signals this way. They cannot ask APEX itself a question yet, and full graph access over MCP is coming.

Wednesday: the product team asks why

Sales has a question for product: why do deals against this rival keep stalling on reporting? The product manager has a roadmap review on Thursday and wants evidence, not anecdotes.

She asks APEX, "Why are we losing deals to them?" The supervisor hands the question to a specialist, and the step line reads Consulting, with that specialist's name, while it works. The answer comes back in the same thread, with reasons and quotes.

Then she asks for the buyer's side: "What do reviewers dislike about their reporting, and which of our features answer that?" APEX searches reviews and product features by meaning, and answers with quoted reviews, their sources, and a list of matching features. Reviews belong in this decision because buyers read them: 54% of technology buyers consult user reviews during the buying journey, according to TrustRadius's 2023 B2B Buying Disconnect research.[5] One limit: the reviews in the graph run to mid-June 2026, so ask for dates when recency matters.

Before she closes the panel, she asks APEX to add a roadmap item for the reporting gap to her team's project, and she takes the quotes to Thursday's review.

Thursday: the bid team drafts an RFP

An RFP arrives with a short deadline. Loopio's 2024 benchmark puts the cost of this work plainly: at 30 hours a bid and an average of 175 bids a year, the average team spends 5,250 hours a year writing RFP responses.[4]

The bid lead uploads the RFP in Accounts › Opportunities, and PYRAMYD extracts every requirement and matches each one to a product capability. In the response workspace she chooses Fill with APEX, and APEX drafts answers across the RFP in one pass from your product's data in the graph. A progress card above the grid shows coverage, a weighted fit score, and risk counts, so the team starts with the requirements marked high risk.

In the APEX panel, the pre-sales lead asks, "How many requirements are still unanswered, and which competitors are in this deal?" and then, "Summarize the gaps in this RFP." The bid lead assigns the high-risk requirements to reviewers with due dates. The product manager from Wednesday reads the gap summary too: it lists what this buyer asks for that the product does not do yet.

When the team is done, PYRAMYD returns the buyer's own Excel or Word file with the answers filled in. Its PDF option downloads a printable page, not a PDF file.

Friday: marketing writes the week's brief

The head of product marketing closes the week with three requests to APEX.

• "What's new with our top three rivals this month?" (or the /news command). APEX searches news articles, digests, and alerts, and lists the main items with their dates and sources. New signals are collected every day, but new digests and alerts are paused for now, so the newest items in the answer are newly collected articles.
• "Create a document called Weekly rival brief." APEX creates it in Workspace › Studio › Documents, and she writes the brief from the items APEX found, each with its date and source.
• "Build a battlecards dashboard." APEX shows a preview of card counts, cards by competitor, the spread of confidence, and a list of cards she can filter. "Save it" keeps it in Workspace › Studio › Dashboards, and the low-confidence cards become next week's list.

Teams are already moving this way. In Crayon's 2026 survey, 80% of teams use AI to generate sales-facing competitive content, up from 61% in 2025 and 25% in 2024.[2] What the AI works from is the part worth checking. Here it is the same graph the rest of the week used, and every answer lists its sources.

What a week of agents adds up to

Nothing in this week was decided by an agent. The product marketer published the card, the seller chose what to say, the product manager chose what to build, and the bid lead chose where the reviewers spent their hours. The agents did the assembly: finding the card, the quotes, the reviews, the requirements, and the news, and showing where each one came from.

That is the split PYRAMYD is built for. Agents gather and draft from the evidence; people check it and decide. Build what matters. Position it to win.