AI slop is a setup problem, not a talent problem
The fastest slop test: the logo swap. If a competitor could drop their logo onto your draft and lose nothing, it was never going to convert. That generic, "in today's landscape" opener is the tell. Start over.
AI never had a chance, because you didn't give it context. No customer language, no positioning, no product specifics, no internal docs. The model fills the gaps you left open with category-generic filler, and a human inherits the cleanup.
More context isn't the answer. Compressed context is. Dump 40 docs into AI and it hallucinates. The skill is compression: turning scattered internal info into tight artifacts the model can extract from reliably. This is the step almost everyone skips.
Bottom-of-funnel is where specificity pays. The closer you get to the buying decision, the more specific you have to be. High intent means you hit the pain immediately, then bridge pains to benefits to capabilities in a single piece.
Content is no longer king. Context is. You can produce all the content you want. If it lacks substance and context, it won't get cited, and it won't book the attributable demos everyone is actually after.
Best practices and key learnings
The session ran strategy first, then a live build in Slate. Here's what's worth stealing.
Stop starting with the content
Lashay's core reframe: people start by writing, and that's exactly why the output is thin. AI can't invent specificity it was never given, so you build from the inside out.
- Start with the innermost layer: cross-functional communication, so the knowledge actually flows between teams.
- Move outward to consistent documentation, compressed into artifacts the model can pull from.
- Only then write the content, at the outer layer, where all that context finally gets expressed.
"You do not start with the content. AI is not a good enough writer to start from scratch with no context." (Lashay Lewis)
Know what good bottom-of-funnel content actually looks like
Before you automate anything, you need to know what "good" is, or you can't spot slop. Lashay ran through the checklist she uses to audit BOFU pages.
- It's relevant to a specific persona, contextual to the use case, and it proves capabilities with real examples, images, and GIFs instead of just claiming them.
- It hits real prospect pain points immediately, then connects pains to benefits to the capabilities that power them.
- It explains what competitors do well without talking down to them, because AI cites your competitors to verify you, so the relationship matters.
"Don't just tell me you do something. Prove it to me. Show it to me in a workflow, on the page." (Lashay Lewis)
Your slop problem is usually a silo problem
Why is BOFU content so hard? Because it needs input from every team, and in most orgs those teams don't talk. Sales knows the objections, CS knows the accounts, product knows the capabilities, and marketing is supposed to fuse it all.
- Recognize that siloed communication, common in bigger orgs, starves your content of the specifics it needs.
- Create a flow of information across teams: sales to marketing, marketing to product, product to CS.
- If you can't fix the communication as a marketer, at least get access to the documentation each team holds.
Compression is the step everyone skips
The mechanics matter here. Messy internal docs scattered across Google Docs and HubSpot get centralized into a knowledge base, then compressed into artifacts, then turned into collateral. The middle step is the one people miss.
- Gather anything that describes what the product is, what it does, and who it serves: ICP docs, personas, case studies, sales decks, battle cards. Make sure it's accurate, because outdated info poisons everything downstream.
- Compress it into tight reference artifacts the model extracts from reliably, rather than dumping raw docs at it.
- Pull in live first-party data through API connections where you can, since first-party data is gold.
Build content like Legos, not essays
Lashay creates content modularly. She breaks every piece into pain points, features, benefits, and capabilities, then stacks and re-stacks those blocks across formats. In her example BOFU article, the H2s were product features mapped to specific pains, not keyword-stuffed headers, so the page read like a story.
- Break content into modular blocks you can reuse in a blog, an email subject line, a LinkedIn header, or a sales one-pager.
- Use framework-based prompting: lead with a dynamic framework and let your internal documentation fill the modular blocks, instead of prompting "write a blog post around this keyword."
- Remember the front-end task stays simple because all the work lives in the framework on the back end. Context matters more than the prompt.
"I have frameworks for every piece of content I create. The internal documentation fills in the blocks." (Lashay Lewis)
Record your sales calls, because that's where the gold is
The live build brought it together. Lashay ran roughly 13 to 14 sales-call transcripts through the same framework in Slate to parse out a customer-language library, formatted by account, role, company size, and date, with timestamps so every pain traces back to who said it.
- Mine sales calls for the exact pains and language your audience uses, then plug those modular pains into any collateral.
- Let the analysis surface your real persona and group personas by the problems they raise most.
- Reuse the extracted themes as email subject lines, LinkedIn headers, and even new workshop ideas.
"People don't take recording sales calls as seriously as they should. It's literally gold." (Lashay Lewis)
How to put this into practice
If you're starting from close to nothing, the minimum viable context is smaller than you think. First question: are you recording your sales calls? If not, start today. Then gather roughly three pieces of documentation on your persona and ICP, three on your product (a demo transcript works if you can't reach the product team), and some competitor intel you can pull with a Perplexity deep-research prompt. Call it six to ten solid pieces.
Treat it as cumulative, not set-and-forget. Positioning and product roadmaps move fast, so keep feeding and refining the system as things change. And if you've become a full-time AI copy editor, build an editing artifact: put a strong existing piece next to a fresh draft, ask AI to learn the difference, and run future drafts through that filter before they reach you.
The context era is here
The takeaway the room left with: if your AI is fed nothing, it writes generic slop. Feed it the right context about your company, your positioning, and your writing style, and it starts to sound like you and sell like you.
The teams that win this next phase won't be the ones producing the most content. They'll be the ones feeding AI what only they know. Content is no longer king. Context is.
Want to feed your customer language, positioning, and internal docs into a system built for it? Book a demo or start a free trial at slatehq.com.
How Slate helps teams beat the slop
Slate is a content automation platform for SEO and content teams, with a knowledge base built directly into the tool. It tracks your AI and search visibility, lets you compress internal docs into reusable artifacts, runs framework-based workflows across as many topics as you need, keeps everything on-brand through the Brand Kit, and connects to Claude over MCP so you can move work between them. It's the context layer most AI tools are missing.
About Lashay Lewis
Lashay Lewis built BOFU.ai after spending years as a content strategist watching well-written content fail to convert because it lacked product depth and structure. She developed the BOFU framework to give lean content teams a systematic way to produce high-quality, product-led content at scale.
She's a frequent podcast guest and speaker on B2B content strategy, and regularly shares frameworks and breakdowns with the content community on LinkedIn.
In this session, she's bringing her clearest thinking on beating AI slop alongside a live Slate build you can put to work immediately.
