The Knowledge Substrate
It doesn't just read your docs. It knows your business.
Most AI tools take a prompt, glance at a handful of files, and hope for the best. Your Solutions Engine is different. Behind every page, deck, and proposal it produces sits a living map of your entire business — your brand, your doctrine, your playbook, your products, the way your best people actually pitch — and every claim it puts on the page traces back to a specific source you can inspect. This page explains what that means and why it changes what AI-assisted pre-sales can look like.
1 · What we take in
Everything you use to teach a new hire.
Documents, in your voice
Product one-pagers, positioning docs, competitive teardowns, case studies, FAQs, pricing rationale, ICP briefs, sales enablement — anything you'd hand a new pre-sales hire on day one.
Brand & visual doctrine
Logos, colours, type, imagery rules, layout conventions. Your engine builds every page inside your visual system — not a generic template.
Voice & tone
Signature phrases, banned phrases, the register you use for a CFO vs an engineering director. Learned once, applied everywhere.
Reference material — on your terms
Reference decks and past pitches are treated as evidence of how you pitch, not as sources of fact. Pulling specifics out of them is always a deliberate, curated step — never a background crawl that could leak a previous client's information into your next proposal.
2 · How it holds all of it in mind
A map of your business, not a stack of files.
The engine reads every document you give it and turns your business into three overlapping views. Any one alone is what makes AI feel forgetful. Together they make it feel like it has actually worked at your company.
Meaning
Semantic map
Every idea in your corpus is placed on a map by what it means, not what words it uses. Ask for "how we win competitive replacements" and it finds the relevant paragraphs even if none of them use those words.
Things
Structured memory
People, products, customers, competitors, proof points, numbers. The engine notices what matters in your business and remembers it as first-class things, so "our SMB motion" or "the platform we compete with in Europe" reliably resolves to the right specifics.
Themes
Recursive summaries
Related material is clustered into themes, each with its own living summary that rolls up into an Company Memory — the one-page distillation of who you are, what you sell, and how you sell it. Always current, never stale.
3 · How it thinks
Not one prompt. Many passes. Always a critic.
A working set, chosen for the job
Before writing anything, the engine assembles a task-specific reading list from all three views — the meaningful passages, the named things, the relevant themes — bounded to what a premium model can actually reason over. No dumping everything into one prompt and hoping.
Draft, merge, critique, revise
Sections are drafted in parallel, then merged by a premium model. A separate critic inspects the result, flags anything that doesn't trace to a source, and the engine rewrites those spots. Long generation is a feature — it's the engine doing more thinking, not stalling.
Minimal-edit doctrine
When you refresh the Company Memory or Pitch Style, the engine only changes what new knowledge actually requires. Your voice, your phrasing, your carefully-tuned wording stays. It edits like a good editor, not like a rewriter.
Nothing important happens in one call
Every meaningful output goes through a durable pipeline — map, reduce, critique, revise — so a single model hiccup can't produce a shallow result. Quality first, speed second, cost third.
4 · Trust, made visible
Every claim, back to a source. In one click.
The single hardest thing about AI-generated pre-sales content is trusting it. Your engine solves this at the substrate level, not with a disclaimer.
Grounded
Inline citations
Every non-trivial claim in your Company Memory and Playbook carries a small superscript. Hover to see the source document, section, and the exact passage the engine drew on.
Honest
Ungrounded claims flagged
If the critic can't tie a claim back to something the engine actually knows, it says so out loud — instead of quietly making it up. You decide whether to keep, edit, or delete before you accept.
Fresh
Freshness by design
Every synthesised document records the exact evidence it was built from. When that evidence changes, the engine knows what's now stale and re-thinks only what needs re-thinking.
5 · What it feels like in use
Like your best pre-sales person, on tap.
You bring a prospect. The engine researches that specific brand, weighs it against everything it knows about your business, and produces a solution — as a web narrative, a deck, a proposal, a companion page for a technical deep-dive — in your voice, on your brand, grounded in your actual capability. Every subsequent pitch makes it sharper: the corrections you make, the phrases you prefer, the moves that land are folded back into the substrate the same way a good team learns from every deal it runs.
That's the difference. Not a generator that writes plausibly. A Solutions Architect that knows your business, reasons over what it knows, and shows its work.