How it works

Your AI is brilliant. It doesn't know your business. Yet.

Two things fix that: a knowledge base and skills. Here's the whole model, in plain language, in five minutes.

Scroll — five minutes, no jargon

One question, asked twice

Same question. Two very different answers.

Ask "how do we onboard a new client?" A generic assistant answers from the internet — smooth, confident, and not your process. Grounded in your knowledge base, it answers from your pages. And shows them.

Any AI
How do we onboard a new client?

"Typically, firms onboard new clients with a kickoff call, a welcome packet, and a regular follow-up cadence to ensure satisfaction…"

Sounds right. Isn't yours.
Yours, grounded
How do we onboard a new client?

Welcome pack goes same-day. Kickoff booked inside 5 business days. Signed agreement filed before any work starts.

client-onboarding-processnew-account-checklist
Cites your actual pages.

Part one

A knowledge base is your company's shared brain.

Everything your company knows — processes, policies, products, roles — written as plain pages in one place. Your team browses it where they already work. Your AI reads it before it answers, and writes new pages as it learns. Including the knowledge that today lives only in people's heads.

CRM Email Meetings Files People's heads what your company already has client-onboarding.md The knowledge base one page per thing your company knows Cited answers Drafts, your voice Skills and it writes back what it learns

Every AI write is checked before it lands — nothing enters your company's record unreviewed.

Google published an open, vendor-neutral standard for exactly this layer — the Open Knowledge Format, June 2026. Early draft, no formal multi-vendor adoption yet: validation of the approach, not a finished standard.

Part two

A skill is an onboarding guide for your AI.

Exactly how you'd train a sharp new hire: here's how we do it, here's what to read, here's what good looks like. Written once. Runs every time, your way. It's how the companies building these AI models package expertise — we write yours.

/client-meeting-prep
Instructions
How we prep, step by step — agenda, history, open items, talking points.
Knowledge
Reads client-history meeting-notes service-policies from the shared brain.
Action
Drafts the brief and files it where the team expects it.
Runs inside the chat your team already uses. One command.

Part three

Skills ship as a plugin — your team's cookbook.

One command each. Every department gets its own set. Same shared brain underneath all of them.

Sales

  • Prep any meeting in one command
  • Draft the follow-up before you're back at your desk
  • Keep every deal's story current

Finance

  • Run the month-end checklist without chasing anyone
  • Draft collection follow-ups that cite the invoice
  • Explain any variance in plain language

Operations

  • Answer "how do we do this?" with the actual SOP
  • Onboard new hires from one playbook
  • Draft vendor emails that match policy

Customer Support

  • Triage tickets against real product docs
  • Answer with citations, not guesses
  • Summarize any escalation for the team

The part that matters most

It gets better every week. That's the point.

Every run teaches the system something. Edge cases become pages. New pages sharpen every skill. The loop feeds itself.

Team runs skills Edge cases + feedback captured The brain grows Skills get sharper 247 your team's pages, growing weekly

Why it compounds

Tools plateau. Systems compound.

A chat that starts from zero is exactly as good in month six as on day one. A brain that remembers isn't.

how much it knows about your business month 1 month 6 chats that start from zero a brain that remembers what compounds

Our point of view

Two kinds of AI. Most companies only buy one.

Efficiency AI

Makes what you already do cost less. Measured in hours returned and errors caught. Ceiling: your current cost base.

  • Inbox triaged before you sit down at 6 a.m.
  • The month-end checklist runs itself
  • A field team's week of inspection-video review, done as a background skill
Opportunity AI

Reaches revenue and service you couldn't touch before. Measured in what wasn't there last quarter. No ceiling.

  • Every client gets the depth of attention your best client gets
  • Quotes back in hours — bids you never had capacity to make
  • Your people get self-serve answers instead of waiting on you

Efficiency pays for the system. Opportunity is why you build it.

Adoption

Three steps per person. About five minutes.

No git, no terminals, no new tabs. Everything else runs centrally. Your team keeps working exactly where it already works.

01

Join the shared drive

The knowledge base shows up as folders in the drive your team already opens every day.

02

Install Claude + your plugin

The desktop app, with your team's cookbook of skills pre-loaded. One installer.

03

Accept one folder

One permission grant. Done — every skill and every page is now live for that person.

Not a concept

Running today.

Environmental services

Knowledge base + team plugin delivered. The business-development team runs its own skills.

Payments — GTM team

A 33-skill plugin in daily use across the sales motion, from call prep to pipeline review.

Our own consultancy

Every deal, decision, and deliverable lives in one shared brain. We run on what we sell.

The method

Discover. Map. Build. Deliver.

01

Discover

Interviews plus a read of your systems. The real workflows surface.

02

Map

The knowledge architecture: what's written, what's in heads, what's missing.

03

Build

The knowledge base and your team's plugin, on your stack.

04

Deliver

Onboarding, adoption, and the weekly improvement loop.

First working system in weeks, not quarters.

Next step

See it on your own workflows.

Bring one meeting, one process, one team. We'll show you what your version looks like.

Tyler Massey
Prescyent · prescyent.ai · tyler@prescyent.ai