What it is
One place that holds what your company is and what it may claim
Your company, written down so your AI can use it.
Cadence is the system I built to do it. One place that holds what your business actually is, what it's allowed to claim, and where every claim came from. Anything your team asks an AI to do starts from that instead of from a guess.
Most AI output comes back generic because nobody has written the company down. I'm Jonathan Engle. I write it, wire it into the tools your people already open, and stay on to keep it true.
What you get
Documents you own, on a private site, wired into your tools
Where it starts
A free consultation
What I do after
Keep it true, and build from it
What it changes
First drafts stop coming back wrong
What your team gets back is about your company, in your language, with your positioning. Reviewing the work turns into approving it.
More of your people can use AI on real work
Once a register decides what may be claimed, a junior can draft something a customer will read without anyone policing every sentence for invented numbers.
The company stops depending on who remembers what
When the person who holds your positioning in their head leaves, the positioning stays. So does the reasoning behind it.
Each new thing costs less than the last
Your tenth page doesn't need the research your first one did, because the answers are already written and every asset draws on the same set.
What you get
- The company on the record. Positioning, who you sell to, the offer, pricing, brand voice, competitors, how sales runs. One version of the business, so two departments stop describing it differently.
- Your market, researched. The category you sell into, written up next to the company itself, so nothing anyone drafts has to guess at where you sit in it.
- A register of claims. What your company may say and what it may not. Nothing goes out carrying an invented statistic, a credential you don't hold, or a customer's name you never cleared.
- The briefs. The written instruction behind each page, campaign, or deck. The thinking survives the person who did it.
- The map. Which document feeds which. Change your pricing and you find out what it just broke.
You own all of it, and it stays readable long after any engagement ends, because none of it is locked inside software. The larger of the two companies I've written down, both of them mine, runs to forty documents about the business and thirty-two about its industry.
What you can buy
Free
Consultation
Where we work out whether any of this is worth doing.
- Where your AI output is failing now, and why
- Your score against the four conditions
- A path worth taking, or an honest no
Fixed price, fixed window
Build
I write your company down and connect it to your tools.
- Intake from your files, plus interviews on what they don't cover
- The company documents, the market research, the register of claims
- Your private site, and the connections into the tools your people use
- The urgent pages and campaigns you came for, built during or right after
Monthly, three-month minimum
Management
I stay responsible for the documents being true.
- Reconciling what your teams hand me, and resolving what they disagree about
- Keeping every document current as the business moves
- At the higher level: sitting with each team to find the workflows worth building, and teaching your people to write their own briefs
Monthly, alongside management
Execution
The work built out of the documents.
- Pages, campaigns, playbooks, decks, content
- Each one briefed against the documents, and checked against them before it ships
- Sold with management rather than on its own, including when budget is the reason
What sets the size
The number of functions the documents have to serve. Each one adds its own subjects to write up, its own contradictions to reconcile, and its own people to train. Marketing, sales, customer success, product, operations, finance, people, engineering, executive.
| Functions served | What that looks like | What grows with it |
|---|---|---|
| Up to two | Go-to-market only. The startup shape. | The shortest build. Two teams rarely contradict each other in writing. |
| Up to four | Go-to-market plus customer success or product. | More subjects and more interviews, and a first round of settling what two departments each believe. |
| Five or more | Company-wide. | Settling disagreements becomes most of the job, and the monthly work carries more training than maintenance. |
I price against your scope rather than posting a number I'd have to walk back. Bring me the scope on the call and I'll give you the number on the call.
The four conditions
The consultation scores you against these, and each low score names a specific piece of work.
Codified
Written down, outside people's heads, in a form a machine can read.
Current
Somebody owns keeping it accurate, and you hear about a stale document before a customer does.
Governed
You've decided in writing what your company may claim and what it may not.
Connected
Your people reach it inside the tools they already open.
How an engagement runs
-
Consultation
Free. I walk your team through the four conditions, we score them together, and we work out which of the three paths is worth your money.
-
Agreement
Your obligations in one column, mine in the other, both dated. Hit both on time and part of the fee comes back to you. Miss my own dates and I absorb it. Your column is a file dump and a round of interviews.
-
Intake
You hand me your files. I interview your people only on what those files don't contain, so nobody spends an hour telling me something Cadence can research.
-
Build
Documents appear on your site one at a time. You read each one as it lands and tell me what's wrong while it's cheap to fix.
-
Handover
I connect the documents to your tools and teach your team to work from them, starting with how to write a brief.
-
Keeping
I take over responsibility for the documents being true, and build the next round of work out of them.
How the method works
Documents by role
Every document has one job, so an AI loads the ones its task needs, in full. It never has to fish for whichever paragraphs happened to look like the question.
A source on every claim
Each document records where its facts came from and the date somebody last checked them. You can always trace a sentence back to who decided it.
A register of what you may say
One file holds the claims your company is approved to make. Anything absent from it doesn't get said, which is what lets you put AI near customer-facing work.
A map of what depends on what
Change your pricing and Cadence names the documents that just started contradicting it. You find out on the day rather than from a customer.
Checking before it ships
Cadence verifies the claims in a finished draft against the documents it was built from, and keeps a record of what it checked.
Two layers of knowledge
Your company sits in one layer and your industry in another, so a new page can draw on both without either one leaking assumptions into the other.
Where it connects
I wire the same folder into the places your people already work: your team's ChatGPT or Claude account, whatever your engineers use to write code with AI, and every page, campaign, playbook, or deck built from then on.
The keeper role
I maintain your company's brain. That's the whole job.
Every team keeps the workflow it already has and points it at the same documents. Nobody on your side has to learn my software, because what they interact with is the writing.
A consultant leaves with the knowledge in their head. I leave it installed in your company.
Part of keeping it true is settling what your teams disagree about, which is strategic work wearing administrative clothing. You rule on it, I write down what you decided, and we name the executive who makes those rulings in the agreement.
Where it lives, and what happens when I go
The documents sit in a repository you own from the first day, and the site that renders them is password-protected. Nothing about the deliverable depends on software I run.
I'm one person, which is a real risk to weigh before you buy an ongoing relationship. The exit packet is how I keep that from being a fatal one. Whenever you stop working with me, you leave with:
- Every document, and the full dated history of how each one changed.
- The index: one file that tells any AI what's in the folder, which document answers which question, and what it may never claim.
- A guide written for your team, including the prompts that only work because this exists.
- A short manual for keeping it accurate, a report on what's current and what's gone stale, and an honest note on what rots first and how fast.
- What to look for in whoever you hire to take it over.
I keep the method, the checking, and the judgment that goes into every brief. That's what you're paying me for.
Against the alternatives
| What you can buy | What you get | Where it stops |
|---|---|---|
| Licences for everyone | Access | Everyone can log in, and the output doesn't change |
| Prompt training | Better questions | Your team asks better questions and the model still doesn't know your business |
| A chatbot over the wiki | Search | It answers confidently out of pages nobody checked and nobody approved |
| A fine-tuned model | Your house style | It picks up how you sound and still doesn't know your facts, and changing one fact means paying to retrain it |
| An AI lead on staff | Someone who can build | They'll ask you where the documentation is |
| Cadence | A company the model actually knows | Somebody has to sit down and write it |
Buy the last one first and the other five get more out of your money.
Where this starts
A free call. I run a side-by-side test on your own company, ask a plain AI a question and then one that's been handed your documents, and you read both answers. It takes a minute, and it decides whether any of the rest is worth your money.