Explore

Business AI resource

The Context Stack: How to Make AI Answer as Your Business Instead of as the Internet

By

A friendly pencil-drawn robot with a soft rounded body and warm expression welcomes you beside working papers and a yellow pencil.

AI does not know anything about your business. Every time you ask it something it starts from nothing, like a new employee on their first morning who has never seen your prices, your templates or your standards. The Context Stack is everything you stack around the request so the answer comes back as your business. Five layers: Context, Task, Source, Rules and Output. Build it once for a job you repeat, and you stop rewriting the same work from a blank page.

Key takeaways

  • The Stack is the induction pack you would hand a new employee, written once.
  • Five layers, bottom to top: Context, Task, Source, Rules, Output.
  • The Source layer is the one almost everybody skips, and it is why their output reads like a brochure.
  • The Rules layer is what stops invented prices, dates and promises.
  • When output comes back wrong, you do not blame the tool. You ask which layer is missing.
  • A Stack is reusable. The second job costs nothing to set up.

The five layers

Layer What it holds On a quote On a customer reply
Context Who you are, who you are talking to, the situation A Coast roofing business quoting a homeowner A clinic replying to a first-time enquiry
Task The job you want done, in one sentence Turn site notes into a quote draft Draft a reply that answers the question asked
Source Your own paperwork and approved facts Site notes, measurements, current price list Services, fees, hours, what to bring
Rules What it must and must not do Use only supplied prices, never guess a measurement, flag gaps Never give health advice, escalate anything clinical
Output The shape of the finished thing Scope, itemised table, GST separate, total, exclusions Four short paragraphs and a next step

Stack all five and you get your document. Skip Source and Rules and you get confident writing about nothing, which is exactly what people mean when they say they tried AI and it was no use to them.

Why it is called a stack

Because the layers sit on top of each other and the whole thing falls over if one is missing. Context without Source produces a plausible document with invented facts. Source without Rules produces facts with invented conclusions. Rules without Output produces a correct answer in the wrong shape, which somebody then reformats by hand and quietly stops using.

The word also does a second job. It tells you the Stack is a thing you own and maintain, not a clever sentence you typed once and lost in a chat window.

Building your first one

Pick a job that repeats weekly, already has written inputs, and gets checked by a person before it leaves the business. Quoting, enquiry replies and the weekly report are the usual three.

One. Write the Context in two or three sentences. Your business, your customer, the situation.

Two. Write the Task as a single sentence naming the document you want. Not "help me with quotes". Turn these site notes into a quote draft.

Three. Gather the Source. Actual files: the price list, the templates, the approved answers, the measurements. This is the work. It is also the step that makes everything after it easy.

Four. Write the Rules. Start with the missing-information rule: flag anything not covered rather than filling the gap. Then the limits that matter in your trade. No invented prices. No promised dates. No advice you are not licensed to give.

Five. Describe the Output. Sections, order, what has to appear, what must stay separate, how long.

Then run it on a job you have already completed, and compare the draft with what you actually sent. That comparison finds your weak rules faster than any amount of theory.

The diagnostic

The reason to name the layers is that it turns a vague complaint into a specific fix.

Symptom Missing layer
Reads generic, could be any business Context
Answers a different question Task
Invents prices, figures or details Source
Fills gaps instead of asking Rules
Right information, unusable format Output
Confidently wrong about your industry Source, then Rules
Good once, useless the next time No Stack at all, just a chat

Work through the full diagnostic.

Where the Stack goes after the first one

Once one Stack works, the rest is repetition. Same five layers, different Source and Rules. A quoting Stack, a reply Stack, a reporting Stack. Teams then share one Stack per job so five people stop improvising five different versions of the same task.

The layers also need maintaining. When your prices change, one layer changes. That is the practical advantage of writing it down instead of retyping instructions into a chat box every time.

Learn it on your own work

The workshop is a full day at Avani Mooloolaba Beach Hotel on the Sunshine Coast, A$495 excluding GST, capped at 15 people. You bring one repeated job and a safe example of it, and you leave with a built and tested Stack plus the written instructions.

Private onsite training builds Stacks with your team at your workplace across South East Queensland, scoped after a conversation about your tasks. See the workshop, or enquire about team training.

For the sentence-level detail of writing each layer, read the five-input guide.

Frequently asked questions

What is the Context Stack?

The five layers of information you supply so an AI assistant produces your work rather than generic writing: Context, Task, Source, Rules and Output. It is the induction pack you would hand a new employee, written once and reused.

Why does a written Stack beat a one-off prompt?

A prompt is one sentence typed into a box and lost. A Stack is written down, reused and maintained, and it names which part to fix when the output is wrong. The difference shows up in month two.

Which layer matters most?

Source. Supplying your own prices, templates and approved answers changes output more than any other single change. It is also the layer people skip, because it is the only one that feels like work.

How long does it take to build a Stack?

An afternoon for the first one, including testing it against a job you already completed. Later Stacks take under an hour because the Context and Rules mostly carry across.

Do you need special software?

No. A paid AI assistant account and your own documents are enough. Software matters when systems need to talk to each other unattended, which is separate engineering work. Compare training with consulting.

Does this work for any industry?

Yes, because every business has repeated documents built from information it already holds. The Source and Rules layers change by trade. Browse the industry pages.

Can a team share one Stack?

Yes, and sharing it is the point of writing it down. One Stack per job, agreed by the people who do that job, so output stops depending on who happened to type the request. How to set team rules.

What if the output is still wrong?

Use the diagnostic table above to find the missing layer, fix that layer only, then rerun the same example. Changing everything at once tells you nothing. Full diagnostic here.

How do you keep a Stack current?

Review the Source layer whenever prices, services or standard answers change, and the Rules layer whenever something gets through that should have been flagged. Maintaining workflows.

Where can I see a finished example?

There is a complete worked quoting example on the site with fictional figures, showing the source notes, the itemised draft and the checked arithmetic. See the worked example.

Choose your next step.

Learn to build it yourself, or have us build it for you. Training and implementation are different engagements, so enquire and we will scope the one that fits.

Take one of the 15 seats.

Wednesday 7 October 2026, 9am to 4pm at Avani Mooloolaba Beach Hotel, The Cove room, Mooloolaba. A$495 excluding GST, added at checkout.

Registration takes under a minute. Three quick questions, then where to send your place. Pay on the spot to hold the seat, or pay from the link in your email after we have spoken. Registration stays open until the seats are taken or the workshop starts.

Training your whole team? Explore private onsite training.