Case study · Board advisory · Australia

Board knowledge, one question away. And it never leaves their hands.

CHAIRS Global is an independent Board leadership organisation. It advises Chairs and the Boards they lead on difficult people, leadership and governance decisions, backed by its own proprietary intelligence drawn from Chairs around the world. We built AskCHAIRS, an assistant that answers members from that intelligence in plain words, with sources, running entirely inside CHAIRS Global's own AWS account.

The problem: Members could not reach the knowledge. The quick fix meant giving up control of it.
What we built: AskCHAIRS. Plain word answers, with sources, on their member site.
The result: Live and in use. Nothing, not even the AI model, runs outside their account.
Who owns it: CHAIRS Global. Their account, their code. It runs without us.

Book an AI Kickoff

Change the industry. The problem is the same.

Professional bodies
Research institutes
Law firms
Industry associations
Private banks

If your own knowledge is what sets your advice apart, your AI assistant should not need a copy of it on someone else's servers.

The knowledge existed. Reaching it was the work.

Before

4steps

Years of writing, expert interviews and member conversations. Some on a hard drive, some on the member site. You had to know it was there to use it.

1Know the piece exists
2Work out where it is kept
3Open the document
4Read until you reach the part you wanted

A director with a question late on a Sunday reaches none of it.

After

2steps

Members ask on the member site, in their own words, and get the answer there.

1Ask in plain words
2Read the answer, and its sources

Work out where it is kept
Open the document
Read until you reach the part you wanted

Every answer names the documents it used, so members can read further.

The problem

Why they turned down the quick option.

They did not come with a failed project. They came with a working option they had decided not to take.

There was already a way to do it

A deal was in place for someone else to build this, faster than starting again.

Their material would have sat outside

It meant their material living on computers CHAIRS Global did not own or control.

They chose control over speed

They walked away rather than let a copy of their work sit where they could not check it or take it back.

Then they asked for the version they own

The same thing, built inside their own AWS account, with all the code handed over.

The two things they asked for

Our material never leaves computers we control.
Members should never be able to tell it was not built by us.

The second is not vanity. A director asking a frank question needs to feel they are asking their own organisation, not an unknown supplier.

It was never about cost. It was about who holds the keys.

Live today

What it does now.

Answers from their own material

Their writing, member conversations, expert interviews and outside material, kept as four separate collections.

You choose what it reads

Members can pick which collections to search. Interviews only, or any mix.

Shows where answers come from

Every answer names its documents. Anything from outside their material is marked as outside.

Remembers you, privately

Conversations are saved for each member, follow the thread, and are never visible to other members.

Changed by settings, not code

Tone, rules, opening questions, access and model are all set from a control panel. No coding needed.

Every conversation can be read back

Their team can review any conversation, its running cost, and what members are asking about.

It wears their name, not ours.

It carries CHAIRS Global's name and look. Nothing a member sees says Avesta Labs.

Where each piece actually runs.

One wall. Everything that touches their material sits inside it, the model included.

Outside

Members, on the site they already use

The assistant appears on their member site, only for logged in members. None of their material travels with the page.

Inside CHAIRS Global's own AWS account

Their document library

Every document read in and prepared
Four collections, kept separate
The search index, owned by CHAIRS Global

The assistant

The link from the assistant to each collection
The chat window members use
The control panel their team uses

The model

The model runs on AWS Bedrock, in the same account
Claude Sonnet 4.6 answers today
Changing model is a setting, not a rebuild

Saved conversations, usage records and the search index all live in the same account. Nothing ever left, so nothing needs moving back.

Who owns it

Most AI projects leave you depending on whoever built them.

This one was built the other way round. It runs in their own AWS account, and day to day changes happen on a control panel, not in a support ticket.

What their team changes without us

1Add documents, and sort them into collections
2Rewrite the assistant's instructions and its rules
3Move to a newer Bedrock model when one arrives
4Decide who has access
Without asking us

None of it needs code or a quote from us, so the assistant does not go stale waiting on a supplier.

The part most suppliers leave out

The code, the set-up files, the release process and the running instructions all pass to CHAIRS Global after go live. If they never speak to us again, it keeps working. That was agreed at the start.

Why it mattered

What was at stake.

Its users chair and sit on the boards of listed, private, public and purpose-led organisations. Their questions are rarely small and often not yet public. The intelligence they ask about is CHAIRS Global's own, drawn from the lived experience of Chairs and found nowhere else. If either leaked, it could not be undone. So the rules were agreed in writing before any code was written.

Every part of it had to be:

Kept in their account

Everything runs in an AWS account CHAIRS Global owns, the model included. No second copy anywhere.

Clear about its sources

Every answer names the documents behind it, so a member can go and read them.

Private to each member

Your conversations are yours. No member can see what another member has asked.

Honest when it does not know

Asked something the collection does not answer, it says so instead of filling the gap.

The hard part

Everything could sit inside their account, except the thinking.

Documents, search and conversation records can all sit in an account you control. But to answer, the question and the relevant passages must go to an AI model. That handover is where the promise usually breaks.

The gap

The step nobody puts on the diagram

Many assistants keep your files in place, then send the question, private passages and all, to a model on someone else's computers.

What we did

We ran the model in the same account

The model runs on AWS Bedrock inside CHAIRS Global's own AWS account, so the question never leaves. Claude Sonnet 4.6 answers today, and switching to a newer model is a setting, not a rebuild.

Keeping the files private but sending them out to get an answer protects nothing. It only moves where they can leak from.

Questions first, then the build

How we worked.

01Three rounds
Before any code
They asked what done looks like, what it does with no answer, how it is kept safe, what it costs to run, and how they could walk away. Three rounds of written answers went into the contract.
02Weeks 1 to 2
Set up in their account
Search set up in their own AWS account, their material split into four collections, the first documents loaded.
03Weeks 1 to 5
Built the assistant on top
The chat window for members and a control panel for their team. Rules, tone, sources and model are settings, not code.
04Weeks 3 to 6, then live
Put into their member site
Added to their existing member site, behind the login members already use. Tested, signed off, live.

What the questions changed

Three things moved from later to now.

Saved history for each member, more document types, and control over when the assistant may look beyond their own material. All three were planned for later. They needed them from day one, so we added engineers and kept the date.

Every one of those questions was cheaper to answer on paper than halfway through the build.

Scope

What we left out on purpose.

Four things were asked about and kept out of the first stage, each for a reason.

Spreadsheets: Tables of figures need looking up, not searching for meaning. Adding them would have added noise, not coverage.
Mandarin: General models are noticeably weaker at Mandarin. It needs its own research and testing, not a rushed add-on.
Members uploading their own files: A different privacy problem to a shared collection. It waits until that private space is designed properly.
Live search across their website: Their call: the benefit was not clear yet at their question volumes.

The build leaves room for all four, so any of them can be added later without pulling it apart.

Start with the knowledge you would never put on someone else's server.

A short workshop, then a trial on your own material in your own environment. You finish with a working assistant and a clear view of whether to go further.