How Bayley Stuart Got Its Whole Team Building in a Week

How Bayley Stuart Got Its Whole Team Building in a Week

A keynote, five days of side-by-side coaching and one working automation per person took Bayley Stuart from installing Claude to running its own workflows inside a single week.

Client: Bayley Stuart · Industry: Commercial property investment management · Location: Melbourne · Duration: One week

Where Bayley Stuart started

Most firms roll out an AI tool and hope adoption follows. Very few get every person building something of their own inside a week. That is what Bayley Stuart set out to do.

Bayley Stuart is a Melbourne-based commercial property investment manager and investor — in their own words, “invested in excellence”. A team of ten holding an Australian Financial Services Licence, running almost entirely on email and documents, with no in-house engineering function.

The work that consumed them was the work that repeats — recurring monthly and quarterly cycles running across funds administration, asset management and office operations. Necessary, exacting, and compounding into nothing. Each cycle moved documents between systems and people in a fixed sequence, and each one absorbed hours that a ten-person firm could not spare.

What they did not have was slack. In a team of ten, nobody can disappear into a week of training, and nobody is spare to maintain something afterwards. Whatever got built had to be built by the people who would use it, and it had to be working by Friday.

One keynote, then five days side by side

Propel ran a single all-company keynote to set the frame, then put a consultant on site for five consecutive days.

The method was deliberately not a training course. Every person on site was taken through building and setting up their own first automation, one to one, start to finish — against work they genuinely did that week, not a worked example. Asked afterwards what had been most useful about the week, participants named the format rather than any part of the curriculum:

“Sitting down with 1:1 training to work through a skill I wanted to implement.”

Bayley Stuart made that unusually fast, and the reason is worth naming. The firm arrived with its operations already in order. A mature, well-categorised file management system — disciplined folder structure, consistent naming conventions, documents where they were supposed to be. Email organised on the same logic, categorised and filed rather than accumulating in a single inbox. And processes that were genuinely documented: written procedures setting out who does what, in what order, and what each step hands to the next.

That matters far more than it sounds. Most of the cost of automating knowledge work is archaeology — establishing where things live, what they are called, which version is current, and what the process actually is once you strip away what everyone assumes it is. Bayley Stuart had done that work already. Automations could be pointed at real folders, real mailboxes and a real documented sequence from day one. There was no discovery phase, no clean-up project standing in the way, and the boundaries of what each automation could and could not reach could be drawn precisely, first time.

That discipline was tested rather than taken on trust. The firm’s invoice batch payment procedure was pulled apart against a live example mid-engagement to confirm the documented process matched what the team actually did. It did — which is rarer than it sounds, and it is why the next automation candidates could be identified in the same session rather than in a separate discovery exercise.

What got built was not file conversion. Each one is a workflow: a sequence of steps that used to occupy a person, now running end to end — triggered on a schedule or by a document arriving, moving through processing and filing, and pausing at the points where judgement is genuinely required. Scheduled tasks that fire without being asked. Pipelines that carry a document from arrival to approval. The work spans funds, asset management and office operations, and in each case the automation follows the firm’s own established process rather than asking anyone to adopt a new one.

Two things were built to outlast the week. The first is a skill library on SharePoint. Every automation is filed by category — mirroring the firm’s own business lines rather than any structure we imposed — and indexed in a reference list that version-controls what exists and updates itself as new work is added. The point is not tidiness. It is that in two years, when someone needs a workflow that already exists, they will find it instead of rebuilding it, and they will be able to see what changed and when. The second is a written FAQ capturing the friction points that came up more than once, including the expensive ones — like why “read all my emails” is a bad instruction.

One working habit mattered more than any single artefact: when someone brought a question, it went back to Claude with them, rather than being answered for them.

What was deployed

The engagement was a deployment as well as a coaching exercise. Across the week:

• Claude installed and configured for every member of staff on site, on desktop and web
• Microsoft 365 connector — admin-consented at tenant level through Microsoft Entra, in coordination with the firm’s external IT provider, giving Claude governed access to SharePoint and mail
• Claude add-ins for Excel, PowerPoint and Word, installed across the team
• HubSpot connector activated through the firm’s admin account
• Claude Skills, with a client-owned skill library on SharePoint and a self-updating, version-controlled reference index
• Claude Cowork for multi-step document and file workflows
• Plaud integration, connecting meeting recordings directly into Claude
• A discovery prompt issued to each person, which interviews them about their own working week to surface what is worth automating — for the people who did not yet know what to ask for
• Custom instructions configured individually for staff who wanted them

The results

Adoption

• 100% of staff on site built and deployed their own first working automation, one to one, start to finish
• 5.00 / 5 — “The week left me confident using Claude on my own work, without someone sitting beside me”
• 5.00 / 5 — “The skill or scheduled task we built together in my 1:1 is something I’ll keep using”
• 4.75 / 5 overall rating for the week
• Adoption spread beyond the individuals expected to lead it, with no internal mandate and no rollout program

Scores from the post-engagement participant survey, four responses.

Deployment

• Claude and the Microsoft 365 connector deployed across the team, admin-consented at tenant level
• Automations live across three functions — funds, asset management and office operations
• Workflow types deployed include batch payment processing from spreadsheet through to signed PDF, monthly asset management schedule reconciliation between PDF and spreadsheet, scheduled document conversion and filing, templated newsletter production, and lease review against the firm’s own precedent

Durability

• Client-owned skill library handed over at close: every automation filed by category and indexed in a self-updating, version-controlled reference list, so the firm can find and maintain its own work without us
• Written FAQ and best-practice reference produced during the engagement and left with the customer

Scale beyond the engagement

• A second session is booked for 29 September 2026 for the staff who could not attend the original week
• Participants named further automations they want built, unprompted — condition reports, board reporting, business cases, and rental reconciliation against the tenancy schedule in a live Modano model. Demand is coming from the team rather than being pushed at them
• After the engagement closed, the firm independently applied the same approach to its leasing workflow
• The 1:1 build-your-own-automation method has been adopted as Propel’s standard delivery model for subsequent Claude enablement engagements

What happened next

The engagement closed after a week. The most interesting part is what the firm did with it once we had gone.

Bayley Stuart turned the same approach on leasing — the gap between a conversation opening and a Heads of Agreement reaching the counterparty. Their own executed leases loaded as reference material. Provisions that carry real commercial consequence — early termination rights, incentive clawbacks, pro-rata repayments — surfaced for a decision rather than left to be caught. Every draft reviewed and approved by a person. Heads of Agreement are not legally binding, which is precisely what makes them the right place for an AI-drafted first pass.

Piers Jalland, Head of Funds and Transactions, frames the prize: “Two things kill deals — time and uncertainty.”

There is a second-order benefit nobody costed in advance. A Heads of Agreement that states plainly what is and is not included cuts the cost of the deal on both sides — the counterparty spends less time, and less legal money, working out what it is being offered. Clarity turns out to be a trust instrument.

Meanwhile the requests kept coming — condition reports, board reporting, rental reconciliation against a live financial model. None of it asked for by us.

“We’re a small team with no IT function, so the risk with anything like this is that it gets built for you and then sits there. That didn’t happen. Everyone here built their own thing during the week, which means we can keep going without needing someone to come back.”

— Piers Jalland, Head of Funds and Transactions, Bayley Stuart

The pattern that works

Bayley Stuart’s path — one keynote to set the frame, five days of one-to-one coaching, and one working automation shipped per person before anyone left the building — is repeatable for any document-heavy firm without an engineering team. Once someone has built one themselves, they know how to ask for the next.

It moves fastest where the fundamentals are already in place. An AI rollout travels at the speed of the operational discipline it lands on: ordered files, ordered email, and a process someone has actually written down.

If that’s the journey you’re on, talk to us at propelventures.ai/contact.

Most small firms are waiting for an AI strategy. Bayley Stuart did the opposite: they gave us a week, opened up their real work, and asked us to sit beside every person in the business until each of them had something working. It took five days.

Where Bayley Stuart started

Bayley Stuart is a ten-person property funds management business in Cremorne, Melbourne. Like most firms its size, the work runs through a familiar set of artefacts — leasing negotiations, heads of agreement, lease reviews, fund models, investor meetings, board and investor reporting. Structured, repetitious, and expensive in hours.

The specifics told the story:

• The leasing workflow — negotiation, minutes, heads of agreement, covering email — took days across multiple touch points, with leakage at every hand-off.
• A lease summary at acquisition cost around external legal fees.
• During a capital raise, three or four investor meetings a day generated actions that took a week or two to surface from the notes.
• A single question of the fund model took a day to turn around through an analyst. One question about the distribution impact of a proposed fund change had been open for two months.

None of that needed software. It needed the tool used well, used safely, and used the same way by everyone.

Governance first — because it has to be

Bayley Stuart holds an AFSL. ASIC, APRA, AUSTRAC, privacy law, cyber resilience — the regulatory floor is not optional, and it is not something you retrofit after the team has already pasted client data into a chat window.

So before the kick-off, their COO was already drafting an AI Framework and Policy with the firm's external compliance advisers, for sign-off by the compliance committee and the board. Seven principles: legality and regulatory compliance, fairness and non-discrimination, transparency and explainability, accountability with mandatory human oversight, privacy and data protection, security and resilience, and proportionality.

We worked to that document, not around it. Data residency, connector permissions, acceptable-use boundaries and record retention were settled with their IT provider in the same week the team was learning to prompt.

One embedded week

The engagement was deliberately small and deliberately deep: five consecutive on-site days for a fixed fee.

Monday — the whole company, in one room. A two-hour kick-off for all ten staff: the AI landscape, security and data confidence, and live demonstrations built beforehand on Bayley Stuart's own leases, heads of agreement and fund model. Not sample data. Their data. The afternoon went on licences and connectors.

Tuesday to Friday — desk-side, one person at a time. Our trainer on site 9 to 5, working shoulder to shoulder with each staff member on their live work, and building a skill every time a repeatable task surfaced.

By Friday the firm had:

• The Heads of Agreement skill — negotiation transcript or leasing minutes in; updated minutes, a drafted HoA and the covering email out, in one pass.
• The Lease Review skill — any lease in, the firm's standard one-page analysis out, with early termination rights, incentive clawbacks, pro-rata repayments and rent-review risks flagged.
• A batch-payment skill turning an Excel schedule into signed PDFs, built with the Head of Funds and Transactions.
• A newsletter builder, a monthly PDF/Excel cross-checking skill for the asset-management schedule, and a labelled email-to-PDF pipeline with a scheduled run — each built one-to-one with the person who actually does the job.
• Claude connected to Microsoft 365 for everyone on site, plus the Office plugins, the CRM connector, and the Managing Director's meeting recorder wired straight through to Claude.
• A shared skill library in SharePoint with an auto-updating index Claude can read directly, and a plain-English FAQ and best-practice handbook written from the friction points that came up repeatedly during the week.

Two method choices did most of the work. First: nobody left their 1:1 without having built and shipped their own first skill. Once you have done it yourself, you know how to ask for the next one. Second: when someone brought a question, we didn't answer it — we put it to Claude together, framed it, and watched it come back. That teaches the mindset as well as the answer.

The results

Every person on site finished the week with a working skill or scheduled automation of their own.

In the post-week feedback, the team rated the engagement 4.75 out of 5, with an NPS of +50 and no detractors. Nothing was scored below four. Three questions came back perfect: that the sessions were easy to follow regardless of where you started, that people intended to keep using what was built in their 1:1, and that they felt confident continuing unaided.

But the numbers that matter more are the ones nobody was asked for.

Adoption spread past the group we expected. The team member the firm thought would resist is now leading it internally.

"It's not to reduce our team. It's to be able to maintain our team — not have to grow exponentially — keep a good team of people, and just get a lot more efficient and do a lot more."

— Andrew MacGillivray, Managing Director, Bayley Stuart

The pattern that works

Three things made this week land, and they generalise to any professional-services firm of this size.

Governance before enthusiasm. Because the policy work was already running, nobody had to slow down to ask whether they were allowed.

Every person builds one. A demo teaches you what is possible. Building your own skill, on your own work, with someone beside you, teaches you what to ask for next.

Then make it the firm's, not the individual's. This is the honest lesson from the week. When ten people each build their own skills, you get duplication and drift — two people independently asked for the same condition-report skill, one of whom had already half-built it. The consistency benefit only arrives when the best version is promoted into the shared library and everyone works from that. That is the step that turns a good week into institutional capability, and it is where we are heading next with Bayley Stuart.

The prize at the end of it is the one Andrew named on day one: a business where a new starter doesn't spend three to six months learning how things are done, because how things are done is already in the skill library — and they can do the job on day two.

If that's the journey you're on, talk to us at propelventures.ai/contact.

Propel Ventures is a partner in Anthropic's Claude Partner Network.