Why publish your own tech stack?

Because the most useful thing a consultant can do is show their work. The AI-first stack we recommend to clients is the one we run ourselves. Publishing it honestly - the tools, the cost, the failure modes - is the operator-led version of credibility. It also helps prospective clients understand what we mean when we say we install AI properly.

What is Better Tuition Academy?

Better Tuition Academy is a 5-campus, 80-staff tutoring business on Sydney's Northern Beaches and North Shore. The founder of XLev is the CEO. It is the test bench for every workflow we ship for XLev clients.

All-in monthly cost for this AI-first stack runs under one entry-level FTE salary across all platforms (Claude.ai Teams, Claude Code, custom builds operating cost, n8n hosting, HubSpot, Xero, ServiceM8 and ancillary tools). The productivity uplift across 80 staff is several times that - the math is comfortably positive.

Can XLev install this stack for our business?

Yes, with the right modifications for your specific business. Most XLev engagements start with an AI strategy workshop to identify which parts of the stack matter most for your context, then a phased rollout. We do not ship a one-size-fits-all template; we adapt the underlying patterns to your industry and stage.

← All insights

Playbook·9 min read

The AI-first SMB tech stack: what we actually run inside an 80-staff Australian business

We are asked “what stack do you actually run?” more than any other question on discovery calls. This is the answer, in full. The AI-first stack that runs our own 5-campus, 80-staff Sydney education business, day-to-day. Nothing hidden, nothing aspirational - the real list.

Why we publish this

The most useful thing a consultant can do is show their work. The AI-first stack we recommend to clients is the one we run ourselves. Publishing it honestly - tools, costs, failure modes - is the operator-led version of credibility. It also helps prospective clients understand what we mean when we say “install AI properly.”

The full stack, by layer

AI surfaces

Claude.ai Teams: the company default for every operations, sales, marketing and finance staff member. Projects per function (Operations, Sales, Marketing, Finance, Customer Success, Tutoring Ops). Skills inside each Project for the recurring workflows.
Claude Code: every engineer. Custom shortcuts and commands for the most-used workflows.
Custom builds on the Anthropic API: roughly half a dozen production builds for the highest-leverage workflows (after-hours triage, structured extraction, content generation, internal knowledge platform).
Small ChatGPT Plus pool: 3-5 seats for the people who do meaningful image generation or voice-mode work.

Automation layer

n8n (self-hosted on Cloudflare): every automation lives here. Properly named, owned, version-controlled. ~40 production workflows wired across the stack.

Systems of record

HubSpot: CRM, marketing automation, customer comms. Connected to n8n and Claude for inbound triage and outbound communications.
Teachworks: practice management (industry-specific - the equivalent for non-tutoring SMBs would be ServiceM8, simPRO, Karbon or a CRM-led equivalent).
Xero: finance. Connected to n8n for invoice and expense automation, and to custom builds for the structured-extraction work.
Google Workspace: docs, sheets, gmail, calendar. Where most of the unstructured business data lives.

Operations tooling

Cloudflare: hosting for n8n and most of the custom builds. Cheap, fast, reliable.
Supabase: shared Postgres database for the custom-build layer.
Slack: internal comms. AI-augmented inbox triage routes the right messages to the right channels automatically.

What it costs

Total monthly AI tooling cost across all of the above runs well under one entry-level FTE salary. Breakdown roughly:

Claude.ai Teams + Claude Code: the largest line, but well under AUD $5,000/month at 80 staff
n8n self-hosted on Cloudflare: under AUD $200/month
Custom-build operating cost (API calls, hosting): AUD $500-1,200/month
HubSpot, Xero, Workspace, Teachworks: the existing systems, not AI-specific
Small ChatGPT Plus pool: AUD $200/month

Total AI tooling spend at 80 staff: well below AUD $10,000/month all-in. The productivity uplift across the team is several times that.

What we tried and dropped

Honest list of things that did not work for us:

One Custom GPT marketplace bought tool that promised to replace a function - turned out to be a thin wrapper, dropped after a month
An early attempt to use Microsoft Copilot - low adoption because BTA is Google-Workspace-led, not M365-led
A bespoke internal-tools build we should have done in n8n instead - too much custom code for the value
Two voice-agent experiments that did not pay back because our call volume was below the threshold; we run AI-summarised voicemail instead

How this translates to other SMBs

Most of the stack scales down cleanly to smaller businesses. A 10-staff Australian SMB can run essentially the same shape at a fraction of the volume: Claude.ai Teams for everyone, Xero for finance, HubSpot Starter for CRM, n8n for automation, and one or two custom Claude builds for the highest-leverage workflows. All- in monthly cost lands AUD $1,500-3,500 for a 10-staff business of similar shape.

For larger SMBs (50-200 staff), the same shape with bigger numbers. The pattern survives the scale-up because the underlying architecture - one Project per function, n8n as the connective layer, custom builds for the highest-leverage workflows - does.

How XLev installs this for clients

When we run a Claude Implementation engagement, the deliverable is a version of this stack adapted to your industry, stage and existing systems. We do not ship a one-size template; we adapt the underlying patterns to your context. Book a free 30-minute discovery call via the Contact page.

Frequently asked questions

Why publish your own tech stack?: Because the most useful thing a consultant can do is show their work. The AI-first stack we recommend to clients is the one we run ourselves. Publishing it honestly - the tools, the cost, the failure modes - is the operator-led version of credibility. It also helps prospective clients understand what we mean when we say we install AI properly.
What is Better Tuition Academy?: Better Tuition Academy is a 5-campus, 80-staff tutoring business on Sydney's Northern Beaches and North Shore. The founder of XLev is the CEO. It is the test bench for every workflow we ship for XLev clients.
Is this stack realistic for a smaller business?: Yes. Most of it scales down cleanly. A 10-staff Australian SMB can run essentially the same stack at a fraction of the volume: Claude.ai Teams for everyone, Xero for finance, HubSpot Starter for CRM, n8n for automation, and one or two custom Claude builds for the highest-leverage workflows. Total monthly cost lands AUD $1,500-3,500 for a 10-staff business of similar shape.
What does it cost?: All-in monthly cost for this AI-first stack runs under one entry-level FTE salary across all platforms (Claude.ai Teams, Claude Code, custom builds operating cost, n8n hosting, HubSpot, Xero, ServiceM8 and ancillary tools). The productivity uplift across 80 staff is several times that - the math is comfortably positive.
Can XLev install this stack for our business?: Yes, with the right modifications for your specific business. Most XLev engagements start with an AI strategy workshop to identify which parts of the stack matter most for your context, then a phased rollout. We do not ship a one-size-fits-all template; we adapt the underlying patterns to your industry and stage.

Where this fits

Consulting

$599/hr operator advisory or monthly retainer. For a single decision, an unblocking session or ongoing AI orbit.

See Consulting →Book a free 30-min discovery call