Before & After: What Client Workflows Look Like After Automation
One of the most common questions we get at Polaris Labs is: "Can you show me what this actually looks like in practice?" It's a fair question. The automation industry is full of abstract promises - "save hours every week," "streamline your operations," "unlock efficiency" - and short on specifics.
This page is the answer. Below are four detailed before-and-after case studies from real Polaris Labs engagements with Australian businesses. Business names and identifying details have been anonymised, but the problems, the builds, and the results are real. We haven't cherry-picked the best outcomes - these are typical results from businesses that had clear processes, engaged teams, and a willingness to change how they operated.
We've structured each case study the same way: who the business is, what the problem was, what the process looked like before automation (step by step), what we built, what it looks like now, and the specific measurable outcomes. No vague percentages. No estimated savings. Real numbers from systems that are currently running.
The Problem
This agency placed approximately 80 candidates per month across technology and professional services roles. They had 6 consultants, each managing a personal pipeline of active candidates. The issue: candidate follow-up was entirely manual and entirely inconsistent. Some consultants were diligent about touching base. Others let candidates go cold. With 50–80 active candidates each in their pipeline at any time, keeping up was genuinely impossible.
The consultants were spending an average of 3 hours per day on follow-up activity alone - checking who hadn't been contacted recently, composing individual emails, logging the contact in their ATS (Bullhorn), and noting next steps. That's 15 hours per week of consultant time on a task that is almost entirely templatable.
- Consultant manually checks Bullhorn each morning for candidates inactive 5+ days
- Individually writes and sends follow-up emails (20–35 min per batch)
- Manually logs email sent in Bullhorn contact record
- Sets reminder to follow up again in Bullhorn
- Repeats across all active candidates - inconsistent cadence depending on workload
- 40% of candidates received no follow-up in a given week due to consultant workload
- Automated sequence triggers when candidate status is set to "Active" in Bullhorn
- Day 1: personalised intro email from consultant's address (AI-generated from candidate profile)
- Day 4: role update email with relevant open positions matched to candidate skills
- Day 10: check-in email asking about job search status
- All contact logged automatically in Bullhorn
- Consultant receives Slack summary of any candidate who replies - ready to continue the conversation
What Was Built
The automation was built in n8n, connected to Bullhorn via its REST API. The personalisation step uses Claude to generate a tailored first-contact email using the candidate's profile data (job title, skills, location, current employer from their CV). Emails are sent via the consultants' connected Gmail accounts - so they appear to come from the consultant, not a no-reply address. All contact is logged back to Bullhorn via the same API connection. The Slack notification uses a webhook that fires on any email reply detection.
The Problem
This 12-person accounting firm had a long-standing debtor problem. Despite billing on standard 30-day terms, their average debtor days were consistently sitting at 41 days. 25% of all invoices were being paid late. The partners knew this was a cash flow problem but had not diagnosed the root cause: invoice follow-up was ad hoc and inconsistent, handled by a single practice manager who also had 40 other things on her plate.
When the practice manager got busy (which was most of the time), payment chasers didn't go out. When they did go out, they were written individually, had inconsistent tone and escalation, and were manually tracked in a spreadsheet that was always slightly out of date. There was no system - just effort applied unevenly.
- Practice manager checks Xero aged receivables report every few days
- Manually identifies invoices 7+ days overdue
- Writes and sends individual reminder emails
- Updates a spreadsheet to track who's been chased
- Escalates to partner verbally when invoice reaches 30+ days
- No consistent process - results entirely dependent on PM availability
- Automated daily check of Xero invoices via API
- 7 days overdue: polite reminder email with invoice PDF attached
- 14 days overdue: firmer reminder, cc'd to practice manager
- 30 days overdue: email from partner's address, phone call prompted via Slack task
- 45 days overdue: internal alert to partners + escalation note in Xero
- All activity logged; sequence pauses when payment is received
What Was Built
Built in Make, connected to Xero via the native Xero module. The scenario runs on a daily schedule, checking the Xero Invoices API for all AUTHORISED invoices where AmountDue is greater than zero and DueDate is in the past. A router applies different email templates based on days overdue. The 30-day touchpoint switches the from-address to the relevant partner (determined by tracking category on the invoice). Payment detection uses a separate Xero webhook that triggers whenever a payment is recorded, which cancels any pending follow-ups for that invoice.
The Problem
A residential building company in South-East Queensland was preparing around 35 quotes per month through Buildxact. Their conversion rate was 22% - reasonable for the industry, but the directors suspected they were leaving jobs on the table. The specific problem: 40% of all quotes that were sent received no follow-up whatsoever. Not because anyone decided not to follow up - because the process of tracking which quotes needed follow-up and when fell through the cracks of a busy estimating and sales function.
The estimator would finish a quote, send it, and move to the next one. Follow-up was someone else's job, but "someone else" was often also the person managing active jobs, handling supplier queries, and attending site inspections. Quotes went cold. Prospects who might have converted with one more touchpoint never heard from the company again.
- Estimator sends quote from Buildxact
- Quote status manually tracked in a shared spreadsheet
- Sales coordinator checks spreadsheet to identify quotes needing follow-up
- Individual follow-up calls or emails sent (when time permits)
- 40% of quotes received no follow-up
- No consistent timing, no template, no escalation
- Quote sent in Buildxact → automation triggers
- Day 3: automated "just checking in" SMS from estimator's number
- Day 7: email follow-up with a personalised note referencing the project
- Day 14: internal Slack task created for sales coordinator to call prospect
- Day 21: final email with adjusted scope offer if appropriate
- Sequence pauses if quote status changes to Won or Lost in Buildxact
What Was Built
Built in Make, with a Buildxact webhook firing when a quote status changes to "Sent." The SMS step uses Twilio connected to a virtual mobile number routed to the estimator (replies route to their phone directly). The email steps use personalised templates that pull in the client's name, project address, and quote reference from the Buildxact webhook payload. The Day 14 step creates a task in their project management tool (Asana) assigned to the sales coordinator, with the client's contact details and quote summary attached. Status change detection runs as a separate nightly scenario that checks all active quote statuses and cancels pending sequence steps for Won/Lost quotes.
The Problem
Every Monday morning, the operations manager at this Perth-based consultancy spent 2.5–3 hours preparing a weekly operations report for the leadership team. The report pulled data from four systems: project time tracking (Harvest), accounting (Xero), their CRM (HubSpot), and a capacity planning spreadsheet. The process was: export from each system, paste into a master Excel template, calculate key metrics, write a short commentary on performance, and email to four directors.
The report covered: billable utilisation by team member, total billed vs. target for the month-to-date, aged receivables summary, pipeline value and stage distribution, and capacity forecast for the next 4 weeks. All of this data was available in near-real-time in the source systems. None of it needed a human to compile - it needed a process to connect the sources and format the output.
- Monday 8am: ops manager logs into Harvest, exports timesheet data to CSV
- Opens Xero, downloads aged receivables report and P&L month-to-date
- Opens HubSpot, exports pipeline deals to CSV
- Copies all data into master Excel template, manually calculates ratios
- Writes 3–4 paragraph commentary on notable variances
- Emails formatted report to four directors (usually around 11am)
- Sunday 11pm: automated scenario runs
- Harvest API: pulls previous week's time entries, calculates utilisation by person
- Xero API: pulls MTD revenue and aged receivables
- HubSpot API: pulls pipeline deals, values by stage
- AI commentary step: generates variance analysis in plain English
- Formatted HTML email delivered to directors at 7am Monday
What Was Built
Built in n8n, scheduled to run every Sunday at 11pm AWST. The workflow makes API calls to Harvest, Xero, and HubSpot in parallel. A code node calculates the derived metrics (billable utilisation as a percentage, MTD revenue vs. target, average debtor days). The data is passed to an AI step (Claude via API) with a structured prompt that generates a 3–4 paragraph commentary, specifically flagging anything more than 10% off target in either direction. An HTML template node formats the data and commentary into a clean email layout. The final node sends the email via SMTP to the leadership team distribution list.
The AI commentary step was the most important design decision. Early versions of the report delivered raw numbers with no interpretation, which directors still needed to spend time processing. Adding the AI commentary step - which flags material variances and provides simple explanatory context - transformed the report from a data dump into an actual decision-support tool.
What These Case Studies Have in Common
Looking across these four engagements, a few patterns emerge that are worth understanding before you scope your own automation project.
The highest-ROI automations eliminate coordination overhead, not core work. In every case study above, the automation eliminated the administrative burden of coordinating, tracking, and reporting - not the work itself. Consultants still have conversations with candidates. Directors still make decisions based on the weekly report. The automation removed the friction that surrounded the work, not the work itself.
Consistency is often more valuable than speed. The recruitment agency's biggest win wasn't that follow-up emails were sent faster - it was that they were sent consistently to 100% of candidates, every time. The construction company's biggest win wasn't faster quoting - it was that every quote received a follow-up, not just the ones that happened to catch the coordinator's attention. Automation makes good processes reliable, not just efficient.
The AI step adds the most value at interpretation, not execution. In three of the four case studies, an AI step was included - and in all three cases, its role was to interpret or contextualise structured data (write a personalised email, generate a commentary on financial data), not to execute the core workflow. The workflow logic was deterministic. The AI made the output more useful to humans.
Every business in these case studies came to us saying some version of "we know we should automate this, but we're not sure where to start." The answer, in every case, was the same: start with the process that runs the most frequently and requires the least judgment. Automate the calendar, not the strategy session.
Is Your Business a Good Candidate for This Kind of Automation?
The businesses in these case studies share a few characteristics that made automation high-impact. They all had: clear, recurring processes that happened on a predictable schedule or trigger; existing software tools with accessible APIs or pre-built integrations; and a team willing to trust the automation and adapt their workflow around it.
If your business has recurring processes, uses mainstream software tools, and you can identify at least one task your team does every day or every week that follows a consistent pattern - you're a good candidate. Book a free discovery call with the Polaris Labs team and we'll tell you honestly what's achievable and what it would cost to achieve it.