Skip to content
Book a free consult
Our work

Real businesses. Lighter work.

Different industries, one pattern: find where the work gets stuck, then redesign it with AI in the right places. These are outcomes from existing engagements. Client names are withheld where requested.

Book a free consultOn Course Golf

All client work, at a glance.

Eleven engagements in eight industries, two of them in flight. Figures are measured outcomes.

Agents

Automation

Training

National renovation contractor · pre-estimate review agent · Australia

Every estimate checked before the client sees it.

Stuckan afternoon of copy-paste

Flowing
1 step for 231 estimate lines to become a client-ready workbook6 locked client Excel bid forms, generated byte-perfect

Where it was stuck

Estimators read 30+ page PDF scopes line by line, built estimates against strict client margin rules and checked hundreds of rows by eye, then re-keyed everything into six locked client Excel formats. All inside a 5-day window from scope to quote. One missed line meant eating the cost on site.

What we built

We embedded with the team as a forward-deployed AI engineer and built a pre-estimate review agent. It catches zero-quantity lines, margin leakage and typo-class errors before anything reaches the client.

What changed

  • Every estimate is checked before it reaches the client.
  • The rules engine improves itself: every reviewer correction adds a new check.
  • It runs on a client-owned, open-source stack on Australian infrastructure.
  1. 01

    A scope becomes an estimate

    Estimators build it from a 30+ page PDF scope, against the client’s margin rules.

  2. 02

    The agent reviews it

    It flags zero-quantity lines, margin leakage and typo-class errors before anything reaches the client.

  3. 03Person in the loop

    An estimator decides

    Nothing reaches the client unreviewed. Every correction becomes a new check the agent writes into its own rules.

  4. 04

    One step to the client’s forms

    The six locked client Excel bid forms are generated from the estimate, byte-perfect.

How we build agents Private AI in Australia
Allied health practice, Victoria · inbound email agent

A lighter inbox, with a person on every reply.

Flowing
about 1,750 emails processed since launch930 drafts prepared, and none sent without a person

What we built

The agent reads incoming email, works out what each sender needs and leaves a threaded draft in the coordinator’s own writing voice.

What changed

  • It never sends: every draft waits for a person to check, edit and approve.
How we build agents
Local medical clinic · automation

Patient reports, drafted from voice notes.

Stuck3–4 h a weekwriting reports

Flowing
under 5 min per report

Where it was stuck

Report writing after consultations cut into family time and created bottlenecks for the practice.

What we built

A workflow that turns consultation voice notes into structured, personalised patient reports.

What changed

  • Reports stay personalised and compliant.
  • Doctors finish work earlier and feel less burnt out.
AI for healthcare
Commercial construction company · automation

Safe work method statements in hours, not days.

Stuckdays

Flowing
hours to produce a safe work method statement (SWMS)

Where it was stuck

Safe work method statements were written by hand for every project. Errors crept in, compliance checks were slow, and each one took days.

What we built

Each statement is now drafted from the project type, location and specific hazards, then routed for approval.

What changed

  • Documents are more consistent, with fewer errors.
  • Compliance with health and safety regulations improved.
  • Project managers spend less time on paperwork and more on site safety.
AI for construction and trades
Government tender consultancy · tender monitor · Queensland

Every open tender, in one morning email.

Stuck41 portalseach with its own login

Flowing
1 digest a dayabout 800 open tenders tracked, each scored against the firm’s bid profile

Where it was stuck

The firm bids on health and community services work put out by governments and Primary Health Networks. Those tenders sit on federal and state portals, aggregators and 31 PHN sites, each with its own login and its own alert emails. Opportunities were easy to miss, and knowing which ones were worth a bid took a person’s time every day.

What we built

A monitor that visited all 41 portals once a day, logging in where it had to, pulled every open tender, removed duplicates and scored each one against the firm’s own bid profile: client type, service type, budget, geography and the work it does not bid on. Each morning the team received one digest of the new tenders with the strong matches flagged, and a dashboard of everything open.

What changed

  • One morning email in place of 41 portals.
  • Strong matches were flagged for the managing director, who decided whether to bid.
  • The bid profile was the scoring rule, so the firm could see why each tender scored as it did.
  • It ran once a day on Australian infrastructure in Sydney.
  1. 01

    The monitor visits 41 portals

    Federal and state portals, two aggregators and the 31 Primary Health Network sites, once a day.

  2. 02

    Each tender is scored

    Against the firm’s bid profile. Clinical delivery, IT systems and construction score out; PHN evaluation and planning work scores in.

  3. 03Person in the loop

    The director decides

    The morning digest flags the strong matches. Whether to bid stays a person’s call.

How we automate
Residential home builder · client update agent · South East QueenslandIn build

A progress update for every client, in the builder's own words.

Where it was stuck

Clients building a home want to know what is happening on site. One team member wrote every update by hand: check the job in the construction management system, work out what had been completed, write a clear, friendly email. Across the whole book of jobs it was a constant task.

What we built

An agent that reads live job data through the vendor’s supported API, picks out the completed work clients care about and drafts a progress email in the builder’s own voice. Each draft lands in the team’s existing Outlook mailbox. Staff review, tweak if needed, and send, as they do now. No new software to learn.

Expected impact

  • Every client receives a regular, personal update without the team writing each one from scratch.
  • Staff time shifts from drafting to a quick review. A person stays in the loop on every email.
  • Consistent communication across all jobs, not just the ones that get chased.
  1. 01

    The agent reads live job data

    From the construction management system, through its supported API.

  2. 02

    It drafts the update

    In the builder’s tone, from the build activities that matter to clients, straight into Outlook.

  3. 03Person in the loop

    A person reviews and sends

    Nothing goes to a client unread.

How we build agents
Petroleum systems engineering firm · AI strategy and training · South East QueenslandIn flight

Finding the real bottleneck before building anything.

Where it was stuck

The firm designs, installs and maintains fuel systems for large energy and resources companies. Work was growing and the field teams were delivering, but a small office team was carrying the load. The instinct was to use AI to speed up quoting.

What we built

We stepped back first and looked for the constraint across the whole job lifecycle. It was the office, not the field: document control held up mobilisation and close-out, and with them invoicing. Faster quoting would only have grown the queue. From that came an AI strategy, a prioritised roadmap of agents starting with document control, and a two-tier rollout: hands-on Claude training and a custom plugin for every office user, then purpose-built agents on Australian-hosted infrastructure.

Expected impact

  • Every piece of AI that follows points at the constraint that frees the most capacity.
  • Office staff are being trained on Claude now, with the plugin shaped live in sessions so the team owns it.
  • The first custom agent, for document control, is in planning.
AI training for your team How we build agents

The pattern behind the results.

What every project has in common, whichever end it starts from.

  • Start with the workflow, not the tool

    We map the process first: where information comes from, where it goes and who needs to see it. Then we add AI.

  • Build on the systems you have

    We connect AI to the tools you already use, such as your email and CRM, so nothing feels foreign.

  • Measure what matters

    Time saved counts, and so does how the work feels. Is it lighter? Are errors down? Do people finish earlier?

  • Refine from real use

    The first version is never the last. We learn from your team and improve the system over time.

On Course Golf · AI training · Queensland and Victoria

AI that works behind the pro shop counter.

On Course Golf is a national network of independent golf retailers and pro shops. Its members wanted AI in their businesses but had no practical starting point. They knew the tools existed, not what to do with them on a Tuesday morning behind the counter.

It started with our talk, “Practical AI for your pro shop”, at the On Course Golf National Conference at Royal Pines. We then ran a four-week Intro to Claude Cowork cohort for pro shop operators and club staff across Queensland and Victoria: four live one-hour sessions, each recorded, with a companion handbook. Participants built working tools for their own shops as they went, including email skills, a golf coaching skill, a branded design system for flyers and promotions, and interactive dashboards.

9.3/10 average recommend score (NPS)7–10 h saved a week, across three responding shops

9.3 is the average answer, out of 10, to “How likely are you to recommend it?”

What else the survey showed

  • Every survey respondent is actively using Claude in their business.
  • Three different sessions were voted most useful, so the program worked for operators at different stages.
  • Participants are asking for more: ready-made skills and hands-on setup in their shops.

What participants said

  • A Queensland pro shop owner said the design system session made his flyers and promotions consistent for the first time.
  • A business development manager at a Sunshine Coast golf club rated the dashboards session most useful because it matched what the club was working on, and reported no barriers to using it.
  • A Victorian golf studio owner found the coaching skill most valuable, and has started using Claude for problem solving and connecting it to Xero.

“AJ from Flowtivity presented at our On Course Golf National Conference to our members who operate small businesses in coaching and retail. His presentation on using Ai in your business, was extremely well received and provided a great overview on the potential scope for Ai to support extremely busy operations, allow our members to save time, and importantly get to tasks that otherwise they may not, to enhance their business. AJ conducted a follow up of 4 weekly sessions with a pilot group to explore implementation and activate Ai across some common key areas of interest. These sessions were very practical and positive, and we will be able to use the knowledge, training and insights to expand the scope of coverage for our members. The expertise, knowledge and thorough methodology behind the training that AJ delivered, certainly made a difference for a very positive outcome.”

The pro shop program AI training for your team

Also in the work.

Training and an agent from existing engagements, told in words. Open one for what changed.

Credit agency · AI training

Weekly AI training for a team starting from zero.

Staff across management, marketing and operations had no exposure to AI and could not see where it fitted. A weekly, hands-on program covered prompting, building simple agents and choosing tools, so each team could find uses in its own work.

What changed
  • Staff moved from AI novices to confident practitioners in weeks.
  • Teams found and set up new AI workflows on their own.
  • Departments shared what worked, which brought teams closer together.
  • Management, marketing and operations each found their own uses for AI.
AI training for your team
Healthcare consulting firm · AI training and strategy

Training and a plan for a consulting practice.

The firm wanted to adopt AI deliberately but had no in-house expertise. We paired hands-on prompt training with an adoption plan built around its consulting work. The brief was to do it without disrupting existing workflows.

What changed
  • Staff built strong, practical prompting skills.
  • The team now uses current AI tools in its daily work.
  • Research and analysis for client deliverables improved.
  • The firm has its own AI capability, rather than relying on outside support.
Boxing gym · AI agent

Enquiries answered and trial classes booked, day or night.

People enquired, then dropped off before their first class, and engagement between classes was low. A chatbot now answers enquiries at any hour, handles common questions, books trial classes and sends encouragement between sessions.

What changed
  • More enquiries go on to become members.
  • Retention and class attendance improved.
  • Front desk staff have more time for the people in the room.
  • Members say they feel more connected to the community.

Where should we start?

From either end. Training works bottom-up; strategy works top-down. Both come back to one process, and the free consult is where we work out which suits you.

Your process matters more than your industry.

If you have repeatable processes and busy people, the approach holds. If you can describe the workflow, we can probably help you redesign it.

One good place to start

What could change in your business?

Every project here started the same way. Someone said, “We’re spending too much time on this.” Then we sat down and redesigned the process together.

Book a free consult

A free 1-hour conversation with AJ. No pressure, no pitch.