Skip to content

For firm owners

Most firms are automating the wrong things.

I've spent twenty-five years doing this to my own practice, in front of real clients, including the parts that didn't work.

01Twenty-five years of this

What actually got built, and what it cost to learn.

Not a product roadmap. A practice that kept being rebuilt underneath its own clients.

  1. 1999

    Walked into a filing cabinet

    The problem
    A Toronto practice run on paper, desktop installs and a cabinet everyone was slightly afraid of. Year-end was archaeology.
    What got built
    Nothing yet. Just the conviction that this was all obviously fixable.
    What happened
    Told repeatedly that the profession didn't work that way. Correct, for about a decade.
  2. 2010s

    Going paperless, properly

    The problem
    Documents arriving by every channel at once, then living in six places. Staff time going to filing rather than advice.
    What got built
    Document capture at source, a single client record, and workflows that assumed digital rather than bolting it on.
    What happened
    Saved money and created value — enough that the trade press wrote it up.
  3. 2010s–2020s

    Cloud-native, not cloud-hosted

    The problem
    Moving desktop habits onto a server changes your bill, not your practice.
    What got built
    Rebuilt around cloud ledgers and bank feeds, with reconciliation running continuously instead of once a quarter.
    What happened
    Ranked among the top 50 cloud accountants in North America; CPA4IT named a top-100 global cloud brand.
  4. 2023

    FridAI — testing the hype in public

    The problem
    Every vendor suddenly had AI. Almost none of it survived contact with a real engagement.
    What got built
    Hands-on labs and live sessions for accountants and bookkeepers, demystifying tools rather than selling them.
    What happened
    A back catalogue of what worked, and a longer one of what didn't.
  5. Now

    NorthStarrs

    The problem
    Firm owners collect ideas at conferences and implement none of them.
    What got built
    A retreat concept: tech stacks and team structures shared openly, EOS principles, automations and AI built together in the room — leaving with an accountability agent rather than a notebook.
    What happened
    In progress, in public.
02Running in my firm right now

Not case studies. Systems my team built, that I have to live with.

Every one of these is running inside CPA4IT today, on real client work, with my name on the returns. If one of them breaks in April, it breaks on me.

Some have run for years, one shipped last month. Between them they give back seven to ten days a month — but the part that matters more is what they catch: HST errors before the invoice reaches the client, which tax forms a client actually needs, and a whole class of transcription error that disappears when nothing is typed twice.

These are a handful of many. They’re the ones that are easiest to explain without sitting down with you.

Invoice processing and revenue analysis

Hundreds of invoices a monthRunning in production for close to two years

One person opening every invoice, keying it into Excel, then cross-referencing each line against Keap and the time-and-billing system by hand.

  1. 01Open every invoice and key it into Excel by hand
  2. 02Look every line up in Keap, then again in time and billing
  3. 03Build the whole revenue analysis by hand, then send each invoice

3 manual steps, every time

03Four things I’d argue for
  1. 01

    Automate the workflow, not the task

    Most firms buy a tool, bolt it onto an unchanged process, and end up with the same work plus a subscription. Fix the workflow, then automate what's left.

  2. 02

    A firm is habits before it's tools

    No system survives contact with a team that didn't ask for it. Adoption is the whole game, and it's a people problem wearing a software costume.

  3. 03

    Pick a niche narrow enough to be obvious

    Being the firm for a specific kind of client beats being a firm for everyone. It makes marketing, pricing and process design easier at once.

  4. 04

    Test the hype yourself

    Most AI claims in this profession don't survive a real engagement. I run the experiments in my own firm and report back — including the ones that fail.

In public

The unedited version, in public.

I work through this in the open — what I’m testing, what’s working, and what turned out to be nonsense.