Growing businesses
Know which AI bets pay off. Then make them part of how you work
Your team already uses AI. Some of it saves real time, some of it doesn't, and nobody has a plan for what comes next. I help firms of 10 to 200 people pick the right bets and build them in.
What limits AI adoption
The blockers are rarely the tools. They're about how the work is organized.
Everyone has their own version: Individual skill doesn't add up to company capability. Five people each getting 20% faster at their own version of a process gives you five processes, not a faster one.
Nothing is written down: The judgment calls that make the work good live in someone's head. You can't hand off a process nobody has described.
No one owns it: A few enthusiasts carry AI on the side of their real job. When they get busy, the experiments stop.
No time protected for it: AI work happens when things are slow, which in a growing business means never.
Nobody can say what's working: Individuals feel faster but nothing is measured, so there's no way to decide what to scale or what to stop paying for.
Where AI actually fits (and where it doesn't)
The best first use cases are repetitive, high-volume, and already understood by the team. The fit only becomes clear once the work is written down.
Write it down before you automate
In one agency workflow, nearly every error traced back to a judgment call someone made every week and had never written down. It wasn't the AI. Document the steps first, then automate what's left.
Repetitive production work
Turning one input into many outputs, like event details into client emails across four platforms. High volume, clear rules, and easy to check.
Information entered in three places
Enter it once and let the systems stay in sync instead of re-keying it by hand.
First drafts in your voice
Proposals, briefs, and client updates drafted from your own best examples, then edited by the person who owns them.
Follow-ups and reminders
The chasing that falls through the cracks when everyone is busy.
Insights from data and conversations
Client calls, email threads, and project data already show which services sell and where work gets stuck. AI pulls those patterns together so you decide on evidence instead of gut feel.
A strategy partner for the owner
Pressure-test a pricing change, a new service line, or a hiring plan before you commit. AI works as a thinking partner for the person making the call, not just a production tool.
The work clients pay you for
The judgment and relationships that set your firm apart. AI should buy your people more time for that work, not replace it.
Processes that change every time
If the team does it differently every time, standardize it first. Automating chaos only makes it faster.
How I measure it
Pick the numbers before you build. Then track them every week.
Time per unit of work: How long one client email, proposal, or report takes from start to finish.
Error rate: How often work comes back for fixes.
Draft survival: How much of each AI draft survives your edit. That number should climb every week. If it doesn't, the written process is the problem, not the model.
Capacity, not cuts: The goal is taking on more client work without hiring, measured in hours back and work your team no longer has to outsource.
How I work with growing businesses
The same five steps, sized for a small team. We start with the tools you already pay for, build on your real work, and hand it over so your team runs it. The one thing I ask for is two protected hours a week from the person who owns the workflow.
Case study
A marketing and communications agency, 18 clients, team of contractors
Everyone had used ChatGPT for something. One person had built her own reusable prompts. The core production workflow, turning event flyers into client emails across four platforms, still ran the way it had for years.
We rebuilt that workflow with the team, set ground rules, and handed it over. The result was the capacity of one additional FTE and one outsourced process eliminated. The team runs it themselves.