AI Workflow Consulting
AI workflows for the repetitive parts of client work.
We help owner-led service firms, such as consultancies, agencies and professional practices, find what's worth automating, build it around the tools they already use, and hand it over to the people who run it.
Use cases
Work that follows the same pattern every time
Much of client work is the same four steps: gather, draft, check and send. We build workflows for those steps.
An AI workflow is a series of steps in your work where an AI model does part of the job, such as reading a document, drafting a reply or pulling out the details that matter. The rest runs in the tools you already use, and a person checks the output wherever a mistake would be costly.
/generate-client-reports
- Pulling this month's results from your spreadsheets
- Removing client details the AI doesn't need
- Loading your writing voice for each client
- Drafting each report in your format
- Checking every figure against its source
Drafts ready for your review
Hi John,
March was your strongest month this quarter. Website inquiries rose to 48, and most came from the spring campaign.
| Measure | Feb | Mar | |
|---|---|---|---|
| Website inquiries | 41 | 48 | |
| New clients | 3 | 5 | |
| Average reply time | 6 hours | 4 hours |
For April, we suggest more budget for the pages that worked best.
How we work
Scoped, tested and handed over
Every engagement is quoted individually, after a free 20-minute call to see whether the work is a good fit.
Discovery
We look at how the work gets done today and decide what's worth changing first. We also agree what information the workflow needs and which outside software or AI providers may handle it. Discovery is included in every engagement. When the work is unusually complex, we may propose a separately priced discovery engagement first, so the quote for the full work rests on a close look at how it really gets done.
Scoped implementation
We build the workflow around the tools you already use, with a person reviewing the output wherever a mistake would be costly. The scope is agreed in writing before work starts.
Validation
We test each workflow on real examples, using only the information you've agreed it may see, and refine it until it meets the agreed requirements.
Training and handover
We document the workflow and train the people who will run it, then hand it over. Every engagement includes a check-in once the workflow has been in real use.
Optional ongoing support
If you want help as your work or tools change, we can keep adjusting the workflow. Beyond the included check-in, ongoing support is optional and quoted separately.
Who leads the work
Workflows built by someone who plans for things breaking
For 15 years, Taylor Turner kept large online systems running, in site reliability, technical leadership and cloud architecture roles, including Staff SRE and Principal Engineer. That work teaches one thing above all: anything automated will eventually get something wrong, so you plan for it.
So every workflow we build is tested on real examples, keeps a person reviewing the output wherever a mistake would be costly, and is documented for the people who run it.
Recent work
For a nonprofit, we built a workflow that turns each week's source material into post ideas, captions and video cut points. A person picks what gets used.
The same approach fits intake, reporting and document checks.
Articles
Recent articles
Taylor writes about what goes wrong in real systems and how to design around it.
Claude Isn't Ignoring Your CLAUDE.md. Your Old Rule Is Still in There.
When your agent follows a rule you already changed, the old version is often still sitting in your `CLAUDE.md` or the agent's memory. A smarter model mostly works around it, and you pay for that in time and tokens. Here's how old rules pile up, the one line that fixed it in my own testing, and where to look in your setup this week.
A Status Line for Claude Code That Shows What Each Prompt Costs You
Most status lines show where your session is. Mine shows what each prompt just did to it: context added, cost added, and whether the cache held. Here's what I watch and the repo if you want a starting point.
An AI Agent Exposed 31 Secrets. Here's What I Changed.
A debugging command in a coding agent session printed 31 credentials. What I rotated, what I rebuilt, and the least-privilege rules I now work by.
Have a repetitive task in mind?
Tell us about it. We'll reply by email within two business days to arrange a free 20-minute fit call.