Yesterday I wrote about why the agency model is broken. The short version: agencies are structurally motivated in favour of the agency, incentivised to slow things down. Fractional AI ops is motivated to speed things up. The incentives point in opposite directions, and I know because I lived on both sides of it for years.
That was the theory. Today I want to tell you what the practice actually looks like, because I had two conversations yesterday that made the whole thing real.
The Head of Legal Who Thought She Had a Summer Project
I've been working out the details with a new enterprise client. Big project without any backend in sight, serious AI stack — we're talking 15 to 20 subscriptions across the organisation. The engagement is coming together, and I was on a call with their head of legal when the conversation took a turn towards the things I might be able to help her with.
The European Union has some significant AI compliance requirements coming for their sector. Internal training documentation, usage policies, guidelines for how the company deploys AI tools — the full regulatory package. She brought it up the way people bring up things they've been dreading. That careful, slightly exhausted tone. "Do you think you could help me with that as well?"
Absolutely. In fact, this is perfect for me, I told her. Not only can I take this on, but it would force me to map their entire AI stack in detail — which I need to do anyway to be effective in the main engagement. Two birds.
She seemed both surprised and relieved. She'd been framing this in her mind as a massive summer project. Something she'd have to carve out weeks for, probably alongside a consultant who'd need three meetings just to understand the landscape.
Here's what I'm thinking: two afternoons, maybe three. Wrap the problem around the right AI tools — Claude for the research and drafting, structured prompts for the policy frameworks, maybe film some Loom walkthroughs to give the training material a human feel instead of just handing over a stack of PDFs nobody reads. Between her domain expertise on the legal requirements and me accelerating the build through the tools, the spine of the whole thing comes together fast. Then you iterate, refine, and deliver something that actually works.
I could be wrong. Maybe it's bigger than I think. But that's the beauty of the fractional model — if it takes four afternoons instead of two, that's fine. I'm on a one-task-at-a-time-until-it's-done retainer. She's leading this as her project and I assist. The work gets done either way. There's no scope creep panic, no change order negotiation, no awkward email about additional fees. The retainer absorbs it.
Now think about this from her perspective. If she had to go find someone to help with EU AI compliance training for a mid-size company, where would she even start? Post on LinkedIn? Google "AI compliance consultant Europe"? She'd get agencies quoting five figures for a discovery phase, solo consultants who'd need weeks to ramp up on the sector, or firms that specialise in compliance but have never actually built anything with AI tools.
Instead, she's already talking to someone who's embedded in the company, already mapping the tech stack, and can just... add it to the board. That's the model working exactly as designed. I feel like it's a new age, and this kind of AI ops is going to be needed by a lot of different types of businesses.
The Nonprofit Founder Who Kept Asking "Can You Also..."
Later that same day, a sixth prospect conversation happened. Warm lead I'd been aware of — someone I followed up with, not a cold outreach. I'm still not doing cold prospecting right now because I want to make sure I nail fulfilment and get my own processes dialled in before I scale the pipeline.
She needs a website for a nonprofit. Since there's a charitable component, I'm offering the permanent 50% discount — that's the World Explorer Fund side of the business, where enterprise revenue funds lighter-touch work for youth nonprofits and charities.
In my agency days, we would have been reluctant to take this on without fully understanding the project first. You'd need a discovery call, a detailed scope document, a quote that accounted for every possible revision cycle. Because if you quoted wrong, you ate the margin difference. And with 30-plus developers on payroll, you couldn't afford to eat anything.
Now it's just me. I built the TestVentures website in two days. I built finnwardman.com in two days. Handling a nonprofit website in a month-long engagement — send me as many tasks as you want, no contract, pause any time — is easy to quote. It's a fixed price. I know what it costs me to deliver. Done.
"Sounds good," she said. Then: "Can you also help me write the business plan for the charity?"
Sure. I'm literally doing this right now myself — I'm applying to register as a charity in Bermuda and writing business plans is something I've been doing for 20 years across a lot of different businesses and offers. With the right AI tools, I can get a solid first draft done in a sitting and refine from there.
"OK great. What about building out the events and payments?"
Yes, I can handle that too. And actually, I'm building a charity payment gateway right now and I need nonprofit beta customers to test it. I'll include that for free.
"Wow, that's great. OK, let's start on Monday!"
Three asks. Three yeses. No scope document. No proposal. No "let me get back to you with a revised estimate." The no-contract, pause-anytime model means I don't have to agonise over whether each additional request fits the original quote. It all fits. That's the point.
The Unsexy Part
Meanwhile, I had no way to actually onboard her.
I'd been locked out of my Stripe account for a week. My main business email was tied up in an old account I couldn't access. My primary Stripe account was linked to an email address I'd deleted. The kind of infrastructure mess that accumulates when you've been running businesses for 17 years and not every migration gets cleaned up properly.
So while I was having these conversations about building websites and writing business plans and mapping AI compliance frameworks, I was also fighting with Stripe support trying to recover account access so I could actually take a payment on Monday.
This is the part nobody posts about on LinkedIn. The glamorous version is "I closed two prospects in one day." The real version is "I'm working on two prospects but I have no way to take payment, and I just spent three hours trying to prove to Stripe that I am, in fact, me."
Building the intake process, fixing the payment infrastructure, setting up the onboarding flow — this is the work that makes the client-facing work possible. It's not exciting. But it's the foundation, and if you skip it, the whole thing falls apart the moment someone actually says yes.
Something I've learned building software: don't build anything until you actually need it. Take real client feedback to build out the process. Well, I needed it now, and I'd be building it live with real client feedback on the onboarding process. It's the only time to attack this problem — when you really need it.
Six Tasks, One Friday
Here's what landed on my plate in a single day — a day that also included a 90-minute ski tour with two mates: EU AI compliance documentation. AI stack audit. Nonprofit website build. Business plan writing. Events and payment integration. Stripe account recovery.
Six completely different domains. Legal, technical, creative, financial, administrative. In the agency model, you'd need a compliance specialist, a technical architect, a web developer, a business writer, a payments engineer, and an ops person to handle the account issues. Six people, six billable rates, six minimum engagements.
In the fractional model, it's one person with the right tools and the right incentive. The tools handle the breadth. The retainer handles the pricing. The speed handles the trust.
Yesterday's post was the argument. Today was the proof.
The Numbers
Revenue: $0. Clients: 1. Prospects: 6.
The prospect count went up by one. The revenue is still zero because I can't take a payment until Stripe lets me back into my own account. There's a metaphor in there somewhere, but I'm too annoyed to find it.
Day 15 of 365.