Powered by Finn
Start here · Deep dives
Pillar · 2,400 words
Five signals it's time. The CTO vs CAIO vs AI advisor split. What a real Tuesday looks like on retainer.
Why this matters →Pillar · 2,700 words
Two skilled operators for $5,000 a month instead of one $80–100k hire. No contract, cancel any time. What you get, what it costs, and who it is not for.
Why this matters →Pillar · 3,000 words
AI-first vs AI-native. How to become one, one workflow at a time. What the operating day looks like from the inside.
Why this matters →Pillar · 4,000 words
What it does, what it costs vs hiring, and the seven guardrails before any script writes to a production ledger.
Why this matters →June 27, 2026 · Day 093
For years I bought hours from agencies and freelancers and wondered why I was tired. The thing I was actually short of has a name: high agency. Here is the five-level chart I now run every contractor against, and why the person they put on your account is the only thing you are really buying.
June 25, 2026 · Day 092
The email I sent my co-founder last month was five words long: looks like we go for US. Attached was the third and final no from the country where my late son grew up. Two years, three denials, one charity that still could not ask anyone for money. Here is the business half of how we got 501(c)(3) status instead, what the US route actually costs a non-US resident, and why I stopped appealing and switched countries.
June 26, 2026 · Day 091
A thirty-page report came across my LinkedIn feed this week, all about the EU jobs market. I almost kept scrolling. Then one finding stopped me: the jobs that pair AI with human skills are holding up better than everything else. Not the roles AI barely touches. The ones where a person uses the tool and still makes the calls. That is the exact bet I have been making for twenty years, and the specific bet of the last few months. Here is what the data says about AI-augmented jobs, and why it matters whether you build, hire, or are quietly worried about your own.
June 24, 2026 · Day 090
Someone I build for asked me why he would pay twelve thousand dollars a year for the Bloomberg API when his phone already pulls stock prices free from Yahoo. Fair question. Half the time the answer is don't pay. The other half is where most people building a data tool get it wrong: the Bloomberg Data License API is not a news-and-prices feed, and the three things it gives that Yahoo cannot are the whole reason it exists. Here is the full comparison, including the part nobody tells you about the licence.
June 23, 2026 · Day 089
Back in the winter I re-explained the same context to my AI every single session, because it woke up each time with no memory. The fix came from two ideas the people who build these tools live by: the bitter lesson, and give the thing a memory. Here is the ten-minute, no-developer setup I use now, one brain file per project that the agent reads first and writes back to last, plus the one part we are still bad at.
June 22, 2026 · Day 088
The AI employee I told you went live has been running on a server I couldn't log into from my own laptop. Here is what the thing actually is, an open source AI assistant I forked and pointed at my prospecting, how the overnight-draft-then-morning-approval loop works, and the unglamorous truth that the agent was the easy twenty percent and the server plumbing was the other eighty.
June 21, 2026 · Day 087
I went looking for grants to fund the youth charity we run and found one I wasn't looking for: Google gives any registered nonprofit ten thousand dollars a month in free search ads. Here's what Google Ad Grants is, who qualifies, why we applied, and how I built the whole landing page and campaign in an afternoon while the application sits in review.
June 20, 2026 · Day 086
I asked five answer engines the questions a buyer asks before hiring someone like me. My company came back in none of them. So I am running a live test in answer engine optimisation: publish the answer today, check in seven days whether any AI cites it.
June 19, 2026 · Day 085
I spent two years trying to register a charity in Bermuda. The reply that finally moved things along told me, two years late, that an application had already been denied. The US 501(c)(3) said yes in months. Here is what the waiting actually cost.
June 18, 2026 · Day 084
There is no 'close the books' button you can buy. Here is how to automate month-end close with AI: five repeatable steps, built custom to your own bank feeds and accounting platform, with a boundary that keeps the figures out of the model.
June 17, 2026 · Day 083
At seven in the morning my CTO and I set a number over WhatsApp: one new client in 45 days, both of us on $15,000 a month by year end. Here is the offer behind it, and why we are done sending cold mail to people who never asked.
June 16, 2026 · Day 082
A finance leader I work with had never opened an AI tool. We turned their forty-page monthly board pack into a single page in under an hour. Here is the exact three-step method, the prompt to copy, and the one safety step most people skip, written for finance leaders new to AI.
June 15, 2026 · Day 081
A card order for the little nonprofit I run failed with a number instead of a sentence. So I rebuilt the whole money stack from scratch: how a small charity actually takes donations across borders, in any currency, and what tripped me assembling it.
June 14, 2026 · Day 080
A COO messaged me at night, worried his firm was losing control of its data. Here is how we run AI inside SOC 2: deterministic code for the data, the model for the judgment, and a boundary the figures never cross.
June 13, 2026 · Day 079
SEO is one of three scoreboards now. I pointed an ai agent at my own Search Console and analytics, audited for AEO and GEO too, and found the blind spots.
June 12, 2026 · Day 078
This morning I got on a call with the CFO half expecting to hear I had wrecked a month of her books. She came on smiling. The ai financial controller I have spent forty-five days building had closed its first full month, and the numbers balanced to the cent.
June 11, 2026 · Day 077
The terminal agent read an entire overnight payment run off a client's live books an hour before I knew about it. It had not hidden anything. It just had nowhere to put what it learned. That gap has a boring name, AI agent memory, and fixing it took two scheduled habits instead of a new tool.
June 10, 2026 · Day 076
The best week to reach someone about fractional AI ops is the week they decide to hire an ops person. Tomorrow my AI sales agent starts watching a database of millions for exactly that signal, writing a tailored draft the morning it shows up. The only thing it cannot do is send.
June 09, 2026 · Day 075
The automation finished, went green, and reported success. Then I counted by hand: twenty-three of fifty bank lines for the month had never made it into the books. Here is what the setup got right, what I assumed wrong, and how one message turned a monthly clean-up into a full-year plan.
June 08, 2026 · Day 074
An AI that admits it is guessing has no business posting to a client's books. Here is how a workflow earns its way up the trust ladder: exception, candidate, trusted, about five clean runs per pattern.
June 07, 2026 · Day 073
A founder told me K-1s were the most painful manual job his team had. He had not even heard me pitch AI yet and he was already pointing at the place it should go. That is the next build.
June 06, 2026 · Day 072
Jan and I are building the prospecting engine on sweat equity. Anything that becomes a company asset, he does not bill for. We are having the hard conversation before there is money to fight over.
June 05, 2026 · Day 071
I queued five posts and went offline for four days right after letting AI into a client's live books. If a system only works when you are watching it, you did not build a system. You built a job.
June 05, 2026 · Day 070
The agent is built, the worker is healthy, the ad is paused at twenty dollars a day waiting to switch on. The only thing in the way is Meta business verification, and this morning it asked me for a document I do not have. So I am emailing a trust company and an internet provider, asking them to put my phone number on a bill.
June 04, 2026 · Day 069
A financial-services research desk was paying $36,000 a year for an AI that reads the news every morning. Most of that was the expensive model doing cheap work. Here is the what-to-do version of cutting it roughly in half while staying SOC 2 compliant.
June 03, 2026 · Day 068
An AI session froze mid-cleanup. It wasn't stuck on the problem, it was stuck behind a locked door. Underneath was the real issue: the folder that holds my whole business memory had quietly turned to cruft.
June 02, 2026 · Day 067
I sat down to write a pillar page on what an AI-first company is. Then I noticed my own company had already done three things that morning without me.
June 01, 2026 · Day 066
Before AI posts a single real journal entry into an enterprise client's books, you do weeks of quiet work nobody sees. Here is the planning behind an AI controller going live: dry runs, the safeguards, AP automation, K-1 reads, and a dashboard that counts the hours saved.
May 31, 2026 · Day 065
At half past four on Friday I joined a call I had asked for. It had started at four and was already over. I had accepted the invite into a work Outlook calendar I never open, and I had the time wrong in my head. Here is how to sync a locked-down work calendar into the one you actually live in, and the system I am putting in place so it never happens again.
May 30, 2026 · Day 064
Wednesday I set a hard date to start prospecting before the weekend. It is Friday and no leads have come in. So I am calling out the plan in public: an openclaw fork on Apollo for outbound, and Meta ads into a WhatsApp ai agent built to qualify leads before I ever pick up the phone.
May 29, 2026 · Day 063
For thirty days the first thing I do is fire claude and plan the day. One of those daily scripts checks crawler traffic against a B2A platform I am building for the charity space. The agents arrived. Thousands of them. They were not doing the one thing I needed them to do. Yesterday I killed the theory in public.
May 28, 2026 · Day 062
Kirsten and I both work 12-13 hour days unless we force ourselves to stop. Yesterday we drove to Bidart for an overhead surf session, moules-frites at a Basque shack, and a calendar block for the next one: every two weeks, no exceptions.
May 27, 2026 · Day 061
Day 61 forces the prospecting work I have been writing about for sixty days. The Mac mini arrived. I borrowed a mouse from a neighbour. Then I started two parallel bets, openclaw forked into Apollo for signal-based cold outreach, and Meta ads to a WhatsApp ai agent I am building this week.
May 26, 2026 · Day 060
On to a new AI project. An AI research assistant for financial analysts. We have some built in Perplexity. I am building on top of it. It sits next to Bloomberg and AlphaSense in the market. Day zero of the 90-day commit to a paying customer.
May 25, 2026 · Day 059
Sunday morning on the deck with coffee, Kirsten, and the birds in the red-berry trees. The BIP energy had been fading. AI slop everywhere, sixty days of writing, not a single direct cold lead. Then a stranger's note landed in LinkedIn overnight and reset the week.
May 24, 2026 · Day 058
Thursday morning Amazon France auto-cancelled three Mac mini orders across two days. Two hours gone, no human to call. I texted my enterprise client's COO and asked him to order it. The cost of wearing every hat as a solo AI ops founder is not the embarrassment, it is the displaced work.
May 23, 2026 · Day 057
How I measure AI for an enterprise client: hours saved on boring repetitive tasks. Two builds before lunch, one custom script and one SaaS swap, both measured the same way.
May 22, 2026 · Day 056
Three big asks land in Slack before 10am. Each one multi-week. Each one from someone I want to say yes to. The one-task-at-a-time rule I have been writing about for fifty-five days does not answer what I do at 10am while I wait for the operator's reply. The honest version is one commitment a week, a queue for the gaps, the queued-yes when the asks stack, and a hire when the queue overflows.
May 21, 2026 · Day 055
MCT oil, creatine, L-theanine in the Amazon basket this week. Trying to get back the head I had on keto two years ago. Same week, a bigger rearrangement: my CTO goes full-time on fulfilment and I go full-time on writing, sales, and marketing. Both moves point at the same year-end number.
May 20, 2026 · Day 054
Yesterday the code review document landed in Slack. Eighteen findings on a live AI build that posts transactions into a real general ledger for an enterprise client. Six high-severity, seven medium, five low. Three of the highs got shipped, tested, and committed the same day. The work behind the slogan.
May 19, 2026 · Day 053
A WhatsApp at 6:54am hyped the AI services gold rush. The reason my mate thought I was not already in it was sitting on my homepage in 60-point type. Day 52, the monthly iteration begins.
May 18, 2026 · Day 052
Anthropic launched its Small Business AI bundle on May 13. The same week, I finished hiring the AI salesperson who is going to sell it. Five weeks of zero outbound, the build underneath, and the pitch the bot starts dialling on May 20.
May 17, 2026 · Day 051
Three years after Finn died, we filed for US 501(c)(3) status for the Finn Wardman World Explorer Fund. The legal structure, the quarterly rhythm, the first grant decisions, and the three real reasons for the US charity layer.
May 16, 2026 · Day 050
A senior partner at one of our enterprise clients asked me last month whether they should hire a financial controller. I didn't have a clean answer in 20 minutes. Today I sat down and wrote the clean one. 4,000 words on what an AI financial controller actually does, what it doesn't, what it costs, and the seven guardrails before any script writes a line to a production ledger.
May 15, 2026 · Day 049
Three senior people on the same client email this morning, each one replying within an hour. Each one said a version of: this is harder than we thought, and the average knowledge worker has no chance of doing what this engineer is doing. The engineer in question is vibe coding the company's first internal AI tool on his laptop, no GitHub, no tests, no proper hosting. We're going to redo most of it. He knows. I know. And I'm letting him keep building. The replies are the cleanest answer I've seen to the question most C-suites are actually asking: how to hire a fractional AI lead.
May 14, 2026 · Day 048
I ask Jan this week if he wants to be CTO of TestVentures. He writes back, "yeah, I'm fine with that." No offer letter, no recruiter, no back-and-forth on title. He's my first human hire after 47 days of building in public with three AI models in three tabs. I can't read his code well enough to judge it, but I can read 12 months of behaviour, and I'd rate it 9 out of 10 on the things I actually care about. Top companies are hiring failed entrepreneurs on purpose right now. The trait they're chasing is the one I'm hiring on: high agency.
May 13, 2026 · Day 047
07:07 CEST and the kettle is on. I open one tab and it is not Gmail. It is a markdown file my daily digest wrote overnight, and yesterday it told me 1,792 named crawlers read giveready.org in 24 hours and zero of them submitted a nonprofit. A daily self-learning loop on a marketing funnel, three numbers, one diagnosis, one shipped fix, tomorrow's number as the test, runs on any funnel with a top, a middle, and a bottom. I run it on GiveReady's agent-discovery funnel and on the TestVentures search funnel.
May 12, 2026 · Day 046
A friend WhatsApped me Monday morning. He's been running AI tools every day for three weeks. Verdict, his words, not mine: "I'm just clicking buttons faster. Am I actually automating anything?" I went looking through six weeks of my own work to answer him honestly. Two things on my list. The discipline that gets the second category there is Christopher Penn's 5Ps.
May 11, 2026 · Day 045
56 regression tests passing in 1.5 seconds. Three months of bank data, byte-identical to a hand-built CSV. And before any of it ships live, a CTO-caliber engineer still reads every line.
May 10, 2026 · Day 044
Day 44 of 365. Kirsten does a So Cal surfer impression every time I say "vibe living". Yeah brah. Fair enough. This week the vibe living meter caught a legacy relationship I had to cut, regardless of the financial exposure. The Royal Enfield earned its first ride of the season.
May 09, 2026 · Day 043
Day 43 of 365. Yesterday I wrote about the corrections log. Today the skill that feeds it. On April 22 I ran a brutal Sedaris-bar critique against four of my own posts. The lowest score was 2.8 out of 10. The skill that came out of that morning is still the single best piece of voice infrastructure on this site.
May 08, 2026 · Day 042
Day 42 of 365. Yesterday I caught myself scrubbing the same three words from a Claude draft I had scrubbed the day before. Shape. Frame. Lane. The corrections log is the durable artefact of publishing daily with an AI as co-author. Without it, every post drifts toward AI-fine. The work is in the gap between AI-fine and yours.
May 07, 2026 · Day 041
Day 41 of 365. Yesterday Claude agreed with every contradictory idea I floated, while Scott Galloway made the case AI is a hype bubble and a friend told me builders are quietly switching models. Eight rules I am running on now, plus the corroboration that says my friend was not imagining it.
May 06, 2026 · Day 040
A cron fires at six-oh-six every morning, pulls twenty-four hours of agent traffic off the live worker, and prints a morning page with yesterday's top hypothesis at the top. This is what B2A marketing looks like when you run it as a self-learning loop in public, plus the seven-step you can copy.
May 05, 2026 · Day 039
Day 39 of 365. A 22-year-old reader messaged me about his head getting in his way. The reply walked through the mixing-board meters, the yellow aspen leaf, and the place I now call ground zero.
May 04, 2026 · Day 038
Day 38 of 365. The vibe coding craze is real and the leverage is real. The trap inside it is also real. The story of the KPI dashboard I lost three months on, the friend whose weekend project gives me cringes, and the 5 out of 10 skill of knowing when to call in the experts.
May 03, 2026 · Day 037
The thin harness, fat skills framework moved from theory to live in the field. An enterprise client sent a full architecture at 5:21am — the fat skills are theirs, the harness is mine, and the operating contract for fractional AI ops became unmistakable.
May 02, 2026 · Day 036
I asked an enterprise client over Trello if the task was done. The reply came back yes, but there is another part I am doing manually in this other application. I had no idea what other application. Here is what that taught me about ai workflows for operations and where the strict async rule has to bend.
May 01, 2026 · Day 035
First substantive task from an enterprise client. Demo tomorrow morning. Tonight's job is seeding test data by hand into a Claude artifact so a research team can see the shape of the workflow before anyone has to commit to a build.
April 30, 2026 · Day 034
A friend mentioned x402 payments on a Saturday morning. A week later I had a small nonprofit donation platform built and a $150K grant application in flight. This post is me coming out about GiveReady, why the demand side of small-charity funding is broken, and why I am building it whether the grant lands or not.
April 29, 2026 · Day 033
A legacy client wants me to break our contract and run on trust. The other party is savvy, the contractors caught in the middle have my respect, and I have most of a week to navigate it without getting pulled back into the practice I left. The first move was sitting down with the coach.
April 28, 2026 · Day 032
Speccing a personal AI assistant on the plane back from Madrid. No wifi, twenty years of street smarts, and a Mac mini in the cart before a single line of code.
April 27, 2026 · Day 031
An old sailing mate wrote me yesterday after my LinkedIn note. Second person this week asking how to make the jump. There is a framework gap between the philosophy shelf and the how-to-start-a-business shelf, and I am writing my way toward it in public.
April 26, 2026 · Day 030
A Sunday BIP, pre-written from a flight on Thursday because the family is driving from Switzerland to the southwest of France this weekend. The lifestyle business is a one-way door. Once you have tasted it, you do not go back.
April 25, 2026 · Day 029
A weekend in Madrid with my son Somers. Bad golf, talk of commodities trading, and the question I keep landing on as a father. The same lens runs on TestVentures. Small team of specialists, 50 to 100 clients, stay in front of the wave.
April 24, 2026 · Day 028
Ran a two-week Facebook ad in Kenya for our $10,000 travel grant for young adults. Forty qualified people wrote in asking how to apply. Zero of them applied. The real demand-side problem is not attention, it is belief.
April 23, 2026 · Day 027
Two weeks ago I started investigating payment rails for the family adventure fund. Yesterday I submitted a charity grant application for a B2A donations platform I built in the gap. Here is how I am writing about it without burning it.
April 22, 2026 · Day 026
An old colleague I hadn't really crossed paths with since 2015 reached out yesterday offering help from his team for free. Buried in the same note was the thing that actually stopped me — my first fan ever. And he told me the people I'd been writing to already exist as a subculture. They have a name. The FIRE movement. I'd never heard of it.
April 21, 2026 · Day 025
Shipped the first pillar page of the 365-day BIP — a 2,400-word piece on what a fractional Chief AI Officer actually does, with the chiefaiofficer.com story told in full for the first time.
April 20, 2026 · Day 024
Three weeks ago I announced a 365-day build-in-public campaign from professional rock bottom. This morning, first revenue landed — $9,200 monthly, likely recurring. What changed, and the one rule I refuse to break.
April 19, 2026 · Day 023
Finishing PDG on hormone therapy with castrate-level testosterone, half-strength legs, and a broken bamboo pole — the Arolla decision and why I continued.
April 18, 2026 · Day 022
Saturday morning. By the time this publishes I'm on a glacier, racing the Patrouille des Glaciers. This post shipped itself. The business model is simple — finish the task at hand, or nothing gets done. Pre-pay to start. One task in the queue. Work it until it's done.
April 17, 2026 · Day 021
First real task kicked off with the enterprise client today. I can't say much about who or what. But the shape of the problem is worth writing about — their CFO spends twenty hours a month reconciling across accounting systems and jurisdictions. Build lite, codify as a skill, close the loop, verify the thing stays in sync.
April 16, 2026 · Day 020
Yesterday I wrote about fat skills and thin harness as theory. Today I had to use it on two real fires, and the rule held. Solve it once, then make sure future-you never has to solve it again.
April 15, 2026 · Day 019
Here I am 19 days into starting a new service and I hit admin hell. This is the stuff I really don't like about being an early stage founder — but it's the stuff that needs to be dealt with in order to move the company forward.
April 14, 2026 · Day 018
Garry Tan named it thin harness, fat skills. How a non-developer used the idea to ship a technical grant in ten working days — what each layer means in plain terms.
April 13, 2026 · Day 017
Today I get my first PSA blood test after 184 days of aggressive hormone therapy for prostate cancer. Between the ADT, a keto diet, and caffeine re-entry, I've made some of the most irrational decisions of my life. All of them feel right.
April 12, 2026 · Day 016
I've been hearing about build in public for years and never actually followed one. Nobody is following this one either. But tomorrow I need to onboard a client, so today I'm borrowing the playbook from a guy who runs 35 design clients solo at $5K a month. Here's how I'm adapting Designjoy's model for fractional AI ops.
April 11, 2026 · Day 015
Yesterday I explained why the fractional model flips the agency incentive. Today I had two conversations that proved it. A head of legal floated a massive EU compliance project like it was a burden. A nonprofit founder needed a website, a business plan, and a payment system. Both conversations ended the same way — yes, I can do that too.
April 10, 2026 · Day 014
I ran a software agency with 75 people across 12 time zones. It was a beautiful lifestyle business, until I saw why the incentives were backwards. Agencies are built for the supplier. Fractional AI ops is built for the client. Here's the difference.
April 09, 2026 · Day 013
I built a working prototype in 48 hours. Then I tried to onboard an actual charity and hit a wall that had nothing to do with code. Charity governance is a bigger problem than I expected — and it might be the most important thing I work on this year.
April 08, 2026 · Day 012
Today, 3 years ago, our lives changed permanently. This post is about a poem my son wrote when he was 13, and a thin voice that's starting to speak to me.
April 07, 2026 · Day 011
We built the AI coach. We designed the check-in. We set up the daily review. Now it's time to put it to work on real decisions. I ran three live business opportunities through my coach last week. One seemed random. One was an entrepreneur's dream. One looked safe. Here's what happened.
April 06, 2026 · Day 010
You designed the day. You check in every morning. But are you actually doing what you said you'd do? The daily review is the part most people skip. It's also the part that tells you whether you're flying to LA or drifting to Winnipeg.
April 05, 2026 · Day 009
You designed your perfect day. Now what? Without a daily check-in, you're a pilot flying from London to LA who's off by half a degree. You won't end up in LA. You'll end up in Anchorage.
April 04, 2026 · Day 008
Three weeks ago I quit a high-paying job because of a vibe from my phone that was making me sick. Then I quit a co-founder role. Then I designed my perfect day 12 months out. Here's where building an AI co-pilot actually starts — and it's not where you think.
April 03, 2026 · Day 007
A prospect bailed on Google Meet mid-call and pulled me into a three-way WhatsApp. My transcription was gone. Except it wasn't. Here's why Granola is the one SaaS I'm not cutting.
April 02, 2026 · Day 006
I set up an AI voice agent called Amy for a prospect's after-hours calls. She scores about 8 out of 10 against a human. Factor in that she never misses a call, and it's closer to 9.
April 01, 2026 · Day 005
I cycled through three jobs in 90 days. Then something clicked. Here's how a 10-day silent retreat in India became the thing I'm building my entire business around.
March 31, 2026 · Day 004
After building a vault for the business, I built one for the mission: the Finn Wardman World Explorer Fund, four years of writing, and a manuscript. Here's how it works.
March 30, 2026 · Day 003
Every Cowork session starts blank. Corrections, decisions, project context, gone. I lost the same fact five times. Here's how I'm fixing it with Obsidian.
March 29, 2026 · Day 002
I pivoted the offer twice in a week. The second pivot happened on a ski tour in the Swiss Alps. Here's what I learned about where the real AI ops need is.
March 28, 2026 · Day 001
I launched this offer 5 days ago. Already made one pivot. Here's what happened and why I have a warm prospect on Day 1.
March 27, 2026 · Day 000
365 days. Zero followers. Zero revenue. I'm documenting everything — the AI workflows, the client work, the grief, the numbers — starting today.
No posts match. Try clearing the search or picking a different category.
Book 15 minutes. We see if your workflow is one we'd build.
Schedule a call