I have a team
Your engineers ship slowly and unevenly. They adopted AI and nothing changed. I move the team you already pay into AI-native workflows with guardrails, fix how you ship, and make the numbers move.
How I move a team into that 23% →A hands-on CTO for founders and teams who need to ship.
I'm Anthony Michel. I give founders and small teams the senior technical leadership a full-time CTO brings, without the full-time hire: the right product decisions, faster delivery, and AI that shows up in the numbers. Hong Kong, working worldwide.
Two situations, one job
Almost everyone has adopted AI, but only 23% see a measurable financial return (Deloitte–HKU, 2026). Wherever you start, the job is the same: make the technology pay.
I don't have a team yet
You have the idea and the budget, not the product or the technical leadership. I build the first version myself and make the technology calls a technical co-founder would, so you reach a real product without betting the runway on an agency.
How I build a first product →The calls I usually get
Different companies, same four conversations. If one of these sounds like yours, I can help.
“I have the vision and the budget. I don't have a CTO.”
You're about to bet your runway on an agency quote you can't evaluate, or a stack a friend recommended. I give you a technical co-founder's judgment without the equity or the full-time salary: what to build, what to skip, and who should build it.
“The product exists, but shipping got slow.”
Releases hurt, bugs come back, estimates drift, and nobody can say why. I've fixed this at a legaltech scale-up (production issues halved, releases every two weeks) and I fix it the same way every time: process, quality gates, and honest priorities, installed alongside your team, not preached at them.
“My developers either ignore AI or drown in it.”
Half the team pastes from chatbots with no review, the other half refuses to touch it, and nobody measures anything. There is a working method between the hype and the fear: tooling, agent workflows, guardrails, and a new definition of done. I've run this transition with a real production team, not in a slide deck.
“I need a CTO in the room.”
Investor due diligence next month. A vendor contract you can't price. A hire you can't interview alone. I stand on your side of the table: I've sat through funding rounds, government grant panels, and enough vendor pitches to know where the bodies are buried.
When to call someone else
The fastest way to trust me is to hear what I won't pretend to be.
I'm not an MNC-scale CTO.
If you need someone to run a 300-person org inside a multinational, that isn't me.
I'm not a CIO.
Internal IT, enterprise procurement, and corporate infrastructure are a different job.
I don't run large engineering organisations.
Above roughly 50 engineers you need a full-time leader, and I'll say so.
I'm not a strategy deck.
If you want slides and nobody on the keyboard, we'll both be disappointed.
If one of those is what you need, tell me on the first call and I'll point you to someone better. Saying no fast is part of the service.
What I do
Four ways to work with me. Same person, same standard of care, whether you get two hours a week of judgment or a quarter of hands-on building.
Flagship program
AI-Native Engineering
Your developers, with AI agents doing the heavy lifting: that is where software is going, and most teams get there by accident, badly. CodeRabbit's State of AI vs Human Code Generation Report found AI-authored pull requests had about 1.7x more issues than human-only ones (470 open-source pull requests, December 2025), so the guardrails are the point, not an afterthought. I run a structured transition I have already executed with a production team of six: audit to daily agent workflows, security included, in about a quarter.
- Audit and baseline. How your team builds today, where AI actually helps, what to measure.
- Tooling and repo setup. Agent-ready repositories, rules, skills, review gates. The boring plumbing that makes AI output trustworthy.
- Working patterns. Training on agent workflows: planning, delegation, validation, and when not to use AI at all.
- Security and governance. Prompt injection, data leakage, supply chain risks. Covered before they cost you.
Engineering acceleration
I embed with the team you already pay: delivery rhythm, release automation, quality gates, code review culture, and mentoring for your future leads. Measured by shipped features and fewer production incidents, not by reports.
For: SMEs and startups whose team ships slower than it should
Start with an engineering audit →MVP & proof-of-concept builds
From idea to working product. I architect it and write the code myself, with AI agents multiplying the output. You get software your future team can inherit, not a prototype you'll pay to rewrite.
For: pre-seed and seed founders before the first tech hire
MVP and POC builds →Product & technology strategy
A standing advisory seat: roadmap, architecture, build-versus-buy, hiring plans, vendor management, and preparation for investor or board conversations. The expensive mistakes get caught while they're still cheap.
For: founders and CEOs who need CTO judgment on tap
Including build-versus-buy calls →Also on the table, without top billing: engineering OKRs and KPIs, org design and recruiting, innovation on an existing product, and vendor and project management.
How an engagement runs
No discovery theater, no lock-in. Three steps, and you keep something useful even if we stop after the second.
-
Intro call
Thirty minutes on your situation. If I'm not the right person, I'll say so and point you somewhere better.
Free, no deck required.
-
Diagnostic
One to two weeks. I read the code, meet the team, and map the product against the business. You get a written assessment with priorities and a plan.
Fixed fee. The plan is yours either way.
-
Engagement
Advisory hours, embedded days, or a scoped build, at a fixed monthly rate sized to the plan. Monthly rolling, one month's notice.
Most clients start small and expand.
Recent work
Real clients, real production systems. Numbers first, adjectives never.
28one · quality-control SaaS for enterprise brands · fractional CTO since 2025
Rebuilt the platform, then rebuilt how the team builds
Twelve siloed client databases merged into one shared platform, with a full backend rebuild and a modernized inspection app. Then the deeper change: their six engineers in India now work AI-native, with agent tooling, review gates, and a training program I wrote for them. AI ships in the product too: supply-chain disruption intelligence, automated sizechart extraction, and rebuilt reporting.
12 → 1
databases unified, zero data loss
6
engineers moved to AI-native workflows
3
AI systems shipped into the product
FanForce · pre-seed social rewards startup · hands-on product & technology lead since 2024
From zero to 20,000 users with a two-person team
Product vision, architecture, and most of the code across three major product iterations. The core is an AI evaluation pipeline that scored roughly one million tweets and ten million replies across 2.5 million accounts for quality and community loyalty, feeding a multichain reward distribution system.
20k+
users on the platform
1M
tweets scored by the AI pipeline
3
product iterations shipped
Earlier: Zegal (legaltech) · Yicom (co-founder) · Zwoop · Société Générale
The same job at every scale
VP of Engineering at Zegal: a remote team of 18, production issues cut in half, releases every two weeks, three squads running without me in the room. Co-founder CTO at Yicom: concept to launch, HK$2M in government funding, a remote ASEAN team hired at 70% lower cost. Before Hong Kong: FX options trading systems at Société Générale in Paris, where "it works on my machine" was never an acceptable answer.
−50%
production issues at Zegal
18
engineers led across time zones
HK$2M
government funding secured at Yicom
As the first company hire, Anthony has been instrumental to the success of Zwoop, which has raised multiple multi-million USD rounds of funding. He shows constant dedication in mentoring junior colleagues and demonstrated the strongest work ethic.
Anthony understands the whole system very well and provides guidance on improving in all aspects. He did a great job planning and prioritizing. A strongly recommended teammate.
What hiring for this actually costs
Three honest ways to get senior technical leadership, and what each one really costs you.
A full-time CTO in Hong Kong
Runs HK$1.2M to HK$3M a year in salary alone (Robert Half, 2026 Hong Kong Salary Guide). Add MPF, equity, a recruiter's fee, and the several months it takes to hire a senior leader. And that assumes the hire is right.
A dev agency
Builds what you specify, but does not own the decision. You still need someone senior to tell it what to build and to check what it ships.
Me, fractional
Fixed monthly, one month's notice. I start in about two weeks, make the calls a CTO makes, and still write the code. Sized after the diagnostic, never billed by the hour for thinking.
The full breakdown, and when a full-time hire is genuinely the right call, is on how a fractional CTO is priced. Hiring in Hong Kong specifically? Here is what a fractional CTO in Hong Kong looks like locally.
Questions founders actually ask
What does this cost?
A fixed monthly rate, sized after the diagnostic, so you're never billed by the hour for thinking. For reference, a full-time CTO in Hong Kong runs HK$1.2M to HK$3M a year in salary alone (Robert Half, 2026 Hong Kong Salary Guide), before MPF, equity, and a recruiter fee. A fractional engagement is a fraction of that, scales down when you need less, and stops with one month's notice. Full breakdown: how a fractional CTO is priced.
Why not just hire someone full-time?
Sometimes you should, and I'll tell you when. But a full-time CTO in Hong Kong runs HK$1.2M to HK$3M a year in salary alone (Robert Half, 2026 Hong Kong Salary Guide), takes several months to hire, and is a bet you can't easily undo. I'm fixed monthly, I start in about two weeks, and you can stop with a month's notice. When you outgrow me into a full-time hire, I help you make it.
I don't have a team yet. Is that a problem?
No. That's the founder path: I build the first product myself, AI-agent augmented, and make the technology decisions a technical co-founder would, then help you decide who to hire and when. If you already have a team, the job becomes making them ship.
How much of your time do we get?
Depends on the engagement: advisory clients get a weekly cadence plus async access; embedded engagements run one to three days a week. I keep a maximum of three active clients so that each one gets real attention, not calendar scraps.
Do you actually write code?
Yes, most days. MVP builds are largely my own hands (with AI agents multiplying the output). In team engagements I code enough to review with authority: architecture spikes, hard bugs, and the example-setting pull requests.
We're not in Hong Kong. Does that matter?
No. I'm in Hong Kong (UTC+8), the same timezone as Singapore and most of Asia, with European mornings and US evenings in comfortable overlap. I've run teams across India, ASEAN, and Europe for years; remote-first is how I've worked since 2019. For clients in Hong Kong, Shenzhen, or Singapore I'm happy to be in the room.
What does "AI-native" actually mean?
A team where AI agents do the mechanical work (drafting code, tests, migrations, documentation) inside guardrails that humans design and enforce: rules, review gates, security policies, and measurement. Not "everyone has a chatbot subscription." The difference shows up in shipping speed and in how few juniors' mistakes reach production.
We already have developers, even a lead. Do you replace them?
Never. The job is to make your existing team stronger: your lead gets a mentor and a sparring partner, your developers get processes that let them ship, and you eventually get an internal CTO promoted from within. I work myself out of the org chart by design.
How fast can you start?
The intro call within a week, the diagnostic usually within two. If I'm at my three-client limit I'll tell you on the call and give you a date.
Tell me where it hurts
Two sentences about your company and the problem is enough. I read every message myself and reply within one business day, usually with a few questions and a suggested time to talk.
Prefer LinkedIn?
Message me on LinkedIn
Hong Kong · UTC+8 (same as Singapore) · EU mornings and US evenings work fine
Bringing an AI assistant?
Point it at anthmichel.com/llms.txt: what I do in machine-readable form, plus how your agent can send me an introduction on your behalf, no form-filling needed.