Make AI a team capability, not a few people’s habit.
Your team already uses AI. We turn those scattered, personal experiments into dependable, shared, documented practice, built on your real codebase, on-site or hybrid in Vietnam.
Right now, AI on your team probably looks like this.
A few early adopters got fast. The rest didn’t.
The gains live in a handful of people, not in the team. When they’re busy or they leave, the capability leaves with them.
Almost nothing is written down.
Prompts, context and hard-won tricks stay in private chats and personal habits. None of it is shared, versioned, or reusable.
Quality is uneven.
AI-written code ships at different standards depending on who used it and how carefully they reviewed it. Trust varies by author.
No one can say what’s actually working.
You feel movement, but you can’t point to it, repeat it, or build on it. There is no shared standard and no honest measure.
None of this is a tooling problem. It’s a practice problem. That’s exactly what we fix.
What changes
From scattered experiments to one dependable practice.
Today
Individual hacks and private prompts
Context and tricks no one wrote down
Quality that changes with the author
“Trust me” reviews
Gains stuck with a few people
With SOPAI
A shared, version-controlled practice
Context and prompt libraries the team owns
Review calibrated to real risk
Quality you can actually depend on
AI as a capability the whole team carries
Same tools. A completely different operating standard.
01The method
The method
Five pillars that make AI dependable on a real team.
Not a catalogue of tips. A structured discipline we build into how your team already works, taught on your codebase, and yours to keep. No lock-in.
01
Spec-driven work
Writing specifications an AI can actually execute — a learned skill: context efficiency, no redundancy, agent-compatible.
02
AI-assisted testing & quality
The review and evaluation discipline that makes AI-written code trustworthy: human review by risk, and tiered evals on your real features. Generating tests with AI is part of the toolkit, an emerging practice we teach honestly, with humans owning the verdict.
03
Team AI coordination
Shared context files, prompt libraries, conventions and workflows so multiple people and their agents work coherently — without chaos.
04
Security & review discipline
What to send (and never send) to an AI, and review by risk level — 100% on sensitive code, spot-checks on boilerplate.
05
Cost & automation control
Choosing the right model for the task, managing API cost, and deciding where on the automation spectrum to sit.
The programme
A 90-day arc, from first wins to team autonomy.
One continuous arc on your own work. Roughly three months, one live session a week, guided practice in between, and a follow-up checkpoint. Four phases, each closing on something real your team shipped.
01 Weeks 1 to 2
Foundations
We set an honest baseline of where AI helps and where it hurts, get the fundamentals of context and prompting right on your repo, and start a shared prompt library.
Real quick wins shipped on your own work
A shared, version-controlled prompt library
An honest baseline of what AI is good and bad at
02 Weeks 3 to 5
Guided expansion
The craft deepens. You spec a real upcoming feature with a human gate at each stage, and bring evaluation discipline to your own code.
A real feature spec’d with a human gate at each stage
Tiered evals standing on your own code
An agent-PR red-flag check on recent pull requests
03 Weeks 6 to 9
Team integration
The practice becomes the team’s, not a few individuals’. You set shared standards and review tiers, and bring them into CI.
Your team’s AGENTS.md, written by humans
One stable rule promoted from prose to lint to CI
Review tiers by risk, and a shared way of working across repos
04 Weeks 10 to 13
Autonomy
We make it stick without us. You take over the measurement, run a retrospective on the method your team adapted, and name internal champions.
Tailored to your team, never pulled from a catalogue.
Every engagement is built on your stack, your maturity and your real projects.
01
Pre-assessment
We watch how your team really works, your stack, and the AI you already use, and capture an honest baseline before we begin.
Questionnaires
On-site observation
Baseline captured
02
Tailored programme
We design the engagement around your stack, your maturity and your goals. Never a generic catalogue, never a template.
Your stack
Your maturity
Your goals
03
Hands-on
Every exercise runs on your team’s actual projects. The deliverables are your own artefacts, never toy examples.
Your real projects
Your own artefacts
Never toy examples
04
Measured
We baseline at the start, run on your real work, measure the real change, then hand you the framework so you keep measuring after we leave.
Baseline, run, measure
You keep the framework
No lock-in. You own every artefact, library and SOP we build together.
02Measurement
Measurement
We measure the real change. And we never invent the number.
The market is full of borrowed statistics. A number that didn’t come from your team means nothing for your team, so we don’t hand you one. We baseline your team, run on your real work, measure what actually changes in your workflow, and you keep the framework. Honesty isn’t a disclaimer here. It’s the whole point.
What we refuse to sell you
+40% faster
Save 10 hours a week
10x your developers
What we do instead
01 Baseline your team
Captured before we begin, on your real workflow.
02 Run on your real work
Every exercise on your actual projects, never toy examples.
03 Measure the real change
Grounded in what happens in your workflow, not how fast people feel.
04 You keep the framework
You go on measuring honestly long after we leave.
“
Partner, not professor. No fabricated metrics.
03Formats
Start where it fits.
From a first taste to a full transformation arc. Each one leads naturally to the next.
01
Discovery session
Start here
Half a day: introduction, a diagnosis of your team, and quick wins. The door-opener.
02
Intensive day
One day: the core methodology plus a hands-on workshop on your team's real work.
03
Multi-session programme
Over several weeks (a ~90-day arc with follow-up): the most transformative — foundations to autonomous scaling.
Who it’s for
Built for English-speaking teams in Vietnam.
Foreign-led and Western-culture companies
FDI subsidiaries, regional offices and expat-founded startups with English-speaking management, in HCMC or across Vietnam.
Vietnamese companies with English-speaking management
Tech-oriented startups and SMEs whose developers are self-learning AI without a shared company framework.
Developer teams are the sharpest fit, typically 10 to 100 people already using AI unevenly. The same method extends to adjacent functions: data, product, operations and leadership.
Optional add-on
For the people who have to model it.
Adoption that lasts needs the people who lead it to practise it. As an optional add-on, we work with your leads, managers and the programme’s sponsor on the part only they can do. Sponsorship isn’t modelling, and modelling is what makes the rest stick.
Use AI on real work
Leaders visibly using AI on their own real work, not just endorsing it.
Clarity-led delegation
Making goal-clarity the heart of how they delegate, to people and to AI alike.
Back the champions
Backing the team’s internal champions so the practice keeps spreading.
“
We don’t hand your team a tool and wish them luck. We sit beside your people, learn how the work really gets done, and build dependable AI practice into it.
SOPAI · Practitioner to practitioner · Ho Chi Minh City
Ready when you are
Ready to make AI actually work?
Book a discovery session for your team, or explore the packages for yourself.
We use cookies to understand how the site is used and improve it. They run only if you accept, never for advertising. You can change your mind anytime. Privacy Policy
