01Start here

Pick your starting point

Three ways in, one operating standard. Most teams prove it in one day, then scale. Every bootcamp ends with a written capability debrief and our recommendation on whether to go further, including “don't.”

1 day Start here

One-day agentic developer bootcamp

One concentrated day inside your approved setup. The operating loop of plan, direct, review, and verify, plus repository instructions (CLAUDE.md / AGENTS.md) drafted against one of your real repos, and your team conventions v1.

CohortUp to 40 developers Leaves withWorking setups, drafted agent files, conventions v1, debrief Best forProving trained agent use to a sceptical team, fast

Fixed fee per cohort · quoted at scoping Book the day →

2–3 days Most chosen

Deep-dive team bootcamp

Two instructor-led days across all eight modules: agent files, plan mode, large codebases, multi-agent and MCP, guardrails, cost. Optional applied day on tasks from your own roadmap.

Cohort10 to 40 engineers; repeatable across teams Leaves withTeam standards, review discipline, multi-agent practice, debrief Best forTeams defining shared conventions properly

Per cohort · quoted after scoping Scope the bootcamp →

Per seat

Enterprise Academy licence

The full self-paced Academy bundle, including the Agentic Development course, with role-based paths, live kickoff, monthly mentorship clinics, office hours, and leadership analytics. The baseline for hundreds to thousands of developers.

ScaleHundreds to thousands of seats Leaves withOne shared foundation, progress analytics, lifetime updates Best forConsistent baseline across a large org

Per seat · volume bands Discuss seats →

Also available: 1:1 coaching for the lead writing your conventions. Cost is driven by participant count, live days, and customisation depth; an indicative quote accompanies the scoping agenda, before you commit to anything.

02The system

Two steps: use AI to build software, then build AI into it

Step 1 makes every developer effective with coding agents and reveals who should go further. Step 2 converts your strongest builders into AI engineers. You are on the Step 1 page.

	Step 1 · Agentic developers ← this page	Step 2 · AI engineers
Who	Every software developer, no AI background needed	Selected builders, typically the top 10–20%
They learn	To plan, direct, review, and verify coding-agent work under one standard	To design, evaluate, ship, and maintain production LLM systems
Outcome	Ship faster on today's roadmap; agent code passes the same bar as human code	New products and AI systems; the capability you cannot hire
Format	1-day or 2–3-day bootcamp · per-seat licence	Part-time cohort waves · senior residency · per-seat licence
When	Now: it pays back on work already in flight	After Step 1: the debrief names who's ready

Explore Step 2: AI engineer conversion → Same instructors, same standard, one scoping call covers both.

03Why it works

Seats are everywhere. Skill is not.

Your developers already spend a median of two hours a day with AI. The field data is blunt about what separates the teams that win from the teams that stall.

AI is an amplifier

DORA's 2025 study of ~5,000 professionals: 90% of developers now use AI, a median of two hours a day, and over 80% report productivity gains, yet delivery instability rises unless the workflow is redesigned. Seven organisational capabilities decide whether individual gains ever reach the org. Amplifiers make disciplined teams faster and undisciplined teams slower.

DORA 2025 · Google Cloud

Untrained use backfires

METR found experienced developers 19% slower with AI on familiar code, while believing they were 20% faster. Self-assessment is miscalibrated, so self-learning doesn't converge. 66% of developers fight “almost right” output; 45% say debugging AI code takes longer; only around one in six report significant productivity gains.

METR 2025 · Stack Overflow 2025

The gap is trainable

Inside Anthropic, Claude now authors 80%+ of merged code and code merged per engineer is up 8x since 2024. Experts drive ~12 agent actions and 3,200 words of output per prompt versus ~5 and 600 for novices. The difference is task framing, context, and verification: exactly what we train. BCG finds 5+ hours of structured training with coaching is what turns tool access into regular, effective use.

Anthropic 2026 · BCG 2025

Waiting compounds the gap

Google's CEO says AI now writes around three-quarters of the company's new code. Microsoft has told staff AI use is “no longer optional.” Mandates without enablement backfire; the winners build capability instead of compulsion, and the first trained cohort sets the standard every later cohort inherits.

Alphabet 2026 · Microsoft 2025

04Straight answers

The questions your CFO will ask

“Our devs can self-learn from docs and YouTube.”

They've had three years. Only ~16% of developers report significant productivity gains from AI, and METR shows devs believed they were 20% faster while measuring 19% slower: miscalibrated self-assessment means self-learning doesn't converge. Structured training with coaching is what moves usage (BCG 2025).

“The tools change too fast; training goes stale.”

The durable skills are tool-agnostic: planning, context engineering, review discipline, verification, team conventions. They transfer across Claude Code, Codex, and whatever ships next. We re-train on the frontier monthly because we ship with it, and Academy content carries lifetime updates.

“We can't take engineers off delivery.”

The entry format is one day. Your engineers already spend a median two hours a day with AI, unmanaged. The bootcamp redirects time already being spent, and applied exercises run on tasks from your own roadmap, not toy repos.

“Security won't allow it.”

Everything runs in your approved environment: your tool surface, your permissions, no production access required, NDAs standard. We've delivered inside Europol. And the biggest security risk is untrained use; review boundaries and guardrails are half the curriculum.

“We already pay for seats, so we're covered.”

Seats are access, not capability. DORA's finding is that AI amplifies the engineering discipline you already have; licences amplify nothing on their own. The 8x gains show up where trained use meets shared standards.

“How do we know it worked?”

We capture a baseline in scoping and close with a written capability debrief: what changed in how your engineers work, the conventions your team keeps, and the metrics to watch: throughput, review load, spend per merged change. Scale to more cohorts only if the debrief earns it.

05Delivered in the field

Trained where mistakes aren't an option

EuropolLaw enforcement · EU

Three-day security-first bootcamp: technical investigation teams went from cautious LLM users to confident builders of secure, reviewable intelligence pipelines.

Outcome: special agents deploying AI across operational intelligence workflows.

Major US public library systemPublic sector · USA

Four-day, two-track programme for engineers and operations staff, backed by 43 hands-on notebooks and a hackathon on real pain points.

Outcome: adoption from foundations to shipped build, owned by the library's own teams, not a vendor.

Global banks & leading fundsFinance

Trader training at a leading global bank; live Claude Cowork sessions for investment teams at leading funds; enterprise bootcamps tailored to each firm's stack.

Outcome: desk-level AI workflows in regulated, latency-sensitive environments.

“Since the training, I’ve built two local apps using your vibe-coding approach. With every Gemini release I test vibe coding capabilities and have been unimpressed. Your approach is brilliant… We’re going to make it our primary developer model.”

VP of IT

Major US public library system

“Excellent in-depth handling of trade-offs in evaluating and deploying agent-based solutions. I've used a lot of the advice in building our enterprise practices to agent use in my organisation.”

Cathal Curtin

Student, Agent Engineering

06Who delivers

No training faculty. Practising engineers.

The people in the room ship client systems with coding agents every working week. The material is the conventions, failure modes, and review habits that survived real codebases.

Louie Peters Co-CEO · Instructor

Co-author, Building LLMs for Production. Newsletter read by 200k+ practitioners. Ex-J.P. Morgan VP.

Louis-François Bouchard CTO · Instructor

Co-author of the book; 210k followers; ex-PhD at MILA, Yoshua Bengio's lab.

Fabio Chiusano Senior AI Engineer

NLP lead; ships the client systems the curriculum is drawn from.

Samridhi Vaid Machine Learning Engineer

Ships production multimodal AI, RAG, and agent systems; M.Sc. in Computing Science, University of Alberta.

500k+AI practitioners taught since 2019

15+in-house AI engineers shipping client systems

10k+copies of Building LLMs for Production sold

21+AI applications shipped to production

07How it runs

Low commitment in, evidence out

01

Scoping call · 30 min

Teams, approved tools, working patterns, and priorities, discussed with the instructors. You get a written plan and an indicative quote within one business day.

02

Baseline

We capture where the team actually is (usage patterns, conventions, metrics worth watching) so the debrief measures change, not vibes.

03

The day(s)

Live workshops inside your approved environment. No surprise tooling, no unapproved repository access, acceptance and merge stay with your engineers.

04

Debrief & decision

A written capability debrief: what changed, which conventions to keep, who's ready for Step 2, and whether further cohorts are worth it. Scale only if it earns it.

+Start with one day. Prove it on one cohort before you commit to a programme.

+Your tools, your rules. Scoped to the tool surface your organisation has approved. No production access required.

+Procurement-ready. NDAs on request, PO and invoicing supported, references in scoping.

+We'll say if it's not for you. If a cheaper format fits, or training isn't the right tool, the instructors tell you on the call.

One day. Up to 40 developers. Your approved tools.

Book the scoping call; get the plan and quote within one business day.

Book a scoping call

08Logistics

Frequently asked questions

Does the programme require one specific coding agent?

No. We scope workshops around the approved tool surface relevant to your team (Claude Code, Codex, or both) and explain differences where practices are tool-specific, including how to standardise across them.

What exactly happens on the applied day?

Only with agreement: after sign-off on approved repositories or prepared tasks, permissions, and review expectations, the cohort works agreed tasks from your roadmap under instructor supervision. It's applied learning, never production delivery. One signature session: an hour where engineers may only plan, direct, and review while the agent writes everything; the diffs are then reviewed against your own bar.

What are the prerequisites?

Software engineering experience and access to your approved coding-agent tooling. No AI background needed; that's the point of Step 1.

What comes after?

The debrief names next steps: further cohorts, team conventions to keep, and the developers ready for Step 2: AI engineer conversion. Beyond training, the same engineers build custom AI systems and provide fractional AI leadership.

09Next step

Book a scoping call

Tell us about your developer cohorts, approved tools, and priorities. We reply within one business day with a short scoping agenda and an indicative quote, and you'll talk with the instructors who deliver the workshops.

1On the call: teams, stacks, baseline, and which format fits. 30 minutes, with instructors.

2You leave with: a written plan, an indicative quote, and a measurement approach.

3No pressure: if one day is enough, we say so.

Every enquiry is read by the Towards AI team. Prefer email? louis@towardsai.net