Haider Ali
Founder & CEO
Leads the company, talks to every client, approves every plan and signs off every deliverable.
Company · roster checked October 2026
Led by a human. Built by a team of AI agents.
Pixel & Oak is run by founder and CEO Haider Ali. The work is done by 22 AI agents: 21 specialists in four divisions and the Conductor, an orchestration agent that hands out the work. Haider checks everything before it ships.
- Human CEO
- 1
- AI agents
- 22
- Divisions
- 4
- CEO replies within
- 24h
At the centre
Haider Ali
Founder & CEO · the only human on the team
Leads the company, talks to every client, approves every plan and signs off every deliverable.
Point at any agent to see what it does.- Claude agent
- Other AI model
- Human
Leadership
A note from the CEO
Pixel & Oak is built on one idea: a single accountable person, and a team that never runs out of hands.
I lead the company and I talk to every client myself. The work is carried out by a team of AI agents that I direct. They plan, build, test, review and measure in parallel, and nothing they produce ships until I have checked it and signed it off.
That is how one person can run a studio that delivers websites, software, films, marketing and audits, with the work done in parallel instead of one task at a time, at prices agreed before it starts. Write to us and you will hear back from me within 24 hours.
Haider Ali, Founder & CEO
- Building for the web since
- 2017
- Production sites run
- 36, on Vercel and Cloudflare (October 2026)
- Everyday stack
- Next.js, React, TypeScript, WordPress, Shopify, Supabase
- Own products
- Restaurant POS, Funeral Kit, CrewCalculator, free browser games
- Open source
- 6 Next.js templates and an explainer video engine
- Replies
- Within 24 hours, from the CEO
The roster
22 AI agents, each set up for one job, with the model it runs on. Codenames, not people: Haider is the only human here.
00 · Office of the CEO
Direction and coordination
Sets the direction, agrees every price and signs off every deliverable. The Conductor turns that direction into work for the divisions.
Human
Conductor
Orchestration agent
Turns the CEO’s brief into a plan, splits it into tasks, hands each one to the right specialist and runs them in parallel, then merges the results into one report.
- Task planning
- Delegation
- Parallel sessions
- Status reports
01 · 5 agents
Engineering
Plans and writes the software: websites, web apps, integrations and deploys.
Architect
Systems planning agent
Designs the build before any code is written: data models, page structure, integrations and the trade-offs between them.
- System design
- Data models
- Implementation plans
- Trade-offs
Builder
Full-stack engineering agent
Writes the production code on a private branch: Next.js and React front ends, APIs and databases, Shopify and WordPress work.
- Next.js & React
- TypeScript & Node
- Supabase & Postgres
- Shopify & WordPress
Scout
Codebase exploration agent
Reads large and unfamiliar codebases fast: finds every caller, every setting and every place a change will land before anyone edits a line.
- Code search
- Dependency maps
- Legacy code
- Impact analysis
Relay
Integrations & automation agent
Connects the systems a business already uses (payments, CRM, email, spreadsheets, ad platforms) through APIs, webhooks and n8n workflows.
- REST & GraphQL APIs
- Webhooks
- n8n workflows
- MCP connectors
Release
Deployment agent
Builds the site, compares every changed file with what is live, deploys only on the CEO’s go and keeps a rollback ready.
- Cloudflare & Vercel
- Build pipelines
- Pre-deploy diffs
- Rollbacks
02 · 5 agents
Quality & Security
Checks the work before the CEO sees it: code review, security, real-browser tests, speed and accessibility.
Reviewer
Code review agent
Reviews each change for bugs, edge cases and needless complexity, and confirms every finding before reporting it.
- Correctness
- Edge cases
- Simplification
- Regression risk
Sentinel
Security review agent
Looks for the ways a site gets broken into: injection, broken login, exposed secrets, unsafe uploads and server-side request forgery.
- OWASP Top 10
- Auth & sessions
- Secrets
- Dependency risk
Tester
QA agent
Tests in a real browser at phone, tablet and desktop widths: no sideways scroll, no console errors, no broken links, every form sent.
- Real-browser tests
- Responsive checks
- Link checks
- Component tests
Pace
Performance agent
Measures load speed the way Google does and removes what slows a page down: heavy scripts, oversized images, fonts that block the first paint.
- Core Web Vitals
- Largest paint
- Bundle size
- Caching
Access
Accessibility agent
Checks that everyone can use the site: keyboard navigation, screen readers, colour contrast and focus order.
- WCAG 2.2
- Keyboard & focus
- Contrast
- ARIA
03 · 4 agents
Design & Film
Makes what people see and hear: interfaces, artwork, launch films and narration.
Designer
Interface design agent
Sets the art direction (type, layout, colour and components) for the customers you want, not from a template.
- Art direction
- Typography
- Design systems
- Responsive layout
Lens
Image agent
Generates the photograph or illustration, then sets the lettering locally, so text on every image is sharp and spelled right.
- Social cards
- Hero images
- Thumbnails
- Ad creative
Director
Film & motion agent
Builds launch films and explainers in code from real product screens, with original music, cut for 16:9, 9:16 and 1:1.
- Launch films
- Motion graphics
- Explainers
- Short-form cuts
Voice
Narration agent
Records the voice-over and times every word, so captions and motion land exactly on the narration.
- Voice-over
- Word timing
- Captions
- Audio clean-up
04 · 7 agents
Growth & Marketing
Finds the customers: research, SEO and AI search, content, ads, social and analytics.
Hire this division: AI digital marketing · AI SEO · AI search (GEO and AEO) · SEO content writing · Google and Meta ads · Social media · Video ads · Email marketing · Competitor analysis
Researcher
Deep research agent
Researches markets, competitors and options across many sources at once, and reports with every claim sourced and dated.
- Market research
- Competitor analysis
- Pricing research
- Source checks
Signal
SEO & AI search agent
Chooses keywords from real search data, checks who already ranks, and builds pages Google can read and AI assistants can quote.
- Keyword research
- Technical SEO
- Structured data
- AI search
Writer
Content agent
Drafts landing pages, articles and emails in plain English from the research, for the CEO to edit and approve.
- Landing pages
- Articles
- Email sequences
- Plain English
Reach
Paid media agent
Builds and tunes ad campaigns (keywords and negatives, audiences, conversion tracking, creative tests) and spends only what the client approves.
- Google Ads
- Meta Ads
- Conversion tracking
- Creative tests
Herald
Social media agent
Turns each launch, film and article into posts and reels for Facebook, Instagram, LinkedIn and YouTube, and publishes each one only on the CEO’s OK.
- Reels & Shorts
- Captions
- Scheduling
- Cross-posting
Analyst
Analytics agent
Reads the numbers that matter (traffic, rankings, conversions), separates real demand from noise, and charts it.
- GA4
- Search Console
- Dashboards
- Reporting
Concierge
Website assistant
Answers visitors on this website from facts the CEO approved, and passes their messages straight to the CEO.
- Visitor questions
- Prices & plans
- Lead capture
- Hand-off to the CEO
How a brief moves through the team
Six steps, two of them human. Agents do the planning, making and checking in parallel; the CEO opens the project and closes it.
-
01 · CEO
Brief
Haider reads your brief, asks what is missing, and agrees the scope and a fixed price with you in writing.
-
02 · 4 agents
Plan
The Conductor turns the brief into tasks. Architect plans the build, Scout maps any existing code, Researcher checks the market.
-
03 · 7 agents
Make
Specialists work in parallel on a private branch and a preview link: code, design, images, film and words.
-
04 · 5 agents
Check
Separate agents review the code, test it in a real browser, and measure security, speed and accessibility.
-
05 · CEO
Sign-off
Haider reviews the preview and every check report. Nothing ships without a yes.
-
06 · 5 agents
Ship & grow
Release deploys and verifies it live. Signal, Analyst, Reach and Herald take over the work of being found.
The rules every agent works under
Not aspirations: this is how the studio already runs, and why a team of AI agents is safe to hire.
A human answers for every result
Haider reads every brief, approves every plan and signs off every deliverable. The agents do the work; the CEO owns it.
Nothing reaches your live site without a person
Builds happen on a private branch and a preview link. Work that has to touch a live site, such as a WordPress fix or a malware clean-up, happens only on the CEO’s go, with a backup or rollback ready.
A second agent checks, then a person
Code is reviewed by a separate review agent and tested in a real browser before it reaches sign-off.
No spending or sending without approval
Agents never spend money, launch ads, or post and send messages on your behalf without an explicit OK for that action.
Every result we quote has a source
Traffic, rankings and other results we publish say where they came from and when. No invented results, reviews or rankings.
You own everything
Your code, your domain, your accounts and your content stay in your name.
The models behind the team
Most agents run on Claude. Where another model is better at a job, that agent uses it, and the roster says which. Checked October 2026.
| Model | Made by | Used for |
|---|---|---|
| Claude Opus 5.5 | Anthropic | leads engineering in Claude Code |
| Claude family | Anthropic | the specialist agents |
| Gemini 3.1 Flash Image | images | |
| GPT-5 nano | OpenAI | image checks |
| Seedance 2.0 | ByteDance | generated footage |
| Grok 4.3 | xAI | the website assistant |
| Kokoro | hexgrad, open weights | voice, run locally |
| Whisper | OpenAI, open weights | word timing, run locally |
Questions about the team
Who will I actually talk to?
Haider Ali, the founder and CEO. Every enquiry is answered by Haider within 24 hours, and the same person signs off your project.
Are the team members real people?
No. Haider is the only human at Pixel & Oak. The rest of the team are AI agents, each set up for one job, and this page lists the model each one runs on. We say so plainly because you should know who, and what, is building your project.
Which AI models does the team use?
Mostly Claude, made by Anthropic: Claude Opus 5.5 leads engineering in Claude Code and the specialist agents run on the Claude family. Images come from Gemini 3.1 Flash Image, checked by GPT-5 nano; the website assistant runs on Grok 4.3; narration uses Kokoro and Whisper, open-weight models run locally. Roster checked October 2026.
If AI does the work, why not use AI tools myself?
For small jobs, you should. What you get from us is the system around the agents: a plan before any code, review and real-browser tests before sign-off, a person who can tell when the output is wrong, and one accountable name on the result.
Is my project safe with AI agents?
Agents build in a private copy of your project wherever the job allows, and nothing changes on your live site without the CEO’s go and a backup or rollback ready. You own the repository, the domain and every account, so access stays in your hands.
What does it cost?
A fixed price for projects and a fixed monthly fee for ongoing work, published on each service page and agreed in writing before work starts. Running a team of agents is how those prices stay low.
Work with the team
Put 22 agents and one accountable CEO on your project.
Tell us what you need. Haider replies within 24 hours to agree the plan and a fixed price. After a website? You get a free homepage concept within 48 hours.
Claude is made by Anthropic, Gemini by Google, GPT-5 and Whisper by OpenAI, Grok by xAI, Seedance by ByteDance, and Kokoro is an open-weight model. Pixel & Oak is an independent studio and is not affiliated with or endorsed by any of them.